Hugging Face's logo

sujr
/
scaling_exp_125k_refGnd

PEFT
Safetensors
Model card Files Community
sujr commited on
Commit
8b47b4e
•
1 Parent(s): 03567d2

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. checkpoint-1200/README.md +202 -0
  2. checkpoint-1200/adapter_config.json +380 -0
  3. checkpoint-1200/adapter_model.safetensors +3 -0
  4. checkpoint-1200/latest +1 -0
  5. checkpoint-1200/qwen.tiktoken +0 -0
  6. checkpoint-1200/rng_state_0.pth +3 -0
  7. checkpoint-1200/rng_state_1.pth +3 -0
  8. checkpoint-1200/rng_state_2.pth +3 -0
  9. checkpoint-1200/rng_state_3.pth +3 -0
  10. checkpoint-1200/scheduler.pt +3 -0
  11. checkpoint-1200/special_tokens_map.json +3 -0
  12. checkpoint-1200/tokenizer_config.json +14 -0
  13. checkpoint-1200/trainer_state.json +873 -0
  14. checkpoint-1200/training_args.bin +3 -0
  15. checkpoint-1200/zero_to_fp32.py +587 -0
  16. checkpoint-1600/README.md +202 -0
  17. checkpoint-1600/adapter_config.json +380 -0
  18. checkpoint-1600/adapter_model.safetensors +3 -0
  19. checkpoint-1600/latest +1 -0
  20. checkpoint-1600/qwen.tiktoken +0 -0
  21. checkpoint-1600/rng_state_0.pth +3 -0
  22. checkpoint-1600/rng_state_1.pth +3 -0
  23. checkpoint-1600/rng_state_2.pth +3 -0
  24. checkpoint-1600/rng_state_3.pth +3 -0
  25. checkpoint-1600/scheduler.pt +3 -0
  26. checkpoint-1600/special_tokens_map.json +3 -0
  27. checkpoint-1600/tokenizer_config.json +14 -0
  28. checkpoint-1600/trainer_state.json +1153 -0
  29. checkpoint-1600/training_args.bin +3 -0
  30. checkpoint-1600/zero_to_fp32.py +587 -0
  31. checkpoint-2000/README.md +202 -0
  32. checkpoint-2000/adapter_config.json +380 -0
  33. checkpoint-2000/adapter_model.safetensors +3 -0
  34. checkpoint-2000/latest +1 -0
  35. checkpoint-2000/qwen.tiktoken +0 -0
  36. checkpoint-2000/rng_state_0.pth +3 -0
  37. checkpoint-2000/rng_state_1.pth +3 -0
  38. checkpoint-2000/rng_state_2.pth +3 -0
  39. checkpoint-2000/rng_state_3.pth +3 -0
  40. checkpoint-2000/scheduler.pt +3 -0
  41. checkpoint-2000/special_tokens_map.json +3 -0
  42. checkpoint-2000/tokenizer_config.json +14 -0
  43. checkpoint-2000/trainer_state.json +1433 -0
  44. checkpoint-2000/training_args.bin +3 -0
  45. checkpoint-2000/zero_to_fp32.py +587 -0
  46. checkpoint-2400/README.md +202 -0
  47. checkpoint-2400/adapter_config.json +380 -0
  48. checkpoint-2400/adapter_model.safetensors +3 -0
  49. checkpoint-2400/latest +1 -0
  50. checkpoint-2400/qwen.tiktoken +0 -0
checkpoint-1200/README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ base_model: Qwen/Qwen-VL-Chat
4
+ ---
5
+
6
+ # Model Card for Model ID
7
+
8
+ <!-- Provide a quick summary of what the model is/does. -->
9
+
10
+
11
+
12
+ ## Model Details
13
+
14
+ ### Model Description
15
+
16
+ <!-- Provide a longer summary of what this model is. -->
17
+
18
+
19
+
20
+ - **Developed by:** [More Information Needed]
21
+ - **Funded by [optional]:** [More Information Needed]
22
+ - **Shared by [optional]:** [More Information Needed]
23
+ - **Model type:** [More Information Needed]
24
+ - **Language(s) (NLP):** [More Information Needed]
25
+ - **License:** [More Information Needed]
26
+ - **Finetuned from model [optional]:** [More Information Needed]
27
+
28
+ ### Model Sources [optional]
29
+
30
+ <!-- Provide the basic links for the model. -->
31
+
32
+ - **Repository:** [More Information Needed]
33
+ - **Paper [optional]:** [More Information Needed]
34
+ - **Demo [optional]:** [More Information Needed]
35
+
36
+ ## Uses
37
+
38
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
+
40
+ ### Direct Use
41
+
42
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
+
44
+ [More Information Needed]
45
+
46
+ ### Downstream Use [optional]
47
+
48
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Out-of-Scope Use
53
+
54
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
+
56
+ [More Information Needed]
57
+
58
+ ## Bias, Risks, and Limitations
59
+
60
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ### Recommendations
65
+
66
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
+
68
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
+
70
+ ## How to Get Started with the Model
71
+
72
+ Use the code below to get started with the model.
73
+
74
+ [More Information Needed]
75
+
76
+ ## Training Details
77
+
78
+ ### Training Data
79
+
80
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
+
82
+ [More Information Needed]
83
+
84
+ ### Training Procedure
85
+
86
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
+
88
+ #### Preprocessing [optional]
89
+
90
+ [More Information Needed]
91
+
92
+
93
+ #### Training Hyperparameters
94
+
95
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
+
97
+ #### Speeds, Sizes, Times [optional]
98
+
99
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
+
101
+ [More Information Needed]
102
+
103
+ ## Evaluation
104
+
105
+ <!-- This section describes the evaluation protocols and provides the results. -->
106
+
107
+ ### Testing Data, Factors & Metrics
108
+
109
+ #### Testing Data
110
+
111
+ <!-- This should link to a Dataset Card if possible. -->
112
+
113
+ [More Information Needed]
114
+
115
+ #### Factors
116
+
117
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Metrics
122
+
123
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
+
125
+ [More Information Needed]
126
+
127
+ ### Results
128
+
129
+ [More Information Needed]
130
+
131
+ #### Summary
132
+
133
+
134
+
135
+ ## Model Examination [optional]
136
+
137
+ <!-- Relevant interpretability work for the model goes here -->
138
+
139
+ [More Information Needed]
140
+
141
+ ## Environmental Impact
142
+
143
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
+
145
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
+
147
+ - **Hardware Type:** [More Information Needed]
148
+ - **Hours used:** [More Information Needed]
149
+ - **Cloud Provider:** [More Information Needed]
150
+ - **Compute Region:** [More Information Needed]
151
+ - **Carbon Emitted:** [More Information Needed]
152
+
153
+ ## Technical Specifications [optional]
154
+
155
+ ### Model Architecture and Objective
156
+
157
+ [More Information Needed]
158
+
159
+ ### Compute Infrastructure
160
+
161
+ [More Information Needed]
162
+
163
+ #### Hardware
164
+
165
+ [More Information Needed]
166
+
167
+ #### Software
168
+
169
+ [More Information Needed]
170
+
171
+ ## Citation [optional]
172
+
173
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
+
175
+ **BibTeX:**
176
+
177
+ [More Information Needed]
178
+
179
+ **APA:**
180
+
181
+ [More Information Needed]
182
+
183
+ ## Glossary [optional]
184
+
185
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
+
187
+ [More Information Needed]
188
+
189
+ ## More Information [optional]
190
+
191
+ [More Information Needed]
192
+
193
+ ## Model Card Authors [optional]
194
+
195
+ [More Information Needed]
196
+
197
+ ## Model Card Contact
198
+
199
+ [More Information Needed]
200
+ ### Framework versions
201
+
202
+ - PEFT 0.10.0
checkpoint-1200/adapter_config.json ADDED
@@ -0,0 +1,380 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "Qwen/Qwen-VL-Chat",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layer_replication": null,
10
+ "layers_pattern": null,
11
+ "layers_to_transform": null,
12
+ "loftq_config": {},
13
+ "lora_alpha": 16,
14
+ "lora_dropout": 0.05,
15
+ "megatron_config": null,
16
+ "megatron_core": "megatron.core",
17
+ "modules_to_save": null,
18
+ "peft_type": "LORA",
19
+ "r": 64,
20
+ "rank_pattern": {},
21
+ "revision": null,
22
+ "target_modules": [
23
+ "transformer.visual.transformer.resblocks.32.mlp.c_fc",
24
+ "transformer.visual.transformer.resblocks.8.attn.in_proj",
25
+ "transformer.visual.transformer.resblocks.16.attn.in_proj",
26
+ "transformer.visual.transformer.resblocks.41.mlp.c_proj",
27
+ "transformer.visual.transformer.resblocks.31.mlp.c_proj",
28
+ "transformer.visual.transformer.resblocks.14.mlp.c_proj",
29
+ "transformer.h.14.mlp.w2",
30
+ "transformer.visual.transformer.resblocks.39.mlp.c_fc",
31
+ "transformer.visual.transformer.resblocks.40.mlp.c_proj",
32
+ "transformer.h.0.attn.c_attn",
33
+ "transformer.visual.conv1",
34
+ "transformer.visual.transformer.resblocks.26.mlp.c_fc",
35
+ "transformer.h.7.mlp.w2",
36
+ "transformer.visual.transformer.resblocks.21.mlp.c_fc",
37
+ "transformer.visual.transformer.resblocks.17.attn.in_proj",
38
+ "transformer.h.29.attn.c_attn",
39
+ "transformer.h.3.attn.c_proj",
40
+ "transformer.visual.transformer.resblocks.0.attn.out_proj",
41
+ "transformer.visual.transformer.resblocks.28.mlp.c_fc",
42
+ "transformer.h.30.attn.c_proj",
43
+ "transformer.h.3.mlp.w2",
44
+ "transformer.h.22.mlp.w1",
45
+ "transformer.visual.transformer.resblocks.11.mlp.c_proj",
46
+ "transformer.h.11.mlp.c_proj",
47
+ "transformer.visual.transformer.resblocks.3.attn.in_proj",
48
+ "transformer.visual.transformer.resblocks.21.attn.out_proj",
49
+ "transformer.visual.transformer.resblocks.33.attn.out_proj",
50
+ "transformer.visual.transformer.resblocks.3.mlp.c_fc",
51
+ "transformer.visual.transformer.resblocks.15.attn.in_proj",
52
+ "transformer.visual.transformer.resblocks.4.attn.out_proj",
53
+ "transformer.visual.transformer.resblocks.34.attn.in_proj",
54
+ "transformer.h.17.mlp.c_proj",
55
+ "transformer.visual.transformer.resblocks.1.mlp.c_proj",
56
+ "transformer.visual.transformer.resblocks.4.mlp.c_proj",
57
+ "transformer.visual.transformer.resblocks.39.attn.out_proj",
58
+ "transformer.h.13.mlp.c_proj",
59
+ "transformer.visual.transformer.resblocks.3.mlp.c_proj",
60
+ "transformer.visual.transformer.resblocks.22.attn.out_proj",
61
+ "transformer.h.27.attn.c_attn",
62
+ "transformer.visual.transformer.resblocks.47.attn.in_proj",
63
+ "transformer.h.1.mlp.c_proj",
64
+ "transformer.h.21.attn.c_attn",
65
+ "transformer.visual.transformer.resblocks.18.attn.out_proj",
66
+ "transformer.visual.transformer.resblocks.5.attn.in_proj",
67
+ "transformer.h.6.attn.c_proj",
68
+ "transformer.visual.transformer.resblocks.17.mlp.c_fc",
69
+ "transformer.h.16.attn.c_attn",
70
+ "transformer.visual.transformer.resblocks.0.attn.in_proj",
71
+ "transformer.visual.transformer.resblocks.22.mlp.c_fc",
72
+ "transformer.visual.transformer.resblocks.27.mlp.c_proj",
73
+ "transformer.h.11.attn.c_attn",
74
+ "transformer.h.22.mlp.w2",
75
+ "transformer.h.8.mlp.w1",
76
+ "transformer.visual.transformer.resblocks.29.mlp.c_proj",
77
+ "transformer.visual.transformer.resblocks.44.attn.out_proj",
78
+ "transformer.h.13.mlp.w2",
79
+ "transformer.visual.transformer.resblocks.19.mlp.c_proj",
80
+ "transformer.visual.transformer.resblocks.27.attn.in_proj",
81
+ "transformer.h.29.mlp.w1",
82
+ "transformer.h.24.mlp.c_proj",
83
+ "transformer.visual.transformer.resblocks.25.mlp.c_proj",
84
+ "transformer.visual.transformer.resblocks.4.attn.in_proj",
85
+ "transformer.visual.transformer.resblocks.42.mlp.c_fc",
86
+ "transformer.h.28.mlp.w2",
87
+ "transformer.visual.transformer.resblocks.20.attn.in_proj",
88
+ "transformer.visual.transformer.resblocks.9.attn.in_proj",
89
+ "transformer.h.10.attn.c_proj",
90
+ "transformer.h.13.attn.c_proj",
91
+ "transformer.visual.transformer.resblocks.26.attn.in_proj",
92
+ "transformer.h.17.attn.c_attn",
93
+ "transformer.visual.transformer.resblocks.7.attn.in_proj",
94
+ "transformer.h.23.attn.c_proj",
95
+ "transformer.visual.transformer.resblocks.38.attn.in_proj",
96
+ "transformer.h.19.attn.c_attn",
97
+ "transformer.h.1.attn.c_proj",
98
+ "transformer.visual.transformer.resblocks.2.attn.in_proj",
99
+ "transformer.h.4.mlp.w2",
100
+ "transformer.h.15.mlp.c_proj",
101
+ "transformer.h.4.mlp.c_proj",
102
+ "transformer.h.19.mlp.w2",
103
+ "transformer.h.12.mlp.w1",
104
+ "transformer.visual.transformer.resblocks.30.attn.in_proj",
105
+ "transformer.visual.transformer.resblocks.9.attn.out_proj",
106
+ "transformer.h.28.mlp.c_proj",
107
+ "transformer.h.1.attn.c_attn",
108
+ "transformer.h.8.attn.c_proj",
109
+ "transformer.visual.transformer.resblocks.40.mlp.c_fc",
110
+ "transformer.visual.transformer.resblocks.25.attn.out_proj",
111
+ "transformer.h.4.mlp.w1",
112
+ "transformer.visual.transformer.resblocks.31.mlp.c_fc",
113
+ "transformer.visual.transformer.resblocks.7.mlp.c_proj",
114
+ "transformer.visual.transformer.resblocks.33.attn.in_proj",
115
+ "transformer.h.0.mlp.w1",
116
+ "transformer.visual.transformer.resblocks.8.mlp.c_proj",
117
+ "transformer.h.20.mlp.w2",
118
+ "transformer.visual.transformer.resblocks.37.mlp.c_fc",
119
+ "transformer.visual.transformer.resblocks.43.attn.in_proj",
120
+ "transformer.visual.transformer.resblocks.37.attn.out_proj",
121
+ "transformer.h.25.mlp.w1",
122
+ "transformer.visual.transformer.resblocks.2.mlp.c_proj",
123
+ "transformer.h.27.mlp.w2",
124
+ "transformer.visual.transformer.resblocks.47.attn.out_proj",
125
+ "transformer.visual.transformer.resblocks.18.attn.in_proj",
126
+ "transformer.visual.transformer.resblocks.12.mlp.c_fc",
127
+ "transformer.visual.transformer.resblocks.11.attn.in_proj",
128
+ "transformer.h.14.mlp.c_proj",
129
+ "transformer.h.7.attn.c_attn",
130
+ "transformer.h.10.mlp.w2",
131
+ "transformer.h.11.mlp.w2",
132
+ "transformer.visual.transformer.resblocks.13.attn.out_proj",
133
+ "transformer.h.24.mlp.w1",
134
+ "transformer.h.0.mlp.c_proj",
135
+ "transformer.h.24.attn.c_proj",
136
+ "transformer.visual.transformer.resblocks.10.mlp.c_fc",
137
+ "transformer.visual.transformer.resblocks.33.mlp.c_proj",
138
+ "transformer.visual.transformer.resblocks.35.attn.in_proj",
139
+ "transformer.visual.transformer.resblocks.38.attn.out_proj",
140
+ "transformer.h.2.mlp.w2",
141
+ "transformer.visual.transformer.resblocks.40.attn.in_proj",
142
+ "transformer.h.25.mlp.w2",
143
+ "transformer.visual.transformer.resblocks.17.mlp.c_proj",
144
+ "transformer.visual.transformer.resblocks.18.mlp.c_proj",
145
+ "transformer.h.2.attn.c_proj",
146
+ "transformer.visual.transformer.resblocks.0.mlp.c_proj",
147
+ "transformer.visual.transformer.resblocks.23.mlp.c_proj",
148
+ "transformer.visual.transformer.resblocks.2.attn.out_proj",
149
+ "transformer.h.16.mlp.w2",
150
+ "transformer.h.29.mlp.c_proj",
151
+ "transformer.visual.transformer.resblocks.21.attn.in_proj",
152
+ "transformer.visual.transformer.resblocks.7.mlp.c_fc",
153
+ "transformer.visual.transformer.resblocks.8.attn.out_proj",
154
+ "transformer.visual.transformer.resblocks.28.attn.in_proj",
155
+ "transformer.h.11.mlp.w1",
156
+ "transformer.visual.transformer.resblocks.41.mlp.c_fc",
157
+ "transformer.h.18.attn.c_proj",
158
+ "transformer.visual.transformer.resblocks.29.mlp.c_fc",
159
+ "transformer.visual.transformer.resblocks.32.mlp.c_proj",
160
+ "transformer.visual.transformer.resblocks.28.attn.out_proj",
161
+ "transformer.h.6.attn.c_attn",
162
+ "transformer.visual.transformer.resblocks.3.attn.out_proj",
163
+ "transformer.h.25.attn.c_attn",
164
+ "transformer.h.28.attn.c_proj",
165
+ "transformer.visual.transformer.resblocks.16.attn.out_proj",
166
+ "transformer.visual.transformer.resblocks.20.mlp.c_proj",
167
+ "transformer.visual.transformer.resblocks.45.attn.in_proj",
168
+ "transformer.h.8.mlp.c_proj",
169
+ "transformer.visual.transformer.resblocks.29.attn.out_proj",
170
+ "transformer.h.27.mlp.c_proj",
171
+ "transformer.visual.transformer.resblocks.11.mlp.c_fc",
172
+ "transformer.visual.transformer.resblocks.11.attn.out_proj",
173
+ "transformer.visual.transformer.resblocks.10.attn.in_proj",
174
+ "transformer.visual.transformer.resblocks.27.mlp.c_fc",
175
+ "transformer.h.4.attn.c_proj",
176
+ "transformer.visual.transformer.resblocks.43.mlp.c_fc",
177
+ "transformer.visual.transformer.resblocks.22.mlp.c_proj",
178
+ "transformer.h.22.attn.c_proj",
179
+ "transformer.visual.transformer.resblocks.5.mlp.c_proj",
180
+ "transformer.h.22.mlp.c_proj",
181
+ "transformer.visual.transformer.resblocks.7.attn.out_proj",
182
+ "transformer.visual.transformer.resblocks.14.attn.out_proj",
183
+ "transformer.h.30.mlp.w1",
184
+ "transformer.h.14.attn.c_attn",
185
+ "transformer.h.4.attn.c_attn",
186
+ "transformer.visual.transformer.resblocks.46.mlp.c_fc",
187
+ "transformer.visual.transformer.resblocks.27.attn.out_proj",
188
+ "transformer.visual.transformer.resblocks.41.attn.in_proj",
189
+ "transformer.visual.transformer.resblocks.13.mlp.c_fc",
190
+ "transformer.h.12.attn.c_proj",
191
+ "transformer.visual.transformer.resblocks.32.attn.out_proj",
192
+ "transformer.visual.transformer.resblocks.18.mlp.c_fc",
193
+ "transformer.visual.transformer.resblocks.45.attn.out_proj",
194
+ "transformer.visual.transformer.resblocks.44.mlp.c_proj",
195
+ "transformer.visual.transformer.resblocks.35.mlp.c_proj",
196
+ "transformer.visual.transformer.resblocks.20.attn.out_proj",
197
+ "transformer.visual.transformer.resblocks.19.attn.in_proj",
198
+ "transformer.h.17.mlp.w2",
199
+ "transformer.h.10.attn.c_attn",
200
+ "transformer.visual.transformer.resblocks.45.mlp.c_proj",
201
+ "transformer.visual.transformer.resblocks.39.mlp.c_proj",
202
+ "transformer.h.7.mlp.w1",
203
+ "transformer.h.14.mlp.w1",
204
+ "transformer.visual.transformer.resblocks.2.mlp.c_fc",
205
+ "transformer.h.21.mlp.w1",
206
+ "transformer.visual.transformer.resblocks.30.attn.out_proj",
207
+ "transformer.h.19.mlp.c_proj",
208
+ "transformer.h.5.mlp.w1",
209
+ "transformer.visual.transformer.resblocks.47.mlp.c_fc",
210
+ "transformer.visual.transformer.resblocks.38.mlp.c_fc",
211
+ "transformer.h.10.mlp.c_proj",
212
+ "transformer.visual.transformer.resblocks.28.mlp.c_proj",
213
+ "transformer.visual.transformer.resblocks.25.attn.in_proj",
214
+ "transformer.h.9.attn.c_attn",
215
+ "transformer.visual.transformer.resblocks.0.mlp.c_fc",
216
+ "transformer.h.29.attn.c_proj",
217
+ "transformer.h.5.mlp.w2",
218
+ "transformer.h.30.attn.c_attn",
219
+ "transformer.h.1.mlp.w1",
220
+ "transformer.visual.transformer.resblocks.8.mlp.c_fc",
221
+ "transformer.h.19.mlp.w1",
222
+ "transformer.h.18.attn.c_attn",
223
+ "transformer.h.11.attn.c_proj",
224
+ "transformer.h.5.mlp.c_proj",
225
+ "transformer.visual.transformer.resblocks.45.mlp.c_fc",
226
+ "transformer.visual.transformer.resblocks.47.mlp.c_proj",
227
+ "transformer.h.9.attn.c_proj",
228
+ "transformer.visual.transformer.resblocks.6.attn.out_proj",
229
+ "transformer.h.26.mlp.c_proj",
230
+ "transformer.visual.transformer.resblocks.24.mlp.c_proj",
231
+ "transformer.h.31.attn.c_attn",
232
+ "transformer.h.13.mlp.w1",
233
+ "transformer.visual.transformer.resblocks.39.attn.in_proj",
234
+ "transformer.h.20.mlp.w1",
235
+ "transformer.visual.transformer.resblocks.22.attn.in_proj",
236
+ "transformer.visual.transformer.resblocks.35.mlp.c_fc",
237
+ "transformer.visual.transformer.resblocks.24.attn.in_proj",
238
+ "transformer.h.16.mlp.w1",
239
+ "transformer.visual.transformer.resblocks.6.mlp.c_proj",
240
+ "transformer.h.6.mlp.c_proj",
241
+ "transformer.visual.transformer.resblocks.26.mlp.c_proj",
242
+ "transformer.visual.transformer.resblocks.35.attn.out_proj",
243
+ "transformer.h.24.mlp.w2",
244
+ "transformer.visual.transformer.resblocks.31.attn.in_proj",
245
+ "transformer.visual.transformer.resblocks.15.mlp.c_fc",
246
+ "transformer.h.2.mlp.w1",
247
+ "transformer.h.31.mlp.c_proj",
248
+ "transformer.h.13.attn.c_attn",
249
+ "transformer.visual.transformer.resblocks.14.attn.in_proj",
250
+ "transformer.h.12.mlp.w2",
251
+ "transformer.visual.transformer.resblocks.34.mlp.c_proj",
252
+ "transformer.h.26.mlp.w2",
253
+ "transformer.visual.transformer.resblocks.5.mlp.c_fc",
254
+ "transformer.h.5.attn.c_proj",
255
+ "transformer.h.9.mlp.w2",
256
+ "transformer.h.15.mlp.w2",
257
+ "transformer.h.12.attn.c_attn",
258
+ "transformer.visual.transformer.resblocks.10.mlp.c_proj",
259
+ "transformer.h.28.mlp.w1",
260
+ "transformer.visual.transformer.resblocks.46.attn.out_proj",
261
+ "transformer.h.18.mlp.c_proj",
262
+ "transformer.h.15.attn.c_proj",
263
+ "transformer.visual.transformer.resblocks.44.attn.in_proj",
264
+ "transformer.visual.transformer.resblocks.12.attn.out_proj",
265
+ "transformer.visual.transformer.resblocks.1.mlp.c_fc",
266
+ "transformer.visual.transformer.resblocks.15.attn.out_proj",
267
+ "transformer.h.17.mlp.w1",
268
+ "transformer.visual.transformer.resblocks.46.attn.in_proj",
269
+ "transformer.h.2.attn.c_attn",
270
+ "transformer.h.25.attn.c_proj",
271
+ "transformer.h.14.attn.c_proj",
272
+ "transformer.h.26.attn.c_proj",
273
+ "transformer.h.31.mlp.w1",
274
+ "transformer.h.23.mlp.w2",
275
+ "transformer.visual.transformer.resblocks.26.attn.out_proj",
276
+ "transformer.h.20.mlp.c_proj",
277
+ "transformer.visual.transformer.resblocks.6.mlp.c_fc",
278
+ "transformer.h.27.mlp.w1",
279
+ "transformer.h.7.attn.c_proj",
280
+ "transformer.visual.transformer.resblocks.15.mlp.c_proj",
281
+ "transformer.h.16.mlp.c_proj",
282
+ "transformer.visual.transformer.resblocks.12.mlp.c_proj",
283
+ "transformer.h.29.mlp.w2",
284
+ "transformer.h.15.mlp.w1",
285
+ "transformer.h.6.mlp.w2",
286
+ "transformer.h.3.attn.c_attn",
287
+ "transformer.h.21.mlp.w2",
288
+ "transformer.visual.transformer.resblocks.16.mlp.c_fc",
289
+ "transformer.visual.transformer.resblocks.23.mlp.c_fc",
290
+ "transformer.visual.transformer.resblocks.34.attn.out_proj",
291
+ "transformer.h.8.attn.c_attn",
292
+ "transformer.visual.transformer.resblocks.43.mlp.c_proj",
293
+ "transformer.visual.transformer.resblocks.14.mlp.c_fc",
294
+ "transformer.visual.transformer.resblocks.24.attn.out_proj",
295
+ "transformer.visual.transformer.resblocks.12.attn.in_proj",
296
+ "transformer.h.25.mlp.c_proj",
297
+ "transformer.visual.transformer.resblocks.46.mlp.c_proj",
298
+ "transformer.h.7.mlp.c_proj",
299
+ "transformer.h.15.attn.c_attn",
300
+ "transformer.visual.transformer.resblocks.36.attn.in_proj",
301
+ "transformer.h.26.attn.c_attn",
302
+ "transformer.h.0.attn.c_proj",
303
+ "transformer.visual.transformer.resblocks.9.mlp.c_fc",
304
+ "transformer.h.19.attn.c_proj",
305
+ "transformer.visual.transformer.resblocks.29.attn.in_proj",
306
+ "transformer.visual.transformer.resblocks.13.attn.in_proj",
307
+ "transformer.visual.transformer.resblocks.23.attn.in_proj",
308
+ "transformer.visual.transformer.resblocks.36.attn.out_proj",
309
+ "transformer.h.3.mlp.c_proj",
310
+ "transformer.h.27.attn.c_proj",
311
+ "transformer.h.31.attn.c_proj",
312
+ "transformer.visual.transformer.resblocks.19.attn.out_proj",
313
+ "transformer.h.0.mlp.w2",
314
+ "transformer.h.17.attn.c_proj",
315
+ "transformer.h.30.mlp.w2",
316
+ "transformer.visual.transformer.resblocks.24.mlp.c_fc",
317
+ "transformer.h.28.attn.c_attn",
318
+ "transformer.visual.transformer.resblocks.16.mlp.c_proj",
319
+ "transformer.visual.transformer.resblocks.36.mlp.c_fc",
320
+ "transformer.h.30.mlp.c_proj",
321
+ "transformer.visual.transformer.resblocks.37.attn.in_proj",
322
+ "transformer.visual.transformer.resblocks.20.mlp.c_fc",
323
+ "transformer.h.9.mlp.c_proj",
324
+ "transformer.visual.transformer.resblocks.43.attn.out_proj",
325
+ "transformer.visual.transformer.resblocks.42.attn.in_proj",
326
+ "transformer.h.1.mlp.w2",
327
+ "transformer.h.6.mlp.w1",
328
+ "transformer.visual.transformer.resblocks.17.attn.out_proj",
329
+ "transformer.visual.transformer.resblocks.33.mlp.c_fc",
330
+ "transformer.h.5.attn.c_attn",
331
+ "transformer.h.8.mlp.w2",
332
+ "transformer.h.23.mlp.c_proj",
333
+ "transformer.h.20.attn.c_attn",
334
+ "transformer.visual.transformer.resblocks.4.mlp.c_fc",
335
+ "transformer.h.31.mlp.w2",
336
+ "transformer.h.9.mlp.w1",
337
+ "transformer.h.12.mlp.c_proj",
338
+ "transformer.visual.transformer.resblocks.1.attn.out_proj",
339
+ "transformer.visual.transformer.resblocks.30.mlp.c_fc",
340
+ "transformer.visual.transformer.resblocks.1.attn.in_proj",
341
+ "transformer.visual.transformer.resblocks.32.attn.in_proj",
342
+ "transformer.h.16.attn.c_proj",
343
+ "transformer.visual.transformer.resblocks.42.mlp.c_proj",
344
+ "transformer.h.3.mlp.w1",
345
+ "transformer.visual.transformer.resblocks.9.mlp.c_proj",
346
+ "transformer.h.18.mlp.w1",
347
+ "transformer.visual.transformer.resblocks.44.mlp.c_fc",
348
+ "transformer.h.21.mlp.c_proj",
349
+ "transformer.visual.transformer.resblocks.5.attn.out_proj",
350
+ "transformer.visual.transformer.resblocks.34.mlp.c_fc",
351
+ "transformer.visual.transformer.resblocks.42.attn.out_proj",
352
+ "transformer.visual.transformer.resblocks.41.attn.out_proj",
353
+ "transformer.h.10.mlp.w1",
354
+ "transformer.visual.transformer.resblocks.19.mlp.c_fc",
355
+ "transformer.visual.transformer.resblocks.13.mlp.c_proj",
356
+ "transformer.h.21.attn.c_proj",
357
+ "transformer.visual.transformer.resblocks.6.attn.in_proj",
358
+ "transformer.visual.transformer.resblocks.23.attn.out_proj",
359
+ "transformer.visual.transformer.resblocks.40.attn.out_proj",
360
+ "transformer.visual.transformer.resblocks.36.mlp.c_proj",
361
+ "transformer.h.2.mlp.c_proj",
362
+ "transformer.visual.transformer.resblocks.25.mlp.c_fc",
363
+ "transformer.h.22.attn.c_attn",
364
+ "transformer.h.23.mlp.w1",
365
+ "transformer.h.20.attn.c_proj",
366
+ "transformer.h.23.attn.c_attn",
367
+ "transformer.visual.transformer.resblocks.37.mlp.c_proj",
368
+ "transformer.h.26.mlp.w1",
369
+ "transformer.h.18.mlp.w2",
370
+ "transformer.visual.transformer.resblocks.21.mlp.c_proj",
371
+ "transformer.visual.transformer.resblocks.31.attn.out_proj",
372
+ "transformer.h.24.attn.c_attn",
373
+ "transformer.visual.transformer.resblocks.10.attn.out_proj",
374
+ "transformer.visual.transformer.resblocks.30.mlp.c_proj",
375
+ "transformer.visual.transformer.resblocks.38.mlp.c_proj"
376
+ ],
377
+ "task_type": "CAUSAL_LM",
378
+ "use_dora": false,
379
+ "use_rslora": false
380
+ }
checkpoint-1200/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db88b90bb7ac9649582bbf46e54c8eada48a8e8bd5b6320a2df24ed173fc1829
3
+ size 469105640
checkpoint-1200/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step1200
checkpoint-1200/qwen.tiktoken ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-1200/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f65bd43ec266e83fbc7c05e0ae3dc56511353bbd178526e8712deba877ff45a5
3
+ size 14960
checkpoint-1200/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce79252b084241544b5446d4368a6dcb69615e386688b4e69d5328e5e26fdef1
3
+ size 14960
checkpoint-1200/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d2e6e0b1b7cd100ff2ce47ac8562fb598dd799f731f8b79d741eab77a0f030c
3
+ size 14960
checkpoint-1200/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd25289c6a8d6fa0d90d929482ac6674692fee99f0b5002b01c466d60b1ab094
3
+ size 14960
checkpoint-1200/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:034ee3da76b692bee85a05be8fac00a73173abe1e3698813e0bd90da0df473f4
3
+ size 1064
checkpoint-1200/special_tokens_map.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "pad_token": "<|endoftext|>"
3
+ }
checkpoint-1200/tokenizer_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {},
3
+ "auto_map": {
4
+ "AutoTokenizer": [
5
+ "Qwen/Qwen-VL-Chat--tokenization_qwen.QWenTokenizer",
6
+ null
7
+ ]
8
+ },
9
+ "clean_up_tokenization_spaces": true,
10
+ "model_max_length": 768,
11
+ "pad_token": "<|endoftext|>",
12
+ "padding_side": "right",
13
+ "tokenizer_class": "QWenTokenizer"
14
+ }
checkpoint-1200/trainer_state.json ADDED
@@ -0,0 +1,873 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.16005868818566807,
5
+ "eval_steps": 500,
6
+ "global_step": 1200,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.001333822401547234,
13
+ "grad_norm": 5.80256772259428,
14
+ "learning_rate": 4e-06,
15
+ "loss": 1.0498,
16
+ "step": 10
17
+ },
18
+ {
19
+ "epoch": 0.002667644803094468,
20
+ "grad_norm": 33.895696082107904,
21
+ "learning_rate": 8e-06,
22
+ "loss": 1.0653,
23
+ "step": 20
24
+ },
25
+ {
26
+ "epoch": 0.004001467204641702,
27
+ "grad_norm": 5.523348234283539,
28
+ "learning_rate": 1.2e-05,
29
+ "loss": 1.0341,
30
+ "step": 30
31
+ },
32
+ {
33
+ "epoch": 0.005335289606188936,
34
+ "grad_norm": 11.1556403156453,
35
+ "learning_rate": 1.6e-05,
36
+ "loss": 0.9692,
37
+ "step": 40
38
+ },
39
+ {
40
+ "epoch": 0.00666911200773617,
41
+ "grad_norm": 3.7375231126561825,
42
+ "learning_rate": 1.9999999999999998e-05,
43
+ "loss": 0.9554,
44
+ "step": 50
45
+ },
46
+ {
47
+ "epoch": 0.008002934409283404,
48
+ "grad_norm": 8.43538339698909,
49
+ "learning_rate": 2.4e-05,
50
+ "loss": 0.8965,
51
+ "step": 60
52
+ },
53
+ {
54
+ "epoch": 0.009336756810830639,
55
+ "grad_norm": 13.403454896011478,
56
+ "learning_rate": 2.8e-05,
57
+ "loss": 0.8273,
58
+ "step": 70
59
+ },
60
+ {
61
+ "epoch": 0.010670579212377872,
62
+ "grad_norm": 3.95522050766088,
63
+ "learning_rate": 2.9999966406213696e-05,
64
+ "loss": 0.7837,
65
+ "step": 80
66
+ },
67
+ {
68
+ "epoch": 0.012004401613925107,
69
+ "grad_norm": 36.799552052300854,
70
+ "learning_rate": 2.9999697656826056e-05,
71
+ "loss": 0.8288,
72
+ "step": 90
73
+ },
74
+ {
75
+ "epoch": 0.01333822401547234,
76
+ "grad_norm": 1.6305479563258536,
77
+ "learning_rate": 2.9999160162865885e-05,
78
+ "loss": 0.7778,
79
+ "step": 100
80
+ },
81
+ {
82
+ "epoch": 0.014672046417019574,
83
+ "grad_norm": 2.159536648784889,
84
+ "learning_rate": 2.9998353933963273e-05,
85
+ "loss": 0.7616,
86
+ "step": 110
87
+ },
88
+ {
89
+ "epoch": 0.016005868818566808,
90
+ "grad_norm": 3.397321425707004,
91
+ "learning_rate": 2.999727898456315e-05,
92
+ "loss": 0.7594,
93
+ "step": 120
94
+ },
95
+ {
96
+ "epoch": 0.017339691220114042,
97
+ "grad_norm": 4.772220837365037,
98
+ "learning_rate": 2.999593533392503e-05,
99
+ "loss": 0.756,
100
+ "step": 130
101
+ },
102
+ {
103
+ "epoch": 0.018673513621661277,
104
+ "grad_norm": 2.4845945633126885,
105
+ "learning_rate": 2.9994323006122654e-05,
106
+ "loss": 0.7601,
107
+ "step": 140
108
+ },
109
+ {
110
+ "epoch": 0.02000733602320851,
111
+ "grad_norm": 3.591682569169127,
112
+ "learning_rate": 2.9992442030043557e-05,
113
+ "loss": 0.7894,
114
+ "step": 150
115
+ },
116
+ {
117
+ "epoch": 0.021341158424755743,
118
+ "grad_norm": 2.5679458807474416,
119
+ "learning_rate": 2.9990292439388565e-05,
120
+ "loss": 0.7093,
121
+ "step": 160
122
+ },
123
+ {
124
+ "epoch": 0.022674980826302978,
125
+ "grad_norm": 1.9412569107551652,
126
+ "learning_rate": 2.9987874272671168e-05,
127
+ "loss": 0.706,
128
+ "step": 170
129
+ },
130
+ {
131
+ "epoch": 0.024008803227850213,
132
+ "grad_norm": 3.2667097270489,
133
+ "learning_rate": 2.9985187573216855e-05,
134
+ "loss": 0.7586,
135
+ "step": 180
136
+ },
137
+ {
138
+ "epoch": 0.025342625629397444,
139
+ "grad_norm": 4.4208737375400675,
140
+ "learning_rate": 2.998223238916232e-05,
141
+ "loss": 0.6985,
142
+ "step": 190
143
+ },
144
+ {
145
+ "epoch": 0.02667644803094468,
146
+ "grad_norm": 5.515966302183704,
147
+ "learning_rate": 2.9979008773454618e-05,
148
+ "loss": 0.7323,
149
+ "step": 200
150
+ },
151
+ {
152
+ "epoch": 0.028010270432491914,
153
+ "grad_norm": 2.964165450396077,
154
+ "learning_rate": 2.997551678385019e-05,
155
+ "loss": 0.7603,
156
+ "step": 210
157
+ },
158
+ {
159
+ "epoch": 0.02934409283403915,
160
+ "grad_norm": 3.0952916783456197,
161
+ "learning_rate": 2.997175648291384e-05,
162
+ "loss": 0.7421,
163
+ "step": 220
164
+ },
165
+ {
166
+ "epoch": 0.03067791523558638,
167
+ "grad_norm": 4.213588693904103,
168
+ "learning_rate": 2.996772793801763e-05,
169
+ "loss": 0.7322,
170
+ "step": 230
171
+ },
172
+ {
173
+ "epoch": 0.032011737637133615,
174
+ "grad_norm": 1.8568586103139084,
175
+ "learning_rate": 2.996343122133965e-05,
176
+ "loss": 0.6922,
177
+ "step": 240
178
+ },
179
+ {
180
+ "epoch": 0.033345560038680847,
181
+ "grad_norm": 4.494146778909846,
182
+ "learning_rate": 2.9958866409862745e-05,
183
+ "loss": 0.7244,
184
+ "step": 250
185
+ },
186
+ {
187
+ "epoch": 0.034679382440228085,
188
+ "grad_norm": 7.438170074282725,
189
+ "learning_rate": 2.9954033585373108e-05,
190
+ "loss": 0.7093,
191
+ "step": 260
192
+ },
193
+ {
194
+ "epoch": 0.036013204841775316,
195
+ "grad_norm": 2.3744787346857015,
196
+ "learning_rate": 2.994893283445885e-05,
197
+ "loss": 0.6983,
198
+ "step": 270
199
+ },
200
+ {
201
+ "epoch": 0.037347027243322554,
202
+ "grad_norm": 1.4722011682616383,
203
+ "learning_rate": 2.9943564248508415e-05,
204
+ "loss": 0.6781,
205
+ "step": 280
206
+ },
207
+ {
208
+ "epoch": 0.038680849644869786,
209
+ "grad_norm": 3.3397620832486075,
210
+ "learning_rate": 2.9937927923708966e-05,
211
+ "loss": 0.7399,
212
+ "step": 290
213
+ },
214
+ {
215
+ "epoch": 0.04001467204641702,
216
+ "grad_norm": 5.05063397044549,
217
+ "learning_rate": 2.993202396104465e-05,
218
+ "loss": 0.7671,
219
+ "step": 300
220
+ },
221
+ {
222
+ "epoch": 0.041348494447964255,
223
+ "grad_norm": 3.0128431385936767,
224
+ "learning_rate": 2.9925852466294795e-05,
225
+ "loss": 0.7015,
226
+ "step": 310
227
+ },
228
+ {
229
+ "epoch": 0.04268231684951149,
230
+ "grad_norm": 2.0161342716764237,
231
+ "learning_rate": 2.9919413550032014e-05,
232
+ "loss": 0.7009,
233
+ "step": 320
234
+ },
235
+ {
236
+ "epoch": 0.04401613925105872,
237
+ "grad_norm": 1.3114004070324985,
238
+ "learning_rate": 2.991270732762022e-05,
239
+ "loss": 0.7153,
240
+ "step": 330
241
+ },
242
+ {
243
+ "epoch": 0.045349961652605957,
244
+ "grad_norm": 18.493625676806268,
245
+ "learning_rate": 2.990573391921255e-05,
246
+ "loss": 0.7518,
247
+ "step": 340
248
+ },
249
+ {
250
+ "epoch": 0.04668378405415319,
251
+ "grad_norm": 2.9526764059703567,
252
+ "learning_rate": 2.989849344974924e-05,
253
+ "loss": 0.7133,
254
+ "step": 350
255
+ },
256
+ {
257
+ "epoch": 0.048017606455700426,
258
+ "grad_norm": 5.26274958582726,
259
+ "learning_rate": 2.9890986048955368e-05,
260
+ "loss": 0.7139,
261
+ "step": 360
262
+ },
263
+ {
264
+ "epoch": 0.04935142885724766,
265
+ "grad_norm": 3.5319788357887933,
266
+ "learning_rate": 2.9883211851338516e-05,
267
+ "loss": 0.7084,
268
+ "step": 370
269
+ },
270
+ {
271
+ "epoch": 0.05068525125879489,
272
+ "grad_norm": 7.607269935902469,
273
+ "learning_rate": 2.9875170996186392e-05,
274
+ "loss": 0.7309,
275
+ "step": 380
276
+ },
277
+ {
278
+ "epoch": 0.05201907366034213,
279
+ "grad_norm": 2.3456663308287253,
280
+ "learning_rate": 2.986686362756431e-05,
281
+ "loss": 0.6827,
282
+ "step": 390
283
+ },
284
+ {
285
+ "epoch": 0.05335289606188936,
286
+ "grad_norm": 2.176182050789012,
287
+ "learning_rate": 2.9858289894312617e-05,
288
+ "loss": 0.6995,
289
+ "step": 400
290
+ },
291
+ {
292
+ "epoch": 0.0546867184634366,
293
+ "grad_norm": 11.171630173781537,
294
+ "learning_rate": 2.9849449950044036e-05,
295
+ "loss": 0.7335,
296
+ "step": 410
297
+ },
298
+ {
299
+ "epoch": 0.05602054086498383,
300
+ "grad_norm": 6.63441431767892,
301
+ "learning_rate": 2.984034395314088e-05,
302
+ "loss": 0.7031,
303
+ "step": 420
304
+ },
305
+ {
306
+ "epoch": 0.05735436326653106,
307
+ "grad_norm": 2.861620412225736,
308
+ "learning_rate": 2.983097206675227e-05,
309
+ "loss": 0.6559,
310
+ "step": 430
311
+ },
312
+ {
313
+ "epoch": 0.0586881856680783,
314
+ "grad_norm": 5.523165036486206,
315
+ "learning_rate": 2.9821334458791156e-05,
316
+ "loss": 0.726,
317
+ "step": 440
318
+ },
319
+ {
320
+ "epoch": 0.06002200806962553,
321
+ "grad_norm": 3.5602243751368197,
322
+ "learning_rate": 2.9811431301931344e-05,
323
+ "loss": 0.7202,
324
+ "step": 450
325
+ },
326
+ {
327
+ "epoch": 0.06135583047117276,
328
+ "grad_norm": 11.333380381168622,
329
+ "learning_rate": 2.9801262773604377e-05,
330
+ "loss": 0.7189,
331
+ "step": 460
332
+ },
333
+ {
334
+ "epoch": 0.06268965287271999,
335
+ "grad_norm": 14.159758615106613,
336
+ "learning_rate": 2.9790829055996398e-05,
337
+ "loss": 0.7267,
338
+ "step": 470
339
+ },
340
+ {
341
+ "epoch": 0.06402347527426723,
342
+ "grad_norm": 9.009079485918289,
343
+ "learning_rate": 2.978013033604483e-05,
344
+ "loss": 0.748,
345
+ "step": 480
346
+ },
347
+ {
348
+ "epoch": 0.06535729767581447,
349
+ "grad_norm": 1.9682648681675994,
350
+ "learning_rate": 2.976916680543506e-05,
351
+ "loss": 0.7369,
352
+ "step": 490
353
+ },
354
+ {
355
+ "epoch": 0.06669112007736169,
356
+ "grad_norm": 2.9278164598232777,
357
+ "learning_rate": 2.975793866059701e-05,
358
+ "loss": 0.7037,
359
+ "step": 500
360
+ },
361
+ {
362
+ "epoch": 0.06802494247890893,
363
+ "grad_norm": 5.5563562303649885,
364
+ "learning_rate": 2.9746446102701606e-05,
365
+ "loss": 0.6986,
366
+ "step": 510
367
+ },
368
+ {
369
+ "epoch": 0.06935876488045617,
370
+ "grad_norm": 4.036767303783137,
371
+ "learning_rate": 2.9734689337657157e-05,
372
+ "loss": 0.7119,
373
+ "step": 520
374
+ },
375
+ {
376
+ "epoch": 0.07069258728200341,
377
+ "grad_norm": 1.9856990692088847,
378
+ "learning_rate": 2.9722668576105703e-05,
379
+ "loss": 0.7205,
380
+ "step": 530
381
+ },
382
+ {
383
+ "epoch": 0.07202640968355063,
384
+ "grad_norm": 5.200308739226583,
385
+ "learning_rate": 2.971038403341921e-05,
386
+ "loss": 0.6918,
387
+ "step": 540
388
+ },
389
+ {
390
+ "epoch": 0.07336023208509787,
391
+ "grad_norm": 2.237349124701919,
392
+ "learning_rate": 2.9697835929695727e-05,
393
+ "loss": 0.7339,
394
+ "step": 550
395
+ },
396
+ {
397
+ "epoch": 0.07469405448664511,
398
+ "grad_norm": 1.6388680632753365,
399
+ "learning_rate": 2.968502448975544e-05,
400
+ "loss": 0.7086,
401
+ "step": 560
402
+ },
403
+ {
404
+ "epoch": 0.07602787688819233,
405
+ "grad_norm": 2.8545575025135244,
406
+ "learning_rate": 2.967194994313663e-05,
407
+ "loss": 0.678,
408
+ "step": 570
409
+ },
410
+ {
411
+ "epoch": 0.07736169928973957,
412
+ "grad_norm": 2.674647983669599,
413
+ "learning_rate": 2.9658612524091594e-05,
414
+ "loss": 0.7119,
415
+ "step": 580
416
+ },
417
+ {
418
+ "epoch": 0.07869552169128681,
419
+ "grad_norm": 2.489047760330112,
420
+ "learning_rate": 2.9645012471582406e-05,
421
+ "loss": 0.7382,
422
+ "step": 590
423
+ },
424
+ {
425
+ "epoch": 0.08002934409283403,
426
+ "grad_norm": 5.509352102248308,
427
+ "learning_rate": 2.9631150029276662e-05,
428
+ "loss": 0.738,
429
+ "step": 600
430
+ },
431
+ {
432
+ "epoch": 0.08136316649438127,
433
+ "grad_norm": 3.6489235270404015,
434
+ "learning_rate": 2.9617025445543114e-05,
435
+ "loss": 0.7018,
436
+ "step": 610
437
+ },
438
+ {
439
+ "epoch": 0.08269698889592851,
440
+ "grad_norm": 2.7813651243235697,
441
+ "learning_rate": 2.9602638973447218e-05,
442
+ "loss": 0.7381,
443
+ "step": 620
444
+ },
445
+ {
446
+ "epoch": 0.08403081129747574,
447
+ "grad_norm": 8.271390523006518,
448
+ "learning_rate": 2.9587990870746574e-05,
449
+ "loss": 0.7168,
450
+ "step": 630
451
+ },
452
+ {
453
+ "epoch": 0.08536463369902297,
454
+ "grad_norm": 1.2460611751687307,
455
+ "learning_rate": 2.9573081399886356e-05,
456
+ "loss": 0.7004,
457
+ "step": 640
458
+ },
459
+ {
460
+ "epoch": 0.08669845610057021,
461
+ "grad_norm": 1.704626418994062,
462
+ "learning_rate": 2.9557910827994568e-05,
463
+ "loss": 0.738,
464
+ "step": 650
465
+ },
466
+ {
467
+ "epoch": 0.08803227850211744,
468
+ "grad_norm": 3.275051693107957,
469
+ "learning_rate": 2.9542479426877283e-05,
470
+ "loss": 0.7017,
471
+ "step": 660
472
+ },
473
+ {
474
+ "epoch": 0.08936610090366467,
475
+ "grad_norm": 11.389990685570503,
476
+ "learning_rate": 2.9526787473013753e-05,
477
+ "loss": 0.7107,
478
+ "step": 670
479
+ },
480
+ {
481
+ "epoch": 0.09069992330521191,
482
+ "grad_norm": 5.591277359184055,
483
+ "learning_rate": 2.9510835247551485e-05,
484
+ "loss": 0.7141,
485
+ "step": 680
486
+ },
487
+ {
488
+ "epoch": 0.09203374570675915,
489
+ "grad_norm": 3.180111568581053,
490
+ "learning_rate": 2.949462303630116e-05,
491
+ "loss": 0.6987,
492
+ "step": 690
493
+ },
494
+ {
495
+ "epoch": 0.09336756810830638,
496
+ "grad_norm": 3.8428068166831753,
497
+ "learning_rate": 2.9478151129731567e-05,
498
+ "loss": 0.7373,
499
+ "step": 700
500
+ },
501
+ {
502
+ "epoch": 0.09470139050985361,
503
+ "grad_norm": 2.231397231771392,
504
+ "learning_rate": 2.9461419822964348e-05,
505
+ "loss": 0.6962,
506
+ "step": 710
507
+ },
508
+ {
509
+ "epoch": 0.09603521291140085,
510
+ "grad_norm": 18.287201889017563,
511
+ "learning_rate": 2.9444429415768726e-05,
512
+ "loss": 0.6723,
513
+ "step": 720
514
+ },
515
+ {
516
+ "epoch": 0.09736903531294808,
517
+ "grad_norm": 4.340932687135137,
518
+ "learning_rate": 2.942718021255617e-05,
519
+ "loss": 0.7151,
520
+ "step": 730
521
+ },
522
+ {
523
+ "epoch": 0.09870285771449532,
524
+ "grad_norm": 2.7813821825484446,
525
+ "learning_rate": 2.940967252237488e-05,
526
+ "loss": 0.7332,
527
+ "step": 740
528
+ },
529
+ {
530
+ "epoch": 0.10003668011604255,
531
+ "grad_norm": 2.3251782912937475,
532
+ "learning_rate": 2.9391906658904296e-05,
533
+ "loss": 0.6751,
534
+ "step": 750
535
+ },
536
+ {
537
+ "epoch": 0.10137050251758978,
538
+ "grad_norm": 8.123799866292751,
539
+ "learning_rate": 2.937388294044946e-05,
540
+ "loss": 0.6886,
541
+ "step": 760
542
+ },
543
+ {
544
+ "epoch": 0.10270432491913702,
545
+ "grad_norm": 1.528579329214318,
546
+ "learning_rate": 2.9355601689935315e-05,
547
+ "loss": 0.7146,
548
+ "step": 770
549
+ },
550
+ {
551
+ "epoch": 0.10403814732068425,
552
+ "grad_norm": 2.0278953433974825,
553
+ "learning_rate": 2.933706323490092e-05,
554
+ "loss": 0.7453,
555
+ "step": 780
556
+ },
557
+ {
558
+ "epoch": 0.10537196972223148,
559
+ "grad_norm": 1.4306270659678864,
560
+ "learning_rate": 2.9318267907493583e-05,
561
+ "loss": 0.6702,
562
+ "step": 790
563
+ },
564
+ {
565
+ "epoch": 0.10670579212377872,
566
+ "grad_norm": 1.5178081087799355,
567
+ "learning_rate": 2.9299216044462903e-05,
568
+ "loss": 0.7346,
569
+ "step": 800
570
+ },
571
+ {
572
+ "epoch": 0.10803961452532596,
573
+ "grad_norm": 9.506616797760028,
574
+ "learning_rate": 2.927990798715475e-05,
575
+ "loss": 0.6558,
576
+ "step": 810
577
+ },
578
+ {
579
+ "epoch": 0.1093734369268732,
580
+ "grad_norm": 2.4597311302505767,
581
+ "learning_rate": 2.926034408150513e-05,
582
+ "loss": 0.726,
583
+ "step": 820
584
+ },
585
+ {
586
+ "epoch": 0.11070725932842042,
587
+ "grad_norm": 12.372180964422007,
588
+ "learning_rate": 2.9240524678034016e-05,
589
+ "loss": 0.7308,
590
+ "step": 830
591
+ },
592
+ {
593
+ "epoch": 0.11204108172996766,
594
+ "grad_norm": 1.4488469801164658,
595
+ "learning_rate": 2.9220450131839037e-05,
596
+ "loss": 0.7072,
597
+ "step": 840
598
+ },
599
+ {
600
+ "epoch": 0.1133749041315149,
601
+ "grad_norm": 8.602946960846197,
602
+ "learning_rate": 2.920012080258912e-05,
603
+ "loss": 0.7234,
604
+ "step": 850
605
+ },
606
+ {
607
+ "epoch": 0.11470872653306212,
608
+ "grad_norm": 1.441195423452674,
609
+ "learning_rate": 2.9179537054518085e-05,
610
+ "loss": 0.6934,
611
+ "step": 860
612
+ },
613
+ {
614
+ "epoch": 0.11604254893460936,
615
+ "grad_norm": 4.318952956999577,
616
+ "learning_rate": 2.9158699256418056e-05,
617
+ "loss": 0.6534,
618
+ "step": 870
619
+ },
620
+ {
621
+ "epoch": 0.1173763713361566,
622
+ "grad_norm": 9.733179695623866,
623
+ "learning_rate": 2.9137607781632913e-05,
624
+ "loss": 0.71,
625
+ "step": 880
626
+ },
627
+ {
628
+ "epoch": 0.11871019373770382,
629
+ "grad_norm": 7.397049093836735,
630
+ "learning_rate": 2.911626300805155e-05,
631
+ "loss": 0.7386,
632
+ "step": 890
633
+ },
634
+ {
635
+ "epoch": 0.12004401613925106,
636
+ "grad_norm": 2.920812240139869,
637
+ "learning_rate": 2.9094665318101155e-05,
638
+ "loss": 0.6789,
639
+ "step": 900
640
+ },
641
+ {
642
+ "epoch": 0.1213778385407983,
643
+ "grad_norm": 1.7031296196271206,
644
+ "learning_rate": 2.9072815098740326e-05,
645
+ "loss": 0.715,
646
+ "step": 910
647
+ },
648
+ {
649
+ "epoch": 0.12271166094234552,
650
+ "grad_norm": 1.5630656172291801,
651
+ "learning_rate": 2.9050712741452136e-05,
652
+ "loss": 0.7136,
653
+ "step": 920
654
+ },
655
+ {
656
+ "epoch": 0.12404548334389276,
657
+ "grad_norm": 7.870543414771234,
658
+ "learning_rate": 2.902835864223715e-05,
659
+ "loss": 0.6669,
660
+ "step": 930
661
+ },
662
+ {
663
+ "epoch": 0.12537930574543998,
664
+ "grad_norm": 4.843671834991794,
665
+ "learning_rate": 2.9005753201606287e-05,
666
+ "loss": 0.7281,
667
+ "step": 940
668
+ },
669
+ {
670
+ "epoch": 0.12671312814698724,
671
+ "grad_norm": 3.010503818258016,
672
+ "learning_rate": 2.8982896824573678e-05,
673
+ "loss": 0.7018,
674
+ "step": 950
675
+ },
676
+ {
677
+ "epoch": 0.12804695054853446,
678
+ "grad_norm": 2.5552186559589654,
679
+ "learning_rate": 2.8959789920649394e-05,
680
+ "loss": 0.7338,
681
+ "step": 960
682
+ },
683
+ {
684
+ "epoch": 0.12938077295008168,
685
+ "grad_norm": 12.306055851495117,
686
+ "learning_rate": 2.893643290383212e-05,
687
+ "loss": 0.6732,
688
+ "step": 970
689
+ },
690
+ {
691
+ "epoch": 0.13071459535162894,
692
+ "grad_norm": 2.16185926525944,
693
+ "learning_rate": 2.891282619260172e-05,
694
+ "loss": 0.7108,
695
+ "step": 980
696
+ },
697
+ {
698
+ "epoch": 0.13204841775317616,
699
+ "grad_norm": 5.992378798792086,
700
+ "learning_rate": 2.8888970209911754e-05,
701
+ "loss": 0.6525,
702
+ "step": 990
703
+ },
704
+ {
705
+ "epoch": 0.13338224015472339,
706
+ "grad_norm": 2.986272238787896,
707
+ "learning_rate": 2.8864865383181893e-05,
708
+ "loss": 0.6655,
709
+ "step": 1000
710
+ },
711
+ {
712
+ "epoch": 0.13471606255627064,
713
+ "grad_norm": 12.855377354582437,
714
+ "learning_rate": 2.8840512144290273e-05,
715
+ "loss": 0.6826,
716
+ "step": 1010
717
+ },
718
+ {
719
+ "epoch": 0.13604988495781786,
720
+ "grad_norm": 2.045979893776702,
721
+ "learning_rate": 2.8815910929565734e-05,
722
+ "loss": 0.6616,
723
+ "step": 1020
724
+ },
725
+ {
726
+ "epoch": 0.1373837073593651,
727
+ "grad_norm": 6.623264301300591,
728
+ "learning_rate": 2.879106217978002e-05,
729
+ "loss": 0.6935,
730
+ "step": 1030
731
+ },
732
+ {
733
+ "epoch": 0.13871752976091234,
734
+ "grad_norm": 2.67990218211766,
735
+ "learning_rate": 2.8765966340139892e-05,
736
+ "loss": 0.6671,
737
+ "step": 1040
738
+ },
739
+ {
740
+ "epoch": 0.14005135216245956,
741
+ "grad_norm": 2.699521523924172,
742
+ "learning_rate": 2.8740623860279116e-05,
743
+ "loss": 0.6763,
744
+ "step": 1050
745
+ },
746
+ {
747
+ "epoch": 0.14138517456400682,
748
+ "grad_norm": 4.1129898011507535,
749
+ "learning_rate": 2.871503519425044e-05,
750
+ "loss": 0.7159,
751
+ "step": 1060
752
+ },
753
+ {
754
+ "epoch": 0.14271899696555404,
755
+ "grad_norm": 2.4592021333659146,
756
+ "learning_rate": 2.8689200800517448e-05,
757
+ "loss": 0.6551,
758
+ "step": 1070
759
+ },
760
+ {
761
+ "epoch": 0.14405281936710126,
762
+ "grad_norm": 5.138500389099849,
763
+ "learning_rate": 2.866312114194634e-05,
764
+ "loss": 0.7214,
765
+ "step": 1080
766
+ },
767
+ {
768
+ "epoch": 0.14538664176864852,
769
+ "grad_norm": 2.822433730666048,
770
+ "learning_rate": 2.8636796685797657e-05,
771
+ "loss": 0.6862,
772
+ "step": 1090
773
+ },
774
+ {
775
+ "epoch": 0.14672046417019574,
776
+ "grad_norm": 3.086468537427806,
777
+ "learning_rate": 2.8610227903717876e-05,
778
+ "loss": 0.6784,
779
+ "step": 1100
780
+ },
781
+ {
782
+ "epoch": 0.14805428657174297,
783
+ "grad_norm": 2.079766793749202,
784
+ "learning_rate": 2.8583415271730994e-05,
785
+ "loss": 0.7065,
786
+ "step": 1110
787
+ },
788
+ {
789
+ "epoch": 0.14938810897329022,
790
+ "grad_norm": 1.659870509072264,
791
+ "learning_rate": 2.855635927022998e-05,
792
+ "loss": 0.7197,
793
+ "step": 1120
794
+ },
795
+ {
796
+ "epoch": 0.15072193137483744,
797
+ "grad_norm": 7.870626779339635,
798
+ "learning_rate": 2.8529060383968175e-05,
799
+ "loss": 0.7305,
800
+ "step": 1130
801
+ },
802
+ {
803
+ "epoch": 0.15205575377638467,
804
+ "grad_norm": 3.0600340899893537,
805
+ "learning_rate": 2.850151910205061e-05,
806
+ "loss": 0.6922,
807
+ "step": 1140
808
+ },
809
+ {
810
+ "epoch": 0.15338957617793192,
811
+ "grad_norm": 3.6147451373702806,
812
+ "learning_rate": 2.847373591792523e-05,
813
+ "loss": 0.7044,
814
+ "step": 1150
815
+ },
816
+ {
817
+ "epoch": 0.15472339857947914,
818
+ "grad_norm": 4.740777951553679,
819
+ "learning_rate": 2.844571132937407e-05,
820
+ "loss": 0.6794,
821
+ "step": 1160
822
+ },
823
+ {
824
+ "epoch": 0.15605722098102637,
825
+ "grad_norm": 3.377522973717319,
826
+ "learning_rate": 2.841744583850431e-05,
827
+ "loss": 0.673,
828
+ "step": 1170
829
+ },
830
+ {
831
+ "epoch": 0.15739104338257362,
832
+ "grad_norm": 4.250656077289992,
833
+ "learning_rate": 2.838893995173932e-05,
834
+ "loss": 0.6975,
835
+ "step": 1180
836
+ },
837
+ {
838
+ "epoch": 0.15872486578412084,
839
+ "grad_norm": 11.73693900915769,
840
+ "learning_rate": 2.836019417980955e-05,
841
+ "loss": 0.6572,
842
+ "step": 1190
843
+ },
844
+ {
845
+ "epoch": 0.16005868818566807,
846
+ "grad_norm": 2.729291714043308,
847
+ "learning_rate": 2.8331209037743387e-05,
848
+ "loss": 0.7247,
849
+ "step": 1200
850
+ }
851
+ ],
852
+ "logging_steps": 10,
853
+ "max_steps": 7497,
854
+ "num_input_tokens_seen": 0,
855
+ "num_train_epochs": 1,
856
+ "save_steps": 400,
857
+ "stateful_callbacks": {
858
+ "TrainerControl": {
859
+ "args": {
860
+ "should_epoch_stop": false,
861
+ "should_evaluate": false,
862
+ "should_log": false,
863
+ "should_save": true,
864
+ "should_training_stop": false
865
+ },
866
+ "attributes": {}
867
+ }
868
+ },
869
+ "total_flos": 3.280284708293837e+18,
870
+ "train_batch_size": 4,
871
+ "trial_name": null,
872
+ "trial_params": null
873
+ }
checkpoint-1200/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81e448587a2aaa72701b15796f660da2d8aedd670c2a2eaae01d5488a2bf1cfe
3
+ size 6520
checkpoint-1200/zero_to_fp32.py ADDED
@@ -0,0 +1,587 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
215
+ elif zero_stage == 3:
216
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
217
+
218
+
219
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
220
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
221
+ return
222
+
223
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
224
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
225
+
226
+ if debug:
227
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
228
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
229
+
230
+ wanted_params = len(frozen_param_shapes)
231
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
232
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
233
+ print(f'Frozen params: Have {avail_numel} numels to process.')
234
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
235
+
236
+ total_params = 0
237
+ total_numel = 0
238
+ for name, shape in frozen_param_shapes.items():
239
+ total_params += 1
240
+ unpartitioned_numel = shape.numel()
241
+ total_numel += unpartitioned_numel
242
+
243
+ state_dict[name] = frozen_param_fragments[name]
244
+
245
+ if debug:
246
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
247
+
248
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
249
+
250
+
251
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
252
+ param_shapes = zero_model_states[0].param_shapes
253
+
254
+ # Reconstruction protocol:
255
+ #
256
+ # XXX: document this
257
+
258
+ if debug:
259
+ for i in range(world_size):
260
+ for j in range(len(fp32_flat_groups[0])):
261
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
262
+
263
+ # XXX: memory usage doubles here (zero2)
264
+ num_param_groups = len(fp32_flat_groups[0])
265
+ merged_single_partition_of_fp32_groups = []
266
+ for i in range(num_param_groups):
267
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
268
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
269
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
270
+ avail_numel = sum(
271
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
272
+
273
+ if debug:
274
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
275
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
276
+ # not asserting if there is a mismatch due to possible padding
277
+ print(f"Have {avail_numel} numels to process.")
278
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
279
+
280
+ # params
281
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
282
+ # out-of-core computing solution
283
+ total_numel = 0
284
+ total_params = 0
285
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
286
+ offset = 0
287
+ avail_numel = full_single_fp32_vector.numel()
288
+ for name, shape in shapes.items():
289
+
290
+ unpartitioned_numel = shape.numel()
291
+ total_numel += unpartitioned_numel
292
+ total_params += 1
293
+
294
+ if debug:
295
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
296
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
297
+ offset += unpartitioned_numel
298
+
299
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
300
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
301
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
302
+ # live optimizer object, so we are checking that the numbers are within the right range
303
+ align_to = 2 * world_size
304
+
305
+ def zero2_align(x):
306
+ return align_to * math.ceil(x / align_to)
307
+
308
+ if debug:
309
+ print(f"original offset={offset}, avail_numel={avail_numel}")
310
+
311
+ offset = zero2_align(offset)
312
+ avail_numel = zero2_align(avail_numel)
313
+
314
+ if debug:
315
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
316
+
317
+ # Sanity check
318
+ if offset != avail_numel:
319
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
320
+
321
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
322
+
323
+
324
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
325
+ state_dict = OrderedDict()
326
+
327
+ # buffers
328
+ buffers = zero_model_states[0].buffers
329
+ state_dict.update(buffers)
330
+ if debug:
331
+ print(f"added {len(buffers)} buffers")
332
+
333
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
334
+
335
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
336
+
337
+ # recover shared parameters
338
+ for pair in zero_model_states[0].shared_params:
339
+ if pair[1] in state_dict:
340
+ state_dict[pair[0]] = state_dict[pair[1]]
341
+
342
+ return state_dict
343
+
344
+
345
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
346
+ remainder = unpartitioned_numel % world_size
347
+ padding_numel = (world_size - remainder) if remainder else 0
348
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
349
+ return partitioned_numel, padding_numel
350
+
351
+
352
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
353
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
354
+ return
355
+
356
+ if debug:
357
+ for i in range(world_size):
358
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
359
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
360
+
361
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
362
+ wanted_params = len(frozen_param_shapes)
363
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
364
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
365
+ print(f'Frozen params: Have {avail_numel} numels to process.')
366
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
367
+
368
+ total_params = 0
369
+ total_numel = 0
370
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
371
+ total_params += 1
372
+ unpartitioned_numel = shape.numel()
373
+ total_numel += unpartitioned_numel
374
+
375
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
376
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
377
+
378
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
379
+
380
+ if debug:
381
+ print(
382
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
383
+ )
384
+
385
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
386
+
387
+
388
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
389
+ param_shapes = zero_model_states[0].param_shapes
390
+ avail_numel = fp32_flat_groups[0].numel() * world_size
391
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
392
+ # param, re-consolidating each param, while dealing with padding if any
393
+
394
+ # merge list of dicts, preserving order
395
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
396
+
397
+ if debug:
398
+ for i in range(world_size):
399
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
400
+
401
+ wanted_params = len(param_shapes)
402
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
403
+ # not asserting if there is a mismatch due to possible padding
404
+ avail_numel = fp32_flat_groups[0].numel() * world_size
405
+ print(f"Trainable params: Have {avail_numel} numels to process.")
406
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
407
+
408
+ # params
409
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
410
+ # out-of-core computing solution
411
+ offset = 0
412
+ total_numel = 0
413
+ total_params = 0
414
+ for name, shape in param_shapes.items():
415
+
416
+ unpartitioned_numel = shape.numel()
417
+ total_numel += unpartitioned_numel
418
+ total_params += 1
419
+
420
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
421
+
422
+ if debug:
423
+ print(
424
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
425
+ )
426
+
427
+ # XXX: memory usage doubles here
428
+ state_dict[name] = torch.cat(
429
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
430
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
431
+ offset += partitioned_numel
432
+
433
+ offset *= world_size
434
+
435
+ # Sanity check
436
+ if offset != avail_numel:
437
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
438
+
439
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
440
+
441
+
442
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
443
+ state_dict = OrderedDict()
444
+
445
+ # buffers
446
+ buffers = zero_model_states[0].buffers
447
+ state_dict.update(buffers)
448
+ if debug:
449
+ print(f"added {len(buffers)} buffers")
450
+
451
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
452
+
453
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
454
+
455
+ # recover shared parameters
456
+ for pair in zero_model_states[0].shared_params:
457
+ if pair[1] in state_dict:
458
+ state_dict[pair[0]] = state_dict[pair[1]]
459
+
460
+ return state_dict
461
+
462
+
463
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
464
+ """
465
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
466
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
467
+ via a model hub.
468
+
469
+ Args:
470
+ - ``checkpoint_dir``: path to the desired checkpoint folder
471
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
472
+
473
+ Returns:
474
+ - pytorch ``state_dict``
475
+
476
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
477
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
478
+ the checkpoint.
479
+
480
+ A typical usage might be ::
481
+
482
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
483
+ # do the training and checkpoint saving
484
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
485
+ model = model.cpu() # move to cpu
486
+ model.load_state_dict(state_dict)
487
+ # submit to model hub or save the model to share with others
488
+
489
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
490
+ application. i.e. you will need to re-initialize the deepspeed engine, since
491
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
492
+
493
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
494
+
495
+ """
496
+ if tag is None:
497
+ latest_path = os.path.join(checkpoint_dir, 'latest')
498
+ if os.path.isfile(latest_path):
499
+ with open(latest_path, 'r') as fd:
500
+ tag = fd.read().strip()
501
+ else:
502
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
503
+
504
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
505
+
506
+ if not os.path.isdir(ds_checkpoint_dir):
507
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
508
+
509
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
510
+
511
+
512
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
513
+ """
514
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
515
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
516
+
517
+ Args:
518
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
519
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
520
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
521
+ """
522
+
523
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
524
+ print(f"Saving fp32 state dict to {output_file}")
525
+ torch.save(state_dict, output_file)
526
+
527
+
528
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
529
+ """
530
+ 1. Put the provided model to cpu
531
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
532
+ 3. Load it into the provided model
533
+
534
+ Args:
535
+ - ``model``: the model object to update
536
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
537
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
538
+
539
+ Returns:
540
+ - ``model`: modified model
541
+
542
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
543
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
544
+ conveniently placed for you in the checkpoint folder.
545
+
546
+ A typical usage might be ::
547
+
548
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
549
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
550
+ # submit to model hub or save the model to share with others
551
+
552
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
553
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
554
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
555
+
556
+ """
557
+ logger.info(f"Extracting fp32 weights")
558
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
559
+
560
+ logger.info(f"Overwriting model with fp32 weights")
561
+ model = model.cpu()
562
+ model.load_state_dict(state_dict, strict=False)
563
+
564
+ return model
565
+
566
+
567
+ if __name__ == "__main__":
568
+
569
+ parser = argparse.ArgumentParser()
570
+ parser.add_argument("checkpoint_dir",
571
+ type=str,
572
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
573
+ parser.add_argument(
574
+ "output_file",
575
+ type=str,
576
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
577
+ parser.add_argument("-t",
578
+ "--tag",
579
+ type=str,
580
+ default=None,
581
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
582
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
583
+ args = parser.parse_args()
584
+
585
+ debug = args.debug
586
+
587
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)
checkpoint-1600/README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ base_model: Qwen/Qwen-VL-Chat
4
+ ---
5
+
6
+ # Model Card for Model ID
7
+
8
+ <!-- Provide a quick summary of what the model is/does. -->
9
+
10
+
11
+
12
+ ## Model Details
13
+
14
+ ### Model Description
15
+
16
+ <!-- Provide a longer summary of what this model is. -->
17
+
18
+
19
+
20
+ - **Developed by:** [More Information Needed]
21
+ - **Funded by [optional]:** [More Information Needed]
22
+ - **Shared by [optional]:** [More Information Needed]
23
+ - **Model type:** [More Information Needed]
24
+ - **Language(s) (NLP):** [More Information Needed]
25
+ - **License:** [More Information Needed]
26
+ - **Finetuned from model [optional]:** [More Information Needed]
27
+
28
+ ### Model Sources [optional]
29
+
30
+ <!-- Provide the basic links for the model. -->
31
+
32
+ - **Repository:** [More Information Needed]
33
+ - **Paper [optional]:** [More Information Needed]
34
+ - **Demo [optional]:** [More Information Needed]
35
+
36
+ ## Uses
37
+
38
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
+
40
+ ### Direct Use
41
+
42
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
+
44
+ [More Information Needed]
45
+
46
+ ### Downstream Use [optional]
47
+
48
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Out-of-Scope Use
53
+
54
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
+
56
+ [More Information Needed]
57
+
58
+ ## Bias, Risks, and Limitations
59
+
60
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ### Recommendations
65
+
66
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
+
68
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
+
70
+ ## How to Get Started with the Model
71
+
72
+ Use the code below to get started with the model.
73
+
74
+ [More Information Needed]
75
+
76
+ ## Training Details
77
+
78
+ ### Training Data
79
+
80
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
+
82
+ [More Information Needed]
83
+
84
+ ### Training Procedure
85
+
86
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
+
88
+ #### Preprocessing [optional]
89
+
90
+ [More Information Needed]
91
+
92
+
93
+ #### Training Hyperparameters
94
+
95
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
+
97
+ #### Speeds, Sizes, Times [optional]
98
+
99
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
+
101
+ [More Information Needed]
102
+
103
+ ## Evaluation
104
+
105
+ <!-- This section describes the evaluation protocols and provides the results. -->
106
+
107
+ ### Testing Data, Factors & Metrics
108
+
109
+ #### Testing Data
110
+
111
+ <!-- This should link to a Dataset Card if possible. -->
112
+
113
+ [More Information Needed]
114
+
115
+ #### Factors
116
+
117
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Metrics
122
+
123
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
+
125
+ [More Information Needed]
126
+
127
+ ### Results
128
+
129
+ [More Information Needed]
130
+
131
+ #### Summary
132
+
133
+
134
+
135
+ ## Model Examination [optional]
136
+
137
+ <!-- Relevant interpretability work for the model goes here -->
138
+
139
+ [More Information Needed]
140
+
141
+ ## Environmental Impact
142
+
143
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
+
145
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
+
147
+ - **Hardware Type:** [More Information Needed]
148
+ - **Hours used:** [More Information Needed]
149
+ - **Cloud Provider:** [More Information Needed]
150
+ - **Compute Region:** [More Information Needed]
151
+ - **Carbon Emitted:** [More Information Needed]
152
+
153
+ ## Technical Specifications [optional]
154
+
155
+ ### Model Architecture and Objective
156
+
157
+ [More Information Needed]
158
+
159
+ ### Compute Infrastructure
160
+
161
+ [More Information Needed]
162
+
163
+ #### Hardware
164
+
165
+ [More Information Needed]
166
+
167
+ #### Software
168
+
169
+ [More Information Needed]
170
+
171
+ ## Citation [optional]
172
+
173
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
+
175
+ **BibTeX:**
176
+
177
+ [More Information Needed]
178
+
179
+ **APA:**
180
+
181
+ [More Information Needed]
182
+
183
+ ## Glossary [optional]
184
+
185
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
+
187
+ [More Information Needed]
188
+
189
+ ## More Information [optional]
190
+
191
+ [More Information Needed]
192
+
193
+ ## Model Card Authors [optional]
194
+
195
+ [More Information Needed]
196
+
197
+ ## Model Card Contact
198
+
199
+ [More Information Needed]
200
+ ### Framework versions
201
+
202
+ - PEFT 0.10.0
checkpoint-1600/adapter_config.json ADDED
@@ -0,0 +1,380 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "Qwen/Qwen-VL-Chat",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layer_replication": null,
10
+ "layers_pattern": null,
11
+ "layers_to_transform": null,
12
+ "loftq_config": {},
13
+ "lora_alpha": 16,
14
+ "lora_dropout": 0.05,
15
+ "megatron_config": null,
16
+ "megatron_core": "megatron.core",
17
+ "modules_to_save": null,
18
+ "peft_type": "LORA",
19
+ "r": 64,
20
+ "rank_pattern": {},
21
+ "revision": null,
22
+ "target_modules": [
23
+ "transformer.visual.transformer.resblocks.32.mlp.c_fc",
24
+ "transformer.visual.transformer.resblocks.8.attn.in_proj",
25
+ "transformer.visual.transformer.resblocks.16.attn.in_proj",
26
+ "transformer.visual.transformer.resblocks.41.mlp.c_proj",
27
+ "transformer.visual.transformer.resblocks.31.mlp.c_proj",
28
+ "transformer.visual.transformer.resblocks.14.mlp.c_proj",
29
+ "transformer.h.14.mlp.w2",
30
+ "transformer.visual.transformer.resblocks.39.mlp.c_fc",
31
+ "transformer.visual.transformer.resblocks.40.mlp.c_proj",
32
+ "transformer.h.0.attn.c_attn",
33
+ "transformer.visual.conv1",
34
+ "transformer.visual.transformer.resblocks.26.mlp.c_fc",
35
+ "transformer.h.7.mlp.w2",
36
+ "transformer.visual.transformer.resblocks.21.mlp.c_fc",
37
+ "transformer.visual.transformer.resblocks.17.attn.in_proj",
38
+ "transformer.h.29.attn.c_attn",
39
+ "transformer.h.3.attn.c_proj",
40
+ "transformer.visual.transformer.resblocks.0.attn.out_proj",
41
+ "transformer.visual.transformer.resblocks.28.mlp.c_fc",
42
+ "transformer.h.30.attn.c_proj",
43
+ "transformer.h.3.mlp.w2",
44
+ "transformer.h.22.mlp.w1",
45
+ "transformer.visual.transformer.resblocks.11.mlp.c_proj",
46
+ "transformer.h.11.mlp.c_proj",
47
+ "transformer.visual.transformer.resblocks.3.attn.in_proj",
48
+ "transformer.visual.transformer.resblocks.21.attn.out_proj",
49
+ "transformer.visual.transformer.resblocks.33.attn.out_proj",
50
+ "transformer.visual.transformer.resblocks.3.mlp.c_fc",
51
+ "transformer.visual.transformer.resblocks.15.attn.in_proj",
52
+ "transformer.visual.transformer.resblocks.4.attn.out_proj",
53
+ "transformer.visual.transformer.resblocks.34.attn.in_proj",
54
+ "transformer.h.17.mlp.c_proj",
55
+ "transformer.visual.transformer.resblocks.1.mlp.c_proj",
56
+ "transformer.visual.transformer.resblocks.4.mlp.c_proj",
57
+ "transformer.visual.transformer.resblocks.39.attn.out_proj",
58
+ "transformer.h.13.mlp.c_proj",
59
+ "transformer.visual.transformer.resblocks.3.mlp.c_proj",
60
+ "transformer.visual.transformer.resblocks.22.attn.out_proj",
61
+ "transformer.h.27.attn.c_attn",
62
+ "transformer.visual.transformer.resblocks.47.attn.in_proj",
63
+ "transformer.h.1.mlp.c_proj",
64
+ "transformer.h.21.attn.c_attn",
65
+ "transformer.visual.transformer.resblocks.18.attn.out_proj",
66
+ "transformer.visual.transformer.resblocks.5.attn.in_proj",
67
+ "transformer.h.6.attn.c_proj",
68
+ "transformer.visual.transformer.resblocks.17.mlp.c_fc",
69
+ "transformer.h.16.attn.c_attn",
70
+ "transformer.visual.transformer.resblocks.0.attn.in_proj",
71
+ "transformer.visual.transformer.resblocks.22.mlp.c_fc",
72
+ "transformer.visual.transformer.resblocks.27.mlp.c_proj",
73
+ "transformer.h.11.attn.c_attn",
74
+ "transformer.h.22.mlp.w2",
75
+ "transformer.h.8.mlp.w1",
76
+ "transformer.visual.transformer.resblocks.29.mlp.c_proj",
77
+ "transformer.visual.transformer.resblocks.44.attn.out_proj",
78
+ "transformer.h.13.mlp.w2",
79
+ "transformer.visual.transformer.resblocks.19.mlp.c_proj",
80
+ "transformer.visual.transformer.resblocks.27.attn.in_proj",
81
+ "transformer.h.29.mlp.w1",
82
+ "transformer.h.24.mlp.c_proj",
83
+ "transformer.visual.transformer.resblocks.25.mlp.c_proj",
84
+ "transformer.visual.transformer.resblocks.4.attn.in_proj",
85
+ "transformer.visual.transformer.resblocks.42.mlp.c_fc",
86
+ "transformer.h.28.mlp.w2",
87
+ "transformer.visual.transformer.resblocks.20.attn.in_proj",
88
+ "transformer.visual.transformer.resblocks.9.attn.in_proj",
89
+ "transformer.h.10.attn.c_proj",
90
+ "transformer.h.13.attn.c_proj",
91
+ "transformer.visual.transformer.resblocks.26.attn.in_proj",
92
+ "transformer.h.17.attn.c_attn",
93
+ "transformer.visual.transformer.resblocks.7.attn.in_proj",
94
+ "transformer.h.23.attn.c_proj",
95
+ "transformer.visual.transformer.resblocks.38.attn.in_proj",
96
+ "transformer.h.19.attn.c_attn",
97
+ "transformer.h.1.attn.c_proj",
98
+ "transformer.visual.transformer.resblocks.2.attn.in_proj",
99
+ "transformer.h.4.mlp.w2",
100
+ "transformer.h.15.mlp.c_proj",
101
+ "transformer.h.4.mlp.c_proj",
102
+ "transformer.h.19.mlp.w2",
103
+ "transformer.h.12.mlp.w1",
104
+ "transformer.visual.transformer.resblocks.30.attn.in_proj",
105
+ "transformer.visual.transformer.resblocks.9.attn.out_proj",
106
+ "transformer.h.28.mlp.c_proj",
107
+ "transformer.h.1.attn.c_attn",
108
+ "transformer.h.8.attn.c_proj",
109
+ "transformer.visual.transformer.resblocks.40.mlp.c_fc",
110
+ "transformer.visual.transformer.resblocks.25.attn.out_proj",
111
+ "transformer.h.4.mlp.w1",
112
+ "transformer.visual.transformer.resblocks.31.mlp.c_fc",
113
+ "transformer.visual.transformer.resblocks.7.mlp.c_proj",
114
+ "transformer.visual.transformer.resblocks.33.attn.in_proj",
115
+ "transformer.h.0.mlp.w1",
116
+ "transformer.visual.transformer.resblocks.8.mlp.c_proj",
117
+ "transformer.h.20.mlp.w2",
118
+ "transformer.visual.transformer.resblocks.37.mlp.c_fc",
119
+ "transformer.visual.transformer.resblocks.43.attn.in_proj",
120
+ "transformer.visual.transformer.resblocks.37.attn.out_proj",
121
+ "transformer.h.25.mlp.w1",
122
+ "transformer.visual.transformer.resblocks.2.mlp.c_proj",
123
+ "transformer.h.27.mlp.w2",
124
+ "transformer.visual.transformer.resblocks.47.attn.out_proj",
125
+ "transformer.visual.transformer.resblocks.18.attn.in_proj",
126
+ "transformer.visual.transformer.resblocks.12.mlp.c_fc",
127
+ "transformer.visual.transformer.resblocks.11.attn.in_proj",
128
+ "transformer.h.14.mlp.c_proj",
129
+ "transformer.h.7.attn.c_attn",
130
+ "transformer.h.10.mlp.w2",
131
+ "transformer.h.11.mlp.w2",
132
+ "transformer.visual.transformer.resblocks.13.attn.out_proj",
133
+ "transformer.h.24.mlp.w1",
134
+ "transformer.h.0.mlp.c_proj",
135
+ "transformer.h.24.attn.c_proj",
136
+ "transformer.visual.transformer.resblocks.10.mlp.c_fc",
137
+ "transformer.visual.transformer.resblocks.33.mlp.c_proj",
138
+ "transformer.visual.transformer.resblocks.35.attn.in_proj",
139
+ "transformer.visual.transformer.resblocks.38.attn.out_proj",
140
+ "transformer.h.2.mlp.w2",
141
+ "transformer.visual.transformer.resblocks.40.attn.in_proj",
142
+ "transformer.h.25.mlp.w2",
143
+ "transformer.visual.transformer.resblocks.17.mlp.c_proj",
144
+ "transformer.visual.transformer.resblocks.18.mlp.c_proj",
145
+ "transformer.h.2.attn.c_proj",
146
+ "transformer.visual.transformer.resblocks.0.mlp.c_proj",
147
+ "transformer.visual.transformer.resblocks.23.mlp.c_proj",
148
+ "transformer.visual.transformer.resblocks.2.attn.out_proj",
149
+ "transformer.h.16.mlp.w2",
150
+ "transformer.h.29.mlp.c_proj",
151
+ "transformer.visual.transformer.resblocks.21.attn.in_proj",
152
+ "transformer.visual.transformer.resblocks.7.mlp.c_fc",
153
+ "transformer.visual.transformer.resblocks.8.attn.out_proj",
154
+ "transformer.visual.transformer.resblocks.28.attn.in_proj",
155
+ "transformer.h.11.mlp.w1",
156
+ "transformer.visual.transformer.resblocks.41.mlp.c_fc",
157
+ "transformer.h.18.attn.c_proj",
158
+ "transformer.visual.transformer.resblocks.29.mlp.c_fc",
159
+ "transformer.visual.transformer.resblocks.32.mlp.c_proj",
160
+ "transformer.visual.transformer.resblocks.28.attn.out_proj",
161
+ "transformer.h.6.attn.c_attn",
162
+ "transformer.visual.transformer.resblocks.3.attn.out_proj",
163
+ "transformer.h.25.attn.c_attn",
164
+ "transformer.h.28.attn.c_proj",
165
+ "transformer.visual.transformer.resblocks.16.attn.out_proj",
166
+ "transformer.visual.transformer.resblocks.20.mlp.c_proj",
167
+ "transformer.visual.transformer.resblocks.45.attn.in_proj",
168
+ "transformer.h.8.mlp.c_proj",
169
+ "transformer.visual.transformer.resblocks.29.attn.out_proj",
170
+ "transformer.h.27.mlp.c_proj",
171
+ "transformer.visual.transformer.resblocks.11.mlp.c_fc",
172
+ "transformer.visual.transformer.resblocks.11.attn.out_proj",
173
+ "transformer.visual.transformer.resblocks.10.attn.in_proj",
174
+ "transformer.visual.transformer.resblocks.27.mlp.c_fc",
175
+ "transformer.h.4.attn.c_proj",
176
+ "transformer.visual.transformer.resblocks.43.mlp.c_fc",
177
+ "transformer.visual.transformer.resblocks.22.mlp.c_proj",
178
+ "transformer.h.22.attn.c_proj",
179
+ "transformer.visual.transformer.resblocks.5.mlp.c_proj",
180
+ "transformer.h.22.mlp.c_proj",
181
+ "transformer.visual.transformer.resblocks.7.attn.out_proj",
182
+ "transformer.visual.transformer.resblocks.14.attn.out_proj",
183
+ "transformer.h.30.mlp.w1",
184
+ "transformer.h.14.attn.c_attn",
185
+ "transformer.h.4.attn.c_attn",
186
+ "transformer.visual.transformer.resblocks.46.mlp.c_fc",
187
+ "transformer.visual.transformer.resblocks.27.attn.out_proj",
188
+ "transformer.visual.transformer.resblocks.41.attn.in_proj",
189
+ "transformer.visual.transformer.resblocks.13.mlp.c_fc",
190
+ "transformer.h.12.attn.c_proj",
191
+ "transformer.visual.transformer.resblocks.32.attn.out_proj",
192
+ "transformer.visual.transformer.resblocks.18.mlp.c_fc",
193
+ "transformer.visual.transformer.resblocks.45.attn.out_proj",
194
+ "transformer.visual.transformer.resblocks.44.mlp.c_proj",
195
+ "transformer.visual.transformer.resblocks.35.mlp.c_proj",
196
+ "transformer.visual.transformer.resblocks.20.attn.out_proj",
197
+ "transformer.visual.transformer.resblocks.19.attn.in_proj",
198
+ "transformer.h.17.mlp.w2",
199
+ "transformer.h.10.attn.c_attn",
200
+ "transformer.visual.transformer.resblocks.45.mlp.c_proj",
201
+ "transformer.visual.transformer.resblocks.39.mlp.c_proj",
202
+ "transformer.h.7.mlp.w1",
203
+ "transformer.h.14.mlp.w1",
204
+ "transformer.visual.transformer.resblocks.2.mlp.c_fc",
205
+ "transformer.h.21.mlp.w1",
206
+ "transformer.visual.transformer.resblocks.30.attn.out_proj",
207
+ "transformer.h.19.mlp.c_proj",
208
+ "transformer.h.5.mlp.w1",
209
+ "transformer.visual.transformer.resblocks.47.mlp.c_fc",
210
+ "transformer.visual.transformer.resblocks.38.mlp.c_fc",
211
+ "transformer.h.10.mlp.c_proj",
212
+ "transformer.visual.transformer.resblocks.28.mlp.c_proj",
213
+ "transformer.visual.transformer.resblocks.25.attn.in_proj",
214
+ "transformer.h.9.attn.c_attn",
215
+ "transformer.visual.transformer.resblocks.0.mlp.c_fc",
216
+ "transformer.h.29.attn.c_proj",
217
+ "transformer.h.5.mlp.w2",
218
+ "transformer.h.30.attn.c_attn",
219
+ "transformer.h.1.mlp.w1",
220
+ "transformer.visual.transformer.resblocks.8.mlp.c_fc",
221
+ "transformer.h.19.mlp.w1",
222
+ "transformer.h.18.attn.c_attn",
223
+ "transformer.h.11.attn.c_proj",
224
+ "transformer.h.5.mlp.c_proj",
225
+ "transformer.visual.transformer.resblocks.45.mlp.c_fc",
226
+ "transformer.visual.transformer.resblocks.47.mlp.c_proj",
227
+ "transformer.h.9.attn.c_proj",
228
+ "transformer.visual.transformer.resblocks.6.attn.out_proj",
229
+ "transformer.h.26.mlp.c_proj",
230
+ "transformer.visual.transformer.resblocks.24.mlp.c_proj",
231
+ "transformer.h.31.attn.c_attn",
232
+ "transformer.h.13.mlp.w1",
233
+ "transformer.visual.transformer.resblocks.39.attn.in_proj",
234
+ "transformer.h.20.mlp.w1",
235
+ "transformer.visual.transformer.resblocks.22.attn.in_proj",
236
+ "transformer.visual.transformer.resblocks.35.mlp.c_fc",
237
+ "transformer.visual.transformer.resblocks.24.attn.in_proj",
238
+ "transformer.h.16.mlp.w1",
239
+ "transformer.visual.transformer.resblocks.6.mlp.c_proj",
240
+ "transformer.h.6.mlp.c_proj",
241
+ "transformer.visual.transformer.resblocks.26.mlp.c_proj",
242
+ "transformer.visual.transformer.resblocks.35.attn.out_proj",
243
+ "transformer.h.24.mlp.w2",
244
+ "transformer.visual.transformer.resblocks.31.attn.in_proj",
245
+ "transformer.visual.transformer.resblocks.15.mlp.c_fc",
246
+ "transformer.h.2.mlp.w1",
247
+ "transformer.h.31.mlp.c_proj",
248
+ "transformer.h.13.attn.c_attn",
249
+ "transformer.visual.transformer.resblocks.14.attn.in_proj",
250
+ "transformer.h.12.mlp.w2",
251
+ "transformer.visual.transformer.resblocks.34.mlp.c_proj",
252
+ "transformer.h.26.mlp.w2",
253
+ "transformer.visual.transformer.resblocks.5.mlp.c_fc",
254
+ "transformer.h.5.attn.c_proj",
255
+ "transformer.h.9.mlp.w2",
256
+ "transformer.h.15.mlp.w2",
257
+ "transformer.h.12.attn.c_attn",
258
+ "transformer.visual.transformer.resblocks.10.mlp.c_proj",
259
+ "transformer.h.28.mlp.w1",
260
+ "transformer.visual.transformer.resblocks.46.attn.out_proj",
261
+ "transformer.h.18.mlp.c_proj",
262
+ "transformer.h.15.attn.c_proj",
263
+ "transformer.visual.transformer.resblocks.44.attn.in_proj",
264
+ "transformer.visual.transformer.resblocks.12.attn.out_proj",
265
+ "transformer.visual.transformer.resblocks.1.mlp.c_fc",
266
+ "transformer.visual.transformer.resblocks.15.attn.out_proj",
267
+ "transformer.h.17.mlp.w1",
268
+ "transformer.visual.transformer.resblocks.46.attn.in_proj",
269
+ "transformer.h.2.attn.c_attn",
270
+ "transformer.h.25.attn.c_proj",
271
+ "transformer.h.14.attn.c_proj",
272
+ "transformer.h.26.attn.c_proj",
273
+ "transformer.h.31.mlp.w1",
274
+ "transformer.h.23.mlp.w2",
275
+ "transformer.visual.transformer.resblocks.26.attn.out_proj",
276
+ "transformer.h.20.mlp.c_proj",
277
+ "transformer.visual.transformer.resblocks.6.mlp.c_fc",
278
+ "transformer.h.27.mlp.w1",
279
+ "transformer.h.7.attn.c_proj",
280
+ "transformer.visual.transformer.resblocks.15.mlp.c_proj",
281
+ "transformer.h.16.mlp.c_proj",
282
+ "transformer.visual.transformer.resblocks.12.mlp.c_proj",
283
+ "transformer.h.29.mlp.w2",
284
+ "transformer.h.15.mlp.w1",
285
+ "transformer.h.6.mlp.w2",
286
+ "transformer.h.3.attn.c_attn",
287
+ "transformer.h.21.mlp.w2",
288
+ "transformer.visual.transformer.resblocks.16.mlp.c_fc",
289
+ "transformer.visual.transformer.resblocks.23.mlp.c_fc",
290
+ "transformer.visual.transformer.resblocks.34.attn.out_proj",
291
+ "transformer.h.8.attn.c_attn",
292
+ "transformer.visual.transformer.resblocks.43.mlp.c_proj",
293
+ "transformer.visual.transformer.resblocks.14.mlp.c_fc",
294
+ "transformer.visual.transformer.resblocks.24.attn.out_proj",
295
+ "transformer.visual.transformer.resblocks.12.attn.in_proj",
296
+ "transformer.h.25.mlp.c_proj",
297
+ "transformer.visual.transformer.resblocks.46.mlp.c_proj",
298
+ "transformer.h.7.mlp.c_proj",
299
+ "transformer.h.15.attn.c_attn",
300
+ "transformer.visual.transformer.resblocks.36.attn.in_proj",
301
+ "transformer.h.26.attn.c_attn",
302
+ "transformer.h.0.attn.c_proj",
303
+ "transformer.visual.transformer.resblocks.9.mlp.c_fc",
304
+ "transformer.h.19.attn.c_proj",
305
+ "transformer.visual.transformer.resblocks.29.attn.in_proj",
306
+ "transformer.visual.transformer.resblocks.13.attn.in_proj",
307
+ "transformer.visual.transformer.resblocks.23.attn.in_proj",
308
+ "transformer.visual.transformer.resblocks.36.attn.out_proj",
309
+ "transformer.h.3.mlp.c_proj",
310
+ "transformer.h.27.attn.c_proj",
311
+ "transformer.h.31.attn.c_proj",
312
+ "transformer.visual.transformer.resblocks.19.attn.out_proj",
313
+ "transformer.h.0.mlp.w2",
314
+ "transformer.h.17.attn.c_proj",
315
+ "transformer.h.30.mlp.w2",
316
+ "transformer.visual.transformer.resblocks.24.mlp.c_fc",
317
+ "transformer.h.28.attn.c_attn",
318
+ "transformer.visual.transformer.resblocks.16.mlp.c_proj",
319
+ "transformer.visual.transformer.resblocks.36.mlp.c_fc",
320
+ "transformer.h.30.mlp.c_proj",
321
+ "transformer.visual.transformer.resblocks.37.attn.in_proj",
322
+ "transformer.visual.transformer.resblocks.20.mlp.c_fc",
323
+ "transformer.h.9.mlp.c_proj",
324
+ "transformer.visual.transformer.resblocks.43.attn.out_proj",
325
+ "transformer.visual.transformer.resblocks.42.attn.in_proj",
326
+ "transformer.h.1.mlp.w2",
327
+ "transformer.h.6.mlp.w1",
328
+ "transformer.visual.transformer.resblocks.17.attn.out_proj",
329
+ "transformer.visual.transformer.resblocks.33.mlp.c_fc",
330
+ "transformer.h.5.attn.c_attn",
331
+ "transformer.h.8.mlp.w2",
332
+ "transformer.h.23.mlp.c_proj",
333
+ "transformer.h.20.attn.c_attn",
334
+ "transformer.visual.transformer.resblocks.4.mlp.c_fc",
335
+ "transformer.h.31.mlp.w2",
336
+ "transformer.h.9.mlp.w1",
337
+ "transformer.h.12.mlp.c_proj",
338
+ "transformer.visual.transformer.resblocks.1.attn.out_proj",
339
+ "transformer.visual.transformer.resblocks.30.mlp.c_fc",
340
+ "transformer.visual.transformer.resblocks.1.attn.in_proj",
341
+ "transformer.visual.transformer.resblocks.32.attn.in_proj",
342
+ "transformer.h.16.attn.c_proj",
343
+ "transformer.visual.transformer.resblocks.42.mlp.c_proj",
344
+ "transformer.h.3.mlp.w1",
345
+ "transformer.visual.transformer.resblocks.9.mlp.c_proj",
346
+ "transformer.h.18.mlp.w1",
347
+ "transformer.visual.transformer.resblocks.44.mlp.c_fc",
348
+ "transformer.h.21.mlp.c_proj",
349
+ "transformer.visual.transformer.resblocks.5.attn.out_proj",
350
+ "transformer.visual.transformer.resblocks.34.mlp.c_fc",
351
+ "transformer.visual.transformer.resblocks.42.attn.out_proj",
352
+ "transformer.visual.transformer.resblocks.41.attn.out_proj",
353
+ "transformer.h.10.mlp.w1",
354
+ "transformer.visual.transformer.resblocks.19.mlp.c_fc",
355
+ "transformer.visual.transformer.resblocks.13.mlp.c_proj",
356
+ "transformer.h.21.attn.c_proj",
357
+ "transformer.visual.transformer.resblocks.6.attn.in_proj",
358
+ "transformer.visual.transformer.resblocks.23.attn.out_proj",
359
+ "transformer.visual.transformer.resblocks.40.attn.out_proj",
360
+ "transformer.visual.transformer.resblocks.36.mlp.c_proj",
361
+ "transformer.h.2.mlp.c_proj",
362
+ "transformer.visual.transformer.resblocks.25.mlp.c_fc",
363
+ "transformer.h.22.attn.c_attn",
364
+ "transformer.h.23.mlp.w1",
365
+ "transformer.h.20.attn.c_proj",
366
+ "transformer.h.23.attn.c_attn",
367
+ "transformer.visual.transformer.resblocks.37.mlp.c_proj",
368
+ "transformer.h.26.mlp.w1",
369
+ "transformer.h.18.mlp.w2",
370
+ "transformer.visual.transformer.resblocks.21.mlp.c_proj",
371
+ "transformer.visual.transformer.resblocks.31.attn.out_proj",
372
+ "transformer.h.24.attn.c_attn",
373
+ "transformer.visual.transformer.resblocks.10.attn.out_proj",
374
+ "transformer.visual.transformer.resblocks.30.mlp.c_proj",
375
+ "transformer.visual.transformer.resblocks.38.mlp.c_proj"
376
+ ],
377
+ "task_type": "CAUSAL_LM",
378
+ "use_dora": false,
379
+ "use_rslora": false
380
+ }
checkpoint-1600/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2e8f02b580a83f9a0dcf2c64cdd40b6224ce4d0c5f7b6bf5c54252469b746fd
3
+ size 469105640
checkpoint-1600/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step1600
checkpoint-1600/qwen.tiktoken ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-1600/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7867300d493c7c70556f317580df469210465c281860ed7afa76d8cc69360a49
3
+ size 14960
checkpoint-1600/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4a69fe0cdc1dde42061be6b4ada5179bad5abd7cb5f1a162b97c6c328a15b49
3
+ size 14960
checkpoint-1600/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f324d2b3935846b844c4dc221df4ac446a4fb48bd06ecca758c0763ce6ce627a
3
+ size 14960
checkpoint-1600/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7150b8ce206b4a26f42a865a11c03f659908080e31492b757e58b801f9fffbc5
3
+ size 14960
checkpoint-1600/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b92b7d41119c1f55a67eb2522899e6a677806ed9022c1df91778b1c215870a1
3
+ size 1064
checkpoint-1600/special_tokens_map.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "pad_token": "<|endoftext|>"
3
+ }
checkpoint-1600/tokenizer_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {},
3
+ "auto_map": {
4
+ "AutoTokenizer": [
5
+ "Qwen/Qwen-VL-Chat--tokenization_qwen.QWenTokenizer",
6
+ null
7
+ ]
8
+ },
9
+ "clean_up_tokenization_spaces": true,
10
+ "model_max_length": 768,
11
+ "pad_token": "<|endoftext|>",
12
+ "padding_side": "right",
13
+ "tokenizer_class": "QWenTokenizer"
14
+ }
checkpoint-1600/trainer_state.json ADDED
@@ -0,0 +1,1153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.21341158424755743,
5
+ "eval_steps": 500,
6
+ "global_step": 1600,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.001333822401547234,
13
+ "grad_norm": 5.80256772259428,
14
+ "learning_rate": 4e-06,
15
+ "loss": 1.0498,
16
+ "step": 10
17
+ },
18
+ {
19
+ "epoch": 0.002667644803094468,
20
+ "grad_norm": 33.895696082107904,
21
+ "learning_rate": 8e-06,
22
+ "loss": 1.0653,
23
+ "step": 20
24
+ },
25
+ {
26
+ "epoch": 0.004001467204641702,
27
+ "grad_norm": 5.523348234283539,
28
+ "learning_rate": 1.2e-05,
29
+ "loss": 1.0341,
30
+ "step": 30
31
+ },
32
+ {
33
+ "epoch": 0.005335289606188936,
34
+ "grad_norm": 11.1556403156453,
35
+ "learning_rate": 1.6e-05,
36
+ "loss": 0.9692,
37
+ "step": 40
38
+ },
39
+ {
40
+ "epoch": 0.00666911200773617,
41
+ "grad_norm": 3.7375231126561825,
42
+ "learning_rate": 1.9999999999999998e-05,
43
+ "loss": 0.9554,
44
+ "step": 50
45
+ },
46
+ {
47
+ "epoch": 0.008002934409283404,
48
+ "grad_norm": 8.43538339698909,
49
+ "learning_rate": 2.4e-05,
50
+ "loss": 0.8965,
51
+ "step": 60
52
+ },
53
+ {
54
+ "epoch": 0.009336756810830639,
55
+ "grad_norm": 13.403454896011478,
56
+ "learning_rate": 2.8e-05,
57
+ "loss": 0.8273,
58
+ "step": 70
59
+ },
60
+ {
61
+ "epoch": 0.010670579212377872,
62
+ "grad_norm": 3.95522050766088,
63
+ "learning_rate": 2.9999966406213696e-05,
64
+ "loss": 0.7837,
65
+ "step": 80
66
+ },
67
+ {
68
+ "epoch": 0.012004401613925107,
69
+ "grad_norm": 36.799552052300854,
70
+ "learning_rate": 2.9999697656826056e-05,
71
+ "loss": 0.8288,
72
+ "step": 90
73
+ },
74
+ {
75
+ "epoch": 0.01333822401547234,
76
+ "grad_norm": 1.6305479563258536,
77
+ "learning_rate": 2.9999160162865885e-05,
78
+ "loss": 0.7778,
79
+ "step": 100
80
+ },
81
+ {
82
+ "epoch": 0.014672046417019574,
83
+ "grad_norm": 2.159536648784889,
84
+ "learning_rate": 2.9998353933963273e-05,
85
+ "loss": 0.7616,
86
+ "step": 110
87
+ },
88
+ {
89
+ "epoch": 0.016005868818566808,
90
+ "grad_norm": 3.397321425707004,
91
+ "learning_rate": 2.999727898456315e-05,
92
+ "loss": 0.7594,
93
+ "step": 120
94
+ },
95
+ {
96
+ "epoch": 0.017339691220114042,
97
+ "grad_norm": 4.772220837365037,
98
+ "learning_rate": 2.999593533392503e-05,
99
+ "loss": 0.756,
100
+ "step": 130
101
+ },
102
+ {
103
+ "epoch": 0.018673513621661277,
104
+ "grad_norm": 2.4845945633126885,
105
+ "learning_rate": 2.9994323006122654e-05,
106
+ "loss": 0.7601,
107
+ "step": 140
108
+ },
109
+ {
110
+ "epoch": 0.02000733602320851,
111
+ "grad_norm": 3.591682569169127,
112
+ "learning_rate": 2.9992442030043557e-05,
113
+ "loss": 0.7894,
114
+ "step": 150
115
+ },
116
+ {
117
+ "epoch": 0.021341158424755743,
118
+ "grad_norm": 2.5679458807474416,
119
+ "learning_rate": 2.9990292439388565e-05,
120
+ "loss": 0.7093,
121
+ "step": 160
122
+ },
123
+ {
124
+ "epoch": 0.022674980826302978,
125
+ "grad_norm": 1.9412569107551652,
126
+ "learning_rate": 2.9987874272671168e-05,
127
+ "loss": 0.706,
128
+ "step": 170
129
+ },
130
+ {
131
+ "epoch": 0.024008803227850213,
132
+ "grad_norm": 3.2667097270489,
133
+ "learning_rate": 2.9985187573216855e-05,
134
+ "loss": 0.7586,
135
+ "step": 180
136
+ },
137
+ {
138
+ "epoch": 0.025342625629397444,
139
+ "grad_norm": 4.4208737375400675,
140
+ "learning_rate": 2.998223238916232e-05,
141
+ "loss": 0.6985,
142
+ "step": 190
143
+ },
144
+ {
145
+ "epoch": 0.02667644803094468,
146
+ "grad_norm": 5.515966302183704,
147
+ "learning_rate": 2.9979008773454618e-05,
148
+ "loss": 0.7323,
149
+ "step": 200
150
+ },
151
+ {
152
+ "epoch": 0.028010270432491914,
153
+ "grad_norm": 2.964165450396077,
154
+ "learning_rate": 2.997551678385019e-05,
155
+ "loss": 0.7603,
156
+ "step": 210
157
+ },
158
+ {
159
+ "epoch": 0.02934409283403915,
160
+ "grad_norm": 3.0952916783456197,
161
+ "learning_rate": 2.997175648291384e-05,
162
+ "loss": 0.7421,
163
+ "step": 220
164
+ },
165
+ {
166
+ "epoch": 0.03067791523558638,
167
+ "grad_norm": 4.213588693904103,
168
+ "learning_rate": 2.996772793801763e-05,
169
+ "loss": 0.7322,
170
+ "step": 230
171
+ },
172
+ {
173
+ "epoch": 0.032011737637133615,
174
+ "grad_norm": 1.8568586103139084,
175
+ "learning_rate": 2.996343122133965e-05,
176
+ "loss": 0.6922,
177
+ "step": 240
178
+ },
179
+ {
180
+ "epoch": 0.033345560038680847,
181
+ "grad_norm": 4.494146778909846,
182
+ "learning_rate": 2.9958866409862745e-05,
183
+ "loss": 0.7244,
184
+ "step": 250
185
+ },
186
+ {
187
+ "epoch": 0.034679382440228085,
188
+ "grad_norm": 7.438170074282725,
189
+ "learning_rate": 2.9954033585373108e-05,
190
+ "loss": 0.7093,
191
+ "step": 260
192
+ },
193
+ {
194
+ "epoch": 0.036013204841775316,
195
+ "grad_norm": 2.3744787346857015,
196
+ "learning_rate": 2.994893283445885e-05,
197
+ "loss": 0.6983,
198
+ "step": 270
199
+ },
200
+ {
201
+ "epoch": 0.037347027243322554,
202
+ "grad_norm": 1.4722011682616383,
203
+ "learning_rate": 2.9943564248508415e-05,
204
+ "loss": 0.6781,
205
+ "step": 280
206
+ },
207
+ {
208
+ "epoch": 0.038680849644869786,
209
+ "grad_norm": 3.3397620832486075,
210
+ "learning_rate": 2.9937927923708966e-05,
211
+ "loss": 0.7399,
212
+ "step": 290
213
+ },
214
+ {
215
+ "epoch": 0.04001467204641702,
216
+ "grad_norm": 5.05063397044549,
217
+ "learning_rate": 2.993202396104465e-05,
218
+ "loss": 0.7671,
219
+ "step": 300
220
+ },
221
+ {
222
+ "epoch": 0.041348494447964255,
223
+ "grad_norm": 3.0128431385936767,
224
+ "learning_rate": 2.9925852466294795e-05,
225
+ "loss": 0.7015,
226
+ "step": 310
227
+ },
228
+ {
229
+ "epoch": 0.04268231684951149,
230
+ "grad_norm": 2.0161342716764237,
231
+ "learning_rate": 2.9919413550032014e-05,
232
+ "loss": 0.7009,
233
+ "step": 320
234
+ },
235
+ {
236
+ "epoch": 0.04401613925105872,
237
+ "grad_norm": 1.3114004070324985,
238
+ "learning_rate": 2.991270732762022e-05,
239
+ "loss": 0.7153,
240
+ "step": 330
241
+ },
242
+ {
243
+ "epoch": 0.045349961652605957,
244
+ "grad_norm": 18.493625676806268,
245
+ "learning_rate": 2.990573391921255e-05,
246
+ "loss": 0.7518,
247
+ "step": 340
248
+ },
249
+ {
250
+ "epoch": 0.04668378405415319,
251
+ "grad_norm": 2.9526764059703567,
252
+ "learning_rate": 2.989849344974924e-05,
253
+ "loss": 0.7133,
254
+ "step": 350
255
+ },
256
+ {
257
+ "epoch": 0.048017606455700426,
258
+ "grad_norm": 5.26274958582726,
259
+ "learning_rate": 2.9890986048955368e-05,
260
+ "loss": 0.7139,
261
+ "step": 360
262
+ },
263
+ {
264
+ "epoch": 0.04935142885724766,
265
+ "grad_norm": 3.5319788357887933,
266
+ "learning_rate": 2.9883211851338516e-05,
267
+ "loss": 0.7084,
268
+ "step": 370
269
+ },
270
+ {
271
+ "epoch": 0.05068525125879489,
272
+ "grad_norm": 7.607269935902469,
273
+ "learning_rate": 2.9875170996186392e-05,
274
+ "loss": 0.7309,
275
+ "step": 380
276
+ },
277
+ {
278
+ "epoch": 0.05201907366034213,
279
+ "grad_norm": 2.3456663308287253,
280
+ "learning_rate": 2.986686362756431e-05,
281
+ "loss": 0.6827,
282
+ "step": 390
283
+ },
284
+ {
285
+ "epoch": 0.05335289606188936,
286
+ "grad_norm": 2.176182050789012,
287
+ "learning_rate": 2.9858289894312617e-05,
288
+ "loss": 0.6995,
289
+ "step": 400
290
+ },
291
+ {
292
+ "epoch": 0.0546867184634366,
293
+ "grad_norm": 11.171630173781537,
294
+ "learning_rate": 2.9849449950044036e-05,
295
+ "loss": 0.7335,
296
+ "step": 410
297
+ },
298
+ {
299
+ "epoch": 0.05602054086498383,
300
+ "grad_norm": 6.63441431767892,
301
+ "learning_rate": 2.984034395314088e-05,
302
+ "loss": 0.7031,
303
+ "step": 420
304
+ },
305
+ {
306
+ "epoch": 0.05735436326653106,
307
+ "grad_norm": 2.861620412225736,
308
+ "learning_rate": 2.983097206675227e-05,
309
+ "loss": 0.6559,
310
+ "step": 430
311
+ },
312
+ {
313
+ "epoch": 0.0586881856680783,
314
+ "grad_norm": 5.523165036486206,
315
+ "learning_rate": 2.9821334458791156e-05,
316
+ "loss": 0.726,
317
+ "step": 440
318
+ },
319
+ {
320
+ "epoch": 0.06002200806962553,
321
+ "grad_norm": 3.5602243751368197,
322
+ "learning_rate": 2.9811431301931344e-05,
323
+ "loss": 0.7202,
324
+ "step": 450
325
+ },
326
+ {
327
+ "epoch": 0.06135583047117276,
328
+ "grad_norm": 11.333380381168622,
329
+ "learning_rate": 2.9801262773604377e-05,
330
+ "loss": 0.7189,
331
+ "step": 460
332
+ },
333
+ {
334
+ "epoch": 0.06268965287271999,
335
+ "grad_norm": 14.159758615106613,
336
+ "learning_rate": 2.9790829055996398e-05,
337
+ "loss": 0.7267,
338
+ "step": 470
339
+ },
340
+ {
341
+ "epoch": 0.06402347527426723,
342
+ "grad_norm": 9.009079485918289,
343
+ "learning_rate": 2.978013033604483e-05,
344
+ "loss": 0.748,
345
+ "step": 480
346
+ },
347
+ {
348
+ "epoch": 0.06535729767581447,
349
+ "grad_norm": 1.9682648681675994,
350
+ "learning_rate": 2.976916680543506e-05,
351
+ "loss": 0.7369,
352
+ "step": 490
353
+ },
354
+ {
355
+ "epoch": 0.06669112007736169,
356
+ "grad_norm": 2.9278164598232777,
357
+ "learning_rate": 2.975793866059701e-05,
358
+ "loss": 0.7037,
359
+ "step": 500
360
+ },
361
+ {
362
+ "epoch": 0.06802494247890893,
363
+ "grad_norm": 5.5563562303649885,
364
+ "learning_rate": 2.9746446102701606e-05,
365
+ "loss": 0.6986,
366
+ "step": 510
367
+ },
368
+ {
369
+ "epoch": 0.06935876488045617,
370
+ "grad_norm": 4.036767303783137,
371
+ "learning_rate": 2.9734689337657157e-05,
372
+ "loss": 0.7119,
373
+ "step": 520
374
+ },
375
+ {
376
+ "epoch": 0.07069258728200341,
377
+ "grad_norm": 1.9856990692088847,
378
+ "learning_rate": 2.9722668576105703e-05,
379
+ "loss": 0.7205,
380
+ "step": 530
381
+ },
382
+ {
383
+ "epoch": 0.07202640968355063,
384
+ "grad_norm": 5.200308739226583,
385
+ "learning_rate": 2.971038403341921e-05,
386
+ "loss": 0.6918,
387
+ "step": 540
388
+ },
389
+ {
390
+ "epoch": 0.07336023208509787,
391
+ "grad_norm": 2.237349124701919,
392
+ "learning_rate": 2.9697835929695727e-05,
393
+ "loss": 0.7339,
394
+ "step": 550
395
+ },
396
+ {
397
+ "epoch": 0.07469405448664511,
398
+ "grad_norm": 1.6388680632753365,
399
+ "learning_rate": 2.968502448975544e-05,
400
+ "loss": 0.7086,
401
+ "step": 560
402
+ },
403
+ {
404
+ "epoch": 0.07602787688819233,
405
+ "grad_norm": 2.8545575025135244,
406
+ "learning_rate": 2.967194994313663e-05,
407
+ "loss": 0.678,
408
+ "step": 570
409
+ },
410
+ {
411
+ "epoch": 0.07736169928973957,
412
+ "grad_norm": 2.674647983669599,
413
+ "learning_rate": 2.9658612524091594e-05,
414
+ "loss": 0.7119,
415
+ "step": 580
416
+ },
417
+ {
418
+ "epoch": 0.07869552169128681,
419
+ "grad_norm": 2.489047760330112,
420
+ "learning_rate": 2.9645012471582406e-05,
421
+ "loss": 0.7382,
422
+ "step": 590
423
+ },
424
+ {
425
+ "epoch": 0.08002934409283403,
426
+ "grad_norm": 5.509352102248308,
427
+ "learning_rate": 2.9631150029276662e-05,
428
+ "loss": 0.738,
429
+ "step": 600
430
+ },
431
+ {
432
+ "epoch": 0.08136316649438127,
433
+ "grad_norm": 3.6489235270404015,
434
+ "learning_rate": 2.9617025445543114e-05,
435
+ "loss": 0.7018,
436
+ "step": 610
437
+ },
438
+ {
439
+ "epoch": 0.08269698889592851,
440
+ "grad_norm": 2.7813651243235697,
441
+ "learning_rate": 2.9602638973447218e-05,
442
+ "loss": 0.7381,
443
+ "step": 620
444
+ },
445
+ {
446
+ "epoch": 0.08403081129747574,
447
+ "grad_norm": 8.271390523006518,
448
+ "learning_rate": 2.9587990870746574e-05,
449
+ "loss": 0.7168,
450
+ "step": 630
451
+ },
452
+ {
453
+ "epoch": 0.08536463369902297,
454
+ "grad_norm": 1.2460611751687307,
455
+ "learning_rate": 2.9573081399886356e-05,
456
+ "loss": 0.7004,
457
+ "step": 640
458
+ },
459
+ {
460
+ "epoch": 0.08669845610057021,
461
+ "grad_norm": 1.704626418994062,
462
+ "learning_rate": 2.9557910827994568e-05,
463
+ "loss": 0.738,
464
+ "step": 650
465
+ },
466
+ {
467
+ "epoch": 0.08803227850211744,
468
+ "grad_norm": 3.275051693107957,
469
+ "learning_rate": 2.9542479426877283e-05,
470
+ "loss": 0.7017,
471
+ "step": 660
472
+ },
473
+ {
474
+ "epoch": 0.08936610090366467,
475
+ "grad_norm": 11.389990685570503,
476
+ "learning_rate": 2.9526787473013753e-05,
477
+ "loss": 0.7107,
478
+ "step": 670
479
+ },
480
+ {
481
+ "epoch": 0.09069992330521191,
482
+ "grad_norm": 5.591277359184055,
483
+ "learning_rate": 2.9510835247551485e-05,
484
+ "loss": 0.7141,
485
+ "step": 680
486
+ },
487
+ {
488
+ "epoch": 0.09203374570675915,
489
+ "grad_norm": 3.180111568581053,
490
+ "learning_rate": 2.949462303630116e-05,
491
+ "loss": 0.6987,
492
+ "step": 690
493
+ },
494
+ {
495
+ "epoch": 0.09336756810830638,
496
+ "grad_norm": 3.8428068166831753,
497
+ "learning_rate": 2.9478151129731567e-05,
498
+ "loss": 0.7373,
499
+ "step": 700
500
+ },
501
+ {
502
+ "epoch": 0.09470139050985361,
503
+ "grad_norm": 2.231397231771392,
504
+ "learning_rate": 2.9461419822964348e-05,
505
+ "loss": 0.6962,
506
+ "step": 710
507
+ },
508
+ {
509
+ "epoch": 0.09603521291140085,
510
+ "grad_norm": 18.287201889017563,
511
+ "learning_rate": 2.9444429415768726e-05,
512
+ "loss": 0.6723,
513
+ "step": 720
514
+ },
515
+ {
516
+ "epoch": 0.09736903531294808,
517
+ "grad_norm": 4.340932687135137,
518
+ "learning_rate": 2.942718021255617e-05,
519
+ "loss": 0.7151,
520
+ "step": 730
521
+ },
522
+ {
523
+ "epoch": 0.09870285771449532,
524
+ "grad_norm": 2.7813821825484446,
525
+ "learning_rate": 2.940967252237488e-05,
526
+ "loss": 0.7332,
527
+ "step": 740
528
+ },
529
+ {
530
+ "epoch": 0.10003668011604255,
531
+ "grad_norm": 2.3251782912937475,
532
+ "learning_rate": 2.9391906658904296e-05,
533
+ "loss": 0.6751,
534
+ "step": 750
535
+ },
536
+ {
537
+ "epoch": 0.10137050251758978,
538
+ "grad_norm": 8.123799866292751,
539
+ "learning_rate": 2.937388294044946e-05,
540
+ "loss": 0.6886,
541
+ "step": 760
542
+ },
543
+ {
544
+ "epoch": 0.10270432491913702,
545
+ "grad_norm": 1.528579329214318,
546
+ "learning_rate": 2.9355601689935315e-05,
547
+ "loss": 0.7146,
548
+ "step": 770
549
+ },
550
+ {
551
+ "epoch": 0.10403814732068425,
552
+ "grad_norm": 2.0278953433974825,
553
+ "learning_rate": 2.933706323490092e-05,
554
+ "loss": 0.7453,
555
+ "step": 780
556
+ },
557
+ {
558
+ "epoch": 0.10537196972223148,
559
+ "grad_norm": 1.4306270659678864,
560
+ "learning_rate": 2.9318267907493583e-05,
561
+ "loss": 0.6702,
562
+ "step": 790
563
+ },
564
+ {
565
+ "epoch": 0.10670579212377872,
566
+ "grad_norm": 1.5178081087799355,
567
+ "learning_rate": 2.9299216044462903e-05,
568
+ "loss": 0.7346,
569
+ "step": 800
570
+ },
571
+ {
572
+ "epoch": 0.10803961452532596,
573
+ "grad_norm": 9.506616797760028,
574
+ "learning_rate": 2.927990798715475e-05,
575
+ "loss": 0.6558,
576
+ "step": 810
577
+ },
578
+ {
579
+ "epoch": 0.1093734369268732,
580
+ "grad_norm": 2.4597311302505767,
581
+ "learning_rate": 2.926034408150513e-05,
582
+ "loss": 0.726,
583
+ "step": 820
584
+ },
585
+ {
586
+ "epoch": 0.11070725932842042,
587
+ "grad_norm": 12.372180964422007,
588
+ "learning_rate": 2.9240524678034016e-05,
589
+ "loss": 0.7308,
590
+ "step": 830
591
+ },
592
+ {
593
+ "epoch": 0.11204108172996766,
594
+ "grad_norm": 1.4488469801164658,
595
+ "learning_rate": 2.9220450131839037e-05,
596
+ "loss": 0.7072,
597
+ "step": 840
598
+ },
599
+ {
600
+ "epoch": 0.1133749041315149,
601
+ "grad_norm": 8.602946960846197,
602
+ "learning_rate": 2.920012080258912e-05,
603
+ "loss": 0.7234,
604
+ "step": 850
605
+ },
606
+ {
607
+ "epoch": 0.11470872653306212,
608
+ "grad_norm": 1.441195423452674,
609
+ "learning_rate": 2.9179537054518085e-05,
610
+ "loss": 0.6934,
611
+ "step": 860
612
+ },
613
+ {
614
+ "epoch": 0.11604254893460936,
615
+ "grad_norm": 4.318952956999577,
616
+ "learning_rate": 2.9158699256418056e-05,
617
+ "loss": 0.6534,
618
+ "step": 870
619
+ },
620
+ {
621
+ "epoch": 0.1173763713361566,
622
+ "grad_norm": 9.733179695623866,
623
+ "learning_rate": 2.9137607781632913e-05,
624
+ "loss": 0.71,
625
+ "step": 880
626
+ },
627
+ {
628
+ "epoch": 0.11871019373770382,
629
+ "grad_norm": 7.397049093836735,
630
+ "learning_rate": 2.911626300805155e-05,
631
+ "loss": 0.7386,
632
+ "step": 890
633
+ },
634
+ {
635
+ "epoch": 0.12004401613925106,
636
+ "grad_norm": 2.920812240139869,
637
+ "learning_rate": 2.9094665318101155e-05,
638
+ "loss": 0.6789,
639
+ "step": 900
640
+ },
641
+ {
642
+ "epoch": 0.1213778385407983,
643
+ "grad_norm": 1.7031296196271206,
644
+ "learning_rate": 2.9072815098740326e-05,
645
+ "loss": 0.715,
646
+ "step": 910
647
+ },
648
+ {
649
+ "epoch": 0.12271166094234552,
650
+ "grad_norm": 1.5630656172291801,
651
+ "learning_rate": 2.9050712741452136e-05,
652
+ "loss": 0.7136,
653
+ "step": 920
654
+ },
655
+ {
656
+ "epoch": 0.12404548334389276,
657
+ "grad_norm": 7.870543414771234,
658
+ "learning_rate": 2.902835864223715e-05,
659
+ "loss": 0.6669,
660
+ "step": 930
661
+ },
662
+ {
663
+ "epoch": 0.12537930574543998,
664
+ "grad_norm": 4.843671834991794,
665
+ "learning_rate": 2.9005753201606287e-05,
666
+ "loss": 0.7281,
667
+ "step": 940
668
+ },
669
+ {
670
+ "epoch": 0.12671312814698724,
671
+ "grad_norm": 3.010503818258016,
672
+ "learning_rate": 2.8982896824573678e-05,
673
+ "loss": 0.7018,
674
+ "step": 950
675
+ },
676
+ {
677
+ "epoch": 0.12804695054853446,
678
+ "grad_norm": 2.5552186559589654,
679
+ "learning_rate": 2.8959789920649394e-05,
680
+ "loss": 0.7338,
681
+ "step": 960
682
+ },
683
+ {
684
+ "epoch": 0.12938077295008168,
685
+ "grad_norm": 12.306055851495117,
686
+ "learning_rate": 2.893643290383212e-05,
687
+ "loss": 0.6732,
688
+ "step": 970
689
+ },
690
+ {
691
+ "epoch": 0.13071459535162894,
692
+ "grad_norm": 2.16185926525944,
693
+ "learning_rate": 2.891282619260172e-05,
694
+ "loss": 0.7108,
695
+ "step": 980
696
+ },
697
+ {
698
+ "epoch": 0.13204841775317616,
699
+ "grad_norm": 5.992378798792086,
700
+ "learning_rate": 2.8888970209911754e-05,
701
+ "loss": 0.6525,
702
+ "step": 990
703
+ },
704
+ {
705
+ "epoch": 0.13338224015472339,
706
+ "grad_norm": 2.986272238787896,
707
+ "learning_rate": 2.8864865383181893e-05,
708
+ "loss": 0.6655,
709
+ "step": 1000
710
+ },
711
+ {
712
+ "epoch": 0.13471606255627064,
713
+ "grad_norm": 12.855377354582437,
714
+ "learning_rate": 2.8840512144290273e-05,
715
+ "loss": 0.6826,
716
+ "step": 1010
717
+ },
718
+ {
719
+ "epoch": 0.13604988495781786,
720
+ "grad_norm": 2.045979893776702,
721
+ "learning_rate": 2.8815910929565734e-05,
722
+ "loss": 0.6616,
723
+ "step": 1020
724
+ },
725
+ {
726
+ "epoch": 0.1373837073593651,
727
+ "grad_norm": 6.623264301300591,
728
+ "learning_rate": 2.879106217978002e-05,
729
+ "loss": 0.6935,
730
+ "step": 1030
731
+ },
732
+ {
733
+ "epoch": 0.13871752976091234,
734
+ "grad_norm": 2.67990218211766,
735
+ "learning_rate": 2.8765966340139892e-05,
736
+ "loss": 0.6671,
737
+ "step": 1040
738
+ },
739
+ {
740
+ "epoch": 0.14005135216245956,
741
+ "grad_norm": 2.699521523924172,
742
+ "learning_rate": 2.8740623860279116e-05,
743
+ "loss": 0.6763,
744
+ "step": 1050
745
+ },
746
+ {
747
+ "epoch": 0.14138517456400682,
748
+ "grad_norm": 4.1129898011507535,
749
+ "learning_rate": 2.871503519425044e-05,
750
+ "loss": 0.7159,
751
+ "step": 1060
752
+ },
753
+ {
754
+ "epoch": 0.14271899696555404,
755
+ "grad_norm": 2.4592021333659146,
756
+ "learning_rate": 2.8689200800517448e-05,
757
+ "loss": 0.6551,
758
+ "step": 1070
759
+ },
760
+ {
761
+ "epoch": 0.14405281936710126,
762
+ "grad_norm": 5.138500389099849,
763
+ "learning_rate": 2.866312114194634e-05,
764
+ "loss": 0.7214,
765
+ "step": 1080
766
+ },
767
+ {
768
+ "epoch": 0.14538664176864852,
769
+ "grad_norm": 2.822433730666048,
770
+ "learning_rate": 2.8636796685797657e-05,
771
+ "loss": 0.6862,
772
+ "step": 1090
773
+ },
774
+ {
775
+ "epoch": 0.14672046417019574,
776
+ "grad_norm": 3.086468537427806,
777
+ "learning_rate": 2.8610227903717876e-05,
778
+ "loss": 0.6784,
779
+ "step": 1100
780
+ },
781
+ {
782
+ "epoch": 0.14805428657174297,
783
+ "grad_norm": 2.079766793749202,
784
+ "learning_rate": 2.8583415271730994e-05,
785
+ "loss": 0.7065,
786
+ "step": 1110
787
+ },
788
+ {
789
+ "epoch": 0.14938810897329022,
790
+ "grad_norm": 1.659870509072264,
791
+ "learning_rate": 2.855635927022998e-05,
792
+ "loss": 0.7197,
793
+ "step": 1120
794
+ },
795
+ {
796
+ "epoch": 0.15072193137483744,
797
+ "grad_norm": 7.870626779339635,
798
+ "learning_rate": 2.8529060383968175e-05,
799
+ "loss": 0.7305,
800
+ "step": 1130
801
+ },
802
+ {
803
+ "epoch": 0.15205575377638467,
804
+ "grad_norm": 3.0600340899893537,
805
+ "learning_rate": 2.850151910205061e-05,
806
+ "loss": 0.6922,
807
+ "step": 1140
808
+ },
809
+ {
810
+ "epoch": 0.15338957617793192,
811
+ "grad_norm": 3.6147451373702806,
812
+ "learning_rate": 2.847373591792523e-05,
813
+ "loss": 0.7044,
814
+ "step": 1150
815
+ },
816
+ {
817
+ "epoch": 0.15472339857947914,
818
+ "grad_norm": 4.740777951553679,
819
+ "learning_rate": 2.844571132937407e-05,
820
+ "loss": 0.6794,
821
+ "step": 1160
822
+ },
823
+ {
824
+ "epoch": 0.15605722098102637,
825
+ "grad_norm": 3.377522973717319,
826
+ "learning_rate": 2.841744583850431e-05,
827
+ "loss": 0.673,
828
+ "step": 1170
829
+ },
830
+ {
831
+ "epoch": 0.15739104338257362,
832
+ "grad_norm": 4.250656077289992,
833
+ "learning_rate": 2.838893995173932e-05,
834
+ "loss": 0.6975,
835
+ "step": 1180
836
+ },
837
+ {
838
+ "epoch": 0.15872486578412084,
839
+ "grad_norm": 11.73693900915769,
840
+ "learning_rate": 2.836019417980955e-05,
841
+ "loss": 0.6572,
842
+ "step": 1190
843
+ },
844
+ {
845
+ "epoch": 0.16005868818566807,
846
+ "grad_norm": 2.729291714043308,
847
+ "learning_rate": 2.8331209037743387e-05,
848
+ "loss": 0.7247,
849
+ "step": 1200
850
+ },
851
+ {
852
+ "epoch": 0.16139251058721532,
853
+ "grad_norm": 2.347985877636318,
854
+ "learning_rate": 2.8301985044857947e-05,
855
+ "loss": 0.7199,
856
+ "step": 1210
857
+ },
858
+ {
859
+ "epoch": 0.16272633298876255,
860
+ "grad_norm": 2.2534314586033113,
861
+ "learning_rate": 2.8272522724749743e-05,
862
+ "loss": 0.6835,
863
+ "step": 1220
864
+ },
865
+ {
866
+ "epoch": 0.16406015539030977,
867
+ "grad_norm": 3.159583116387406,
868
+ "learning_rate": 2.8242822605285323e-05,
869
+ "loss": 0.7122,
870
+ "step": 1230
871
+ },
872
+ {
873
+ "epoch": 0.16539397779185702,
874
+ "grad_norm": 2.086588782887239,
875
+ "learning_rate": 2.8212885218591812e-05,
876
+ "loss": 0.6949,
877
+ "step": 1240
878
+ },
879
+ {
880
+ "epoch": 0.16672780019340425,
881
+ "grad_norm": 7.284236966547317,
882
+ "learning_rate": 2.8182711101047362e-05,
883
+ "loss": 0.6641,
884
+ "step": 1250
885
+ },
886
+ {
887
+ "epoch": 0.16806162259495147,
888
+ "grad_norm": 3.0369619450249594,
889
+ "learning_rate": 2.815230079327156e-05,
890
+ "loss": 0.6731,
891
+ "step": 1260
892
+ },
893
+ {
894
+ "epoch": 0.16939544499649872,
895
+ "grad_norm": 1.4144726574636068,
896
+ "learning_rate": 2.8121654840115734e-05,
897
+ "loss": 0.6898,
898
+ "step": 1270
899
+ },
900
+ {
901
+ "epoch": 0.17072926739804595,
902
+ "grad_norm": 3.66202356670303,
903
+ "learning_rate": 2.809077379065319e-05,
904
+ "loss": 0.7174,
905
+ "step": 1280
906
+ },
907
+ {
908
+ "epoch": 0.17206308979959317,
909
+ "grad_norm": 4.778073521019285,
910
+ "learning_rate": 2.805965819816937e-05,
911
+ "loss": 0.6186,
912
+ "step": 1290
913
+ },
914
+ {
915
+ "epoch": 0.17339691220114042,
916
+ "grad_norm": 3.9620427201734576,
917
+ "learning_rate": 2.802830862015196e-05,
918
+ "loss": 0.684,
919
+ "step": 1300
920
+ },
921
+ {
922
+ "epoch": 0.17473073460268765,
923
+ "grad_norm": 4.170199740083487,
924
+ "learning_rate": 2.799672561828087e-05,
925
+ "loss": 0.7102,
926
+ "step": 1310
927
+ },
928
+ {
929
+ "epoch": 0.17606455700423487,
930
+ "grad_norm": 2.2612205048804714,
931
+ "learning_rate": 2.79649097584182e-05,
932
+ "loss": 0.7451,
933
+ "step": 1320
934
+ },
935
+ {
936
+ "epoch": 0.17739837940578213,
937
+ "grad_norm": 1.7156828128822517,
938
+ "learning_rate": 2.7932861610598077e-05,
939
+ "loss": 0.6641,
940
+ "step": 1330
941
+ },
942
+ {
943
+ "epoch": 0.17873220180732935,
944
+ "grad_norm": 7.960733847217257,
945
+ "learning_rate": 2.7900581749016466e-05,
946
+ "loss": 0.7365,
947
+ "step": 1340
948
+ },
949
+ {
950
+ "epoch": 0.1800660242088766,
951
+ "grad_norm": 2.5364939682563756,
952
+ "learning_rate": 2.7868070752020865e-05,
953
+ "loss": 0.7078,
954
+ "step": 1350
955
+ },
956
+ {
957
+ "epoch": 0.18139984661042383,
958
+ "grad_norm": 2.7446281678776137,
959
+ "learning_rate": 2.7835329202099944e-05,
960
+ "loss": 0.7214,
961
+ "step": 1360
962
+ },
963
+ {
964
+ "epoch": 0.18273366901197105,
965
+ "grad_norm": 3.2416602016145886,
966
+ "learning_rate": 2.7802357685873117e-05,
967
+ "loss": 0.6757,
968
+ "step": 1370
969
+ },
970
+ {
971
+ "epoch": 0.1840674914135183,
972
+ "grad_norm": 5.225459736579946,
973
+ "learning_rate": 2.7769156794080033e-05,
974
+ "loss": 0.7381,
975
+ "step": 1380
976
+ },
977
+ {
978
+ "epoch": 0.18540131381506553,
979
+ "grad_norm": 5.176692689501482,
980
+ "learning_rate": 2.7735727121569967e-05,
981
+ "loss": 0.7354,
982
+ "step": 1390
983
+ },
984
+ {
985
+ "epoch": 0.18673513621661275,
986
+ "grad_norm": 2.7441883232342574,
987
+ "learning_rate": 2.770206926729121e-05,
988
+ "loss": 0.6937,
989
+ "step": 1400
990
+ },
991
+ {
992
+ "epoch": 0.18806895861816,
993
+ "grad_norm": 2.9792116246243525,
994
+ "learning_rate": 2.7668183834280284e-05,
995
+ "loss": 0.6641,
996
+ "step": 1410
997
+ },
998
+ {
999
+ "epoch": 0.18940278101970723,
1000
+ "grad_norm": 2.4645298487410723,
1001
+ "learning_rate": 2.763407142965117e-05,
1002
+ "loss": 0.6274,
1003
+ "step": 1420
1004
+ },
1005
+ {
1006
+ "epoch": 0.19073660342125445,
1007
+ "grad_norm": 7.245032878035033,
1008
+ "learning_rate": 2.759973266458444e-05,
1009
+ "loss": 0.6962,
1010
+ "step": 1430
1011
+ },
1012
+ {
1013
+ "epoch": 0.1920704258228017,
1014
+ "grad_norm": 5.642209662597534,
1015
+ "learning_rate": 2.756516815431627e-05,
1016
+ "loss": 0.7016,
1017
+ "step": 1440
1018
+ },
1019
+ {
1020
+ "epoch": 0.19340424822434893,
1021
+ "grad_norm": 2.9804981875184526,
1022
+ "learning_rate": 2.7530378518127445e-05,
1023
+ "loss": 0.7331,
1024
+ "step": 1450
1025
+ },
1026
+ {
1027
+ "epoch": 0.19473807062589615,
1028
+ "grad_norm": 7.496561660992361,
1029
+ "learning_rate": 2.7495364379332256e-05,
1030
+ "loss": 0.7234,
1031
+ "step": 1460
1032
+ },
1033
+ {
1034
+ "epoch": 0.1960718930274434,
1035
+ "grad_norm": 1.6139389803246291,
1036
+ "learning_rate": 2.7460126365267335e-05,
1037
+ "loss": 0.7013,
1038
+ "step": 1470
1039
+ },
1040
+ {
1041
+ "epoch": 0.19740571542899063,
1042
+ "grad_norm": 4.618678334755141,
1043
+ "learning_rate": 2.7424665107280402e-05,
1044
+ "loss": 0.6892,
1045
+ "step": 1480
1046
+ },
1047
+ {
1048
+ "epoch": 0.19873953783053785,
1049
+ "grad_norm": 15.494190234738744,
1050
+ "learning_rate": 2.738898124071898e-05,
1051
+ "loss": 0.6785,
1052
+ "step": 1490
1053
+ },
1054
+ {
1055
+ "epoch": 0.2000733602320851,
1056
+ "grad_norm": 3.1680363319798954,
1057
+ "learning_rate": 2.735307540491898e-05,
1058
+ "loss": 0.669,
1059
+ "step": 1500
1060
+ },
1061
+ {
1062
+ "epoch": 0.20140718263363233,
1063
+ "grad_norm": 2.5397562341036224,
1064
+ "learning_rate": 2.7316948243193273e-05,
1065
+ "loss": 0.6726,
1066
+ "step": 1510
1067
+ },
1068
+ {
1069
+ "epoch": 0.20274100503517956,
1070
+ "grad_norm": 4.139021422606072,
1071
+ "learning_rate": 2.7280600402820146e-05,
1072
+ "loss": 0.6706,
1073
+ "step": 1520
1074
+ },
1075
+ {
1076
+ "epoch": 0.2040748274367268,
1077
+ "grad_norm": 2.7422468825646065,
1078
+ "learning_rate": 2.724403253503171e-05,
1079
+ "loss": 0.7078,
1080
+ "step": 1530
1081
+ },
1082
+ {
1083
+ "epoch": 0.20540864983827403,
1084
+ "grad_norm": 2.744225768808104,
1085
+ "learning_rate": 2.7207245295002242e-05,
1086
+ "loss": 0.6821,
1087
+ "step": 1540
1088
+ },
1089
+ {
1090
+ "epoch": 0.20674247223982126,
1091
+ "grad_norm": 2.234040668790152,
1092
+ "learning_rate": 2.7170239341836436e-05,
1093
+ "loss": 0.7451,
1094
+ "step": 1550
1095
+ },
1096
+ {
1097
+ "epoch": 0.2080762946413685,
1098
+ "grad_norm": 2.531733996425376,
1099
+ "learning_rate": 2.7133015338557585e-05,
1100
+ "loss": 0.7205,
1101
+ "step": 1560
1102
+ },
1103
+ {
1104
+ "epoch": 0.20941011704291573,
1105
+ "grad_norm": 2.9772483856455616,
1106
+ "learning_rate": 2.7095573952095727e-05,
1107
+ "loss": 0.7274,
1108
+ "step": 1570
1109
+ },
1110
+ {
1111
+ "epoch": 0.21074393944446296,
1112
+ "grad_norm": 3.317235333047955,
1113
+ "learning_rate": 2.705791585327568e-05,
1114
+ "loss": 0.7309,
1115
+ "step": 1580
1116
+ },
1117
+ {
1118
+ "epoch": 0.2120777618460102,
1119
+ "grad_norm": 1.9652386793628944,
1120
+ "learning_rate": 2.7020041716805014e-05,
1121
+ "loss": 0.7157,
1122
+ "step": 1590
1123
+ },
1124
+ {
1125
+ "epoch": 0.21341158424755743,
1126
+ "grad_norm": 2.93724058913164,
1127
+ "learning_rate": 2.6981952221261986e-05,
1128
+ "loss": 0.7123,
1129
+ "step": 1600
1130
+ }
1131
+ ],
1132
+ "logging_steps": 10,
1133
+ "max_steps": 7497,
1134
+ "num_input_tokens_seen": 0,
1135
+ "num_train_epochs": 1,
1136
+ "save_steps": 400,
1137
+ "stateful_callbacks": {
1138
+ "TrainerControl": {
1139
+ "args": {
1140
+ "should_epoch_stop": false,
1141
+ "should_evaluate": false,
1142
+ "should_log": false,
1143
+ "should_save": true,
1144
+ "should_training_stop": false
1145
+ },
1146
+ "attributes": {}
1147
+ }
1148
+ },
1149
+ "total_flos": 4.3737129443917824e+18,
1150
+ "train_batch_size": 4,
1151
+ "trial_name": null,
1152
+ "trial_params": null
1153
+ }
checkpoint-1600/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81e448587a2aaa72701b15796f660da2d8aedd670c2a2eaae01d5488a2bf1cfe
3
+ size 6520
checkpoint-1600/zero_to_fp32.py ADDED
@@ -0,0 +1,587 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
215
+ elif zero_stage == 3:
216
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
217
+
218
+
219
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
220
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
221
+ return
222
+
223
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
224
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
225
+
226
+ if debug:
227
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
228
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
229
+
230
+ wanted_params = len(frozen_param_shapes)
231
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
232
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
233
+ print(f'Frozen params: Have {avail_numel} numels to process.')
234
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
235
+
236
+ total_params = 0
237
+ total_numel = 0
238
+ for name, shape in frozen_param_shapes.items():
239
+ total_params += 1
240
+ unpartitioned_numel = shape.numel()
241
+ total_numel += unpartitioned_numel
242
+
243
+ state_dict[name] = frozen_param_fragments[name]
244
+
245
+ if debug:
246
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
247
+
248
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
249
+
250
+
251
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
252
+ param_shapes = zero_model_states[0].param_shapes
253
+
254
+ # Reconstruction protocol:
255
+ #
256
+ # XXX: document this
257
+
258
+ if debug:
259
+ for i in range(world_size):
260
+ for j in range(len(fp32_flat_groups[0])):
261
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
262
+
263
+ # XXX: memory usage doubles here (zero2)
264
+ num_param_groups = len(fp32_flat_groups[0])
265
+ merged_single_partition_of_fp32_groups = []
266
+ for i in range(num_param_groups):
267
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
268
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
269
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
270
+ avail_numel = sum(
271
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
272
+
273
+ if debug:
274
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
275
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
276
+ # not asserting if there is a mismatch due to possible padding
277
+ print(f"Have {avail_numel} numels to process.")
278
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
279
+
280
+ # params
281
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
282
+ # out-of-core computing solution
283
+ total_numel = 0
284
+ total_params = 0
285
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
286
+ offset = 0
287
+ avail_numel = full_single_fp32_vector.numel()
288
+ for name, shape in shapes.items():
289
+
290
+ unpartitioned_numel = shape.numel()
291
+ total_numel += unpartitioned_numel
292
+ total_params += 1
293
+
294
+ if debug:
295
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
296
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
297
+ offset += unpartitioned_numel
298
+
299
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
300
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
301
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
302
+ # live optimizer object, so we are checking that the numbers are within the right range
303
+ align_to = 2 * world_size
304
+
305
+ def zero2_align(x):
306
+ return align_to * math.ceil(x / align_to)
307
+
308
+ if debug:
309
+ print(f"original offset={offset}, avail_numel={avail_numel}")
310
+
311
+ offset = zero2_align(offset)
312
+ avail_numel = zero2_align(avail_numel)
313
+
314
+ if debug:
315
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
316
+
317
+ # Sanity check
318
+ if offset != avail_numel:
319
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
320
+
321
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
322
+
323
+
324
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
325
+ state_dict = OrderedDict()
326
+
327
+ # buffers
328
+ buffers = zero_model_states[0].buffers
329
+ state_dict.update(buffers)
330
+ if debug:
331
+ print(f"added {len(buffers)} buffers")
332
+
333
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
334
+
335
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
336
+
337
+ # recover shared parameters
338
+ for pair in zero_model_states[0].shared_params:
339
+ if pair[1] in state_dict:
340
+ state_dict[pair[0]] = state_dict[pair[1]]
341
+
342
+ return state_dict
343
+
344
+
345
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
346
+ remainder = unpartitioned_numel % world_size
347
+ padding_numel = (world_size - remainder) if remainder else 0
348
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
349
+ return partitioned_numel, padding_numel
350
+
351
+
352
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
353
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
354
+ return
355
+
356
+ if debug:
357
+ for i in range(world_size):
358
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
359
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
360
+
361
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
362
+ wanted_params = len(frozen_param_shapes)
363
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
364
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
365
+ print(f'Frozen params: Have {avail_numel} numels to process.')
366
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
367
+
368
+ total_params = 0
369
+ total_numel = 0
370
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
371
+ total_params += 1
372
+ unpartitioned_numel = shape.numel()
373
+ total_numel += unpartitioned_numel
374
+
375
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
376
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
377
+
378
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
379
+
380
+ if debug:
381
+ print(
382
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
383
+ )
384
+
385
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
386
+
387
+
388
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
389
+ param_shapes = zero_model_states[0].param_shapes
390
+ avail_numel = fp32_flat_groups[0].numel() * world_size
391
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
392
+ # param, re-consolidating each param, while dealing with padding if any
393
+
394
+ # merge list of dicts, preserving order
395
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
396
+
397
+ if debug:
398
+ for i in range(world_size):
399
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
400
+
401
+ wanted_params = len(param_shapes)
402
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
403
+ # not asserting if there is a mismatch due to possible padding
404
+ avail_numel = fp32_flat_groups[0].numel() * world_size
405
+ print(f"Trainable params: Have {avail_numel} numels to process.")
406
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
407
+
408
+ # params
409
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
410
+ # out-of-core computing solution
411
+ offset = 0
412
+ total_numel = 0
413
+ total_params = 0
414
+ for name, shape in param_shapes.items():
415
+
416
+ unpartitioned_numel = shape.numel()
417
+ total_numel += unpartitioned_numel
418
+ total_params += 1
419
+
420
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
421
+
422
+ if debug:
423
+ print(
424
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
425
+ )
426
+
427
+ # XXX: memory usage doubles here
428
+ state_dict[name] = torch.cat(
429
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
430
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
431
+ offset += partitioned_numel
432
+
433
+ offset *= world_size
434
+
435
+ # Sanity check
436
+ if offset != avail_numel:
437
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
438
+
439
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
440
+
441
+
442
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
443
+ state_dict = OrderedDict()
444
+
445
+ # buffers
446
+ buffers = zero_model_states[0].buffers
447
+ state_dict.update(buffers)
448
+ if debug:
449
+ print(f"added {len(buffers)} buffers")
450
+
451
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
452
+
453
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
454
+
455
+ # recover shared parameters
456
+ for pair in zero_model_states[0].shared_params:
457
+ if pair[1] in state_dict:
458
+ state_dict[pair[0]] = state_dict[pair[1]]
459
+
460
+ return state_dict
461
+
462
+
463
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
464
+ """
465
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
466
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
467
+ via a model hub.
468
+
469
+ Args:
470
+ - ``checkpoint_dir``: path to the desired checkpoint folder
471
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
472
+
473
+ Returns:
474
+ - pytorch ``state_dict``
475
+
476
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
477
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
478
+ the checkpoint.
479
+
480
+ A typical usage might be ::
481
+
482
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
483
+ # do the training and checkpoint saving
484
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
485
+ model = model.cpu() # move to cpu
486
+ model.load_state_dict(state_dict)
487
+ # submit to model hub or save the model to share with others
488
+
489
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
490
+ application. i.e. you will need to re-initialize the deepspeed engine, since
491
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
492
+
493
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
494
+
495
+ """
496
+ if tag is None:
497
+ latest_path = os.path.join(checkpoint_dir, 'latest')
498
+ if os.path.isfile(latest_path):
499
+ with open(latest_path, 'r') as fd:
500
+ tag = fd.read().strip()
501
+ else:
502
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
503
+
504
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
505
+
506
+ if not os.path.isdir(ds_checkpoint_dir):
507
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
508
+
509
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
510
+
511
+
512
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
513
+ """
514
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
515
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
516
+
517
+ Args:
518
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
519
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
520
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
521
+ """
522
+
523
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
524
+ print(f"Saving fp32 state dict to {output_file}")
525
+ torch.save(state_dict, output_file)
526
+
527
+
528
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
529
+ """
530
+ 1. Put the provided model to cpu
531
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
532
+ 3. Load it into the provided model
533
+
534
+ Args:
535
+ - ``model``: the model object to update
536
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
537
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
538
+
539
+ Returns:
540
+ - ``model`: modified model
541
+
542
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
543
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
544
+ conveniently placed for you in the checkpoint folder.
545
+
546
+ A typical usage might be ::
547
+
548
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
549
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
550
+ # submit to model hub or save the model to share with others
551
+
552
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
553
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
554
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
555
+
556
+ """
557
+ logger.info(f"Extracting fp32 weights")
558
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
559
+
560
+ logger.info(f"Overwriting model with fp32 weights")
561
+ model = model.cpu()
562
+ model.load_state_dict(state_dict, strict=False)
563
+
564
+ return model
565
+
566
+
567
+ if __name__ == "__main__":
568
+
569
+ parser = argparse.ArgumentParser()
570
+ parser.add_argument("checkpoint_dir",
571
+ type=str,
572
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
573
+ parser.add_argument(
574
+ "output_file",
575
+ type=str,
576
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
577
+ parser.add_argument("-t",
578
+ "--tag",
579
+ type=str,
580
+ default=None,
581
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
582
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
583
+ args = parser.parse_args()
584
+
585
+ debug = args.debug
586
+
587
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)
checkpoint-2000/README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ base_model: Qwen/Qwen-VL-Chat
4
+ ---
5
+
6
+ # Model Card for Model ID
7
+
8
+ <!-- Provide a quick summary of what the model is/does. -->
9
+
10
+
11
+
12
+ ## Model Details
13
+
14
+ ### Model Description
15
+
16
+ <!-- Provide a longer summary of what this model is. -->
17
+
18
+
19
+
20
+ - **Developed by:** [More Information Needed]
21
+ - **Funded by [optional]:** [More Information Needed]
22
+ - **Shared by [optional]:** [More Information Needed]
23
+ - **Model type:** [More Information Needed]
24
+ - **Language(s) (NLP):** [More Information Needed]
25
+ - **License:** [More Information Needed]
26
+ - **Finetuned from model [optional]:** [More Information Needed]
27
+
28
+ ### Model Sources [optional]
29
+
30
+ <!-- Provide the basic links for the model. -->
31
+
32
+ - **Repository:** [More Information Needed]
33
+ - **Paper [optional]:** [More Information Needed]
34
+ - **Demo [optional]:** [More Information Needed]
35
+
36
+ ## Uses
37
+
38
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
+
40
+ ### Direct Use
41
+
42
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
+
44
+ [More Information Needed]
45
+
46
+ ### Downstream Use [optional]
47
+
48
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Out-of-Scope Use
53
+
54
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
+
56
+ [More Information Needed]
57
+
58
+ ## Bias, Risks, and Limitations
59
+
60
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ### Recommendations
65
+
66
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
+
68
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
+
70
+ ## How to Get Started with the Model
71
+
72
+ Use the code below to get started with the model.
73
+
74
+ [More Information Needed]
75
+
76
+ ## Training Details
77
+
78
+ ### Training Data
79
+
80
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
+
82
+ [More Information Needed]
83
+
84
+ ### Training Procedure
85
+
86
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
+
88
+ #### Preprocessing [optional]
89
+
90
+ [More Information Needed]
91
+
92
+
93
+ #### Training Hyperparameters
94
+
95
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
+
97
+ #### Speeds, Sizes, Times [optional]
98
+
99
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
+
101
+ [More Information Needed]
102
+
103
+ ## Evaluation
104
+
105
+ <!-- This section describes the evaluation protocols and provides the results. -->
106
+
107
+ ### Testing Data, Factors & Metrics
108
+
109
+ #### Testing Data
110
+
111
+ <!-- This should link to a Dataset Card if possible. -->
112
+
113
+ [More Information Needed]
114
+
115
+ #### Factors
116
+
117
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Metrics
122
+
123
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
+
125
+ [More Information Needed]
126
+
127
+ ### Results
128
+
129
+ [More Information Needed]
130
+
131
+ #### Summary
132
+
133
+
134
+
135
+ ## Model Examination [optional]
136
+
137
+ <!-- Relevant interpretability work for the model goes here -->
138
+
139
+ [More Information Needed]
140
+
141
+ ## Environmental Impact
142
+
143
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
+
145
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
+
147
+ - **Hardware Type:** [More Information Needed]
148
+ - **Hours used:** [More Information Needed]
149
+ - **Cloud Provider:** [More Information Needed]
150
+ - **Compute Region:** [More Information Needed]
151
+ - **Carbon Emitted:** [More Information Needed]
152
+
153
+ ## Technical Specifications [optional]
154
+
155
+ ### Model Architecture and Objective
156
+
157
+ [More Information Needed]
158
+
159
+ ### Compute Infrastructure
160
+
161
+ [More Information Needed]
162
+
163
+ #### Hardware
164
+
165
+ [More Information Needed]
166
+
167
+ #### Software
168
+
169
+ [More Information Needed]
170
+
171
+ ## Citation [optional]
172
+
173
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
+
175
+ **BibTeX:**
176
+
177
+ [More Information Needed]
178
+
179
+ **APA:**
180
+
181
+ [More Information Needed]
182
+
183
+ ## Glossary [optional]
184
+
185
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
+
187
+ [More Information Needed]
188
+
189
+ ## More Information [optional]
190
+
191
+ [More Information Needed]
192
+
193
+ ## Model Card Authors [optional]
194
+
195
+ [More Information Needed]
196
+
197
+ ## Model Card Contact
198
+
199
+ [More Information Needed]
200
+ ### Framework versions
201
+
202
+ - PEFT 0.10.0
checkpoint-2000/adapter_config.json ADDED
@@ -0,0 +1,380 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "Qwen/Qwen-VL-Chat",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layer_replication": null,
10
+ "layers_pattern": null,
11
+ "layers_to_transform": null,
12
+ "loftq_config": {},
13
+ "lora_alpha": 16,
14
+ "lora_dropout": 0.05,
15
+ "megatron_config": null,
16
+ "megatron_core": "megatron.core",
17
+ "modules_to_save": null,
18
+ "peft_type": "LORA",
19
+ "r": 64,
20
+ "rank_pattern": {},
21
+ "revision": null,
22
+ "target_modules": [
23
+ "transformer.visual.transformer.resblocks.32.mlp.c_fc",
24
+ "transformer.visual.transformer.resblocks.8.attn.in_proj",
25
+ "transformer.visual.transformer.resblocks.16.attn.in_proj",
26
+ "transformer.visual.transformer.resblocks.41.mlp.c_proj",
27
+ "transformer.visual.transformer.resblocks.31.mlp.c_proj",
28
+ "transformer.visual.transformer.resblocks.14.mlp.c_proj",
29
+ "transformer.h.14.mlp.w2",
30
+ "transformer.visual.transformer.resblocks.39.mlp.c_fc",
31
+ "transformer.visual.transformer.resblocks.40.mlp.c_proj",
32
+ "transformer.h.0.attn.c_attn",
33
+ "transformer.visual.conv1",
34
+ "transformer.visual.transformer.resblocks.26.mlp.c_fc",
35
+ "transformer.h.7.mlp.w2",
36
+ "transformer.visual.transformer.resblocks.21.mlp.c_fc",
37
+ "transformer.visual.transformer.resblocks.17.attn.in_proj",
38
+ "transformer.h.29.attn.c_attn",
39
+ "transformer.h.3.attn.c_proj",
40
+ "transformer.visual.transformer.resblocks.0.attn.out_proj",
41
+ "transformer.visual.transformer.resblocks.28.mlp.c_fc",
42
+ "transformer.h.30.attn.c_proj",
43
+ "transformer.h.3.mlp.w2",
44
+ "transformer.h.22.mlp.w1",
45
+ "transformer.visual.transformer.resblocks.11.mlp.c_proj",
46
+ "transformer.h.11.mlp.c_proj",
47
+ "transformer.visual.transformer.resblocks.3.attn.in_proj",
48
+ "transformer.visual.transformer.resblocks.21.attn.out_proj",
49
+ "transformer.visual.transformer.resblocks.33.attn.out_proj",
50
+ "transformer.visual.transformer.resblocks.3.mlp.c_fc",
51
+ "transformer.visual.transformer.resblocks.15.attn.in_proj",
52
+ "transformer.visual.transformer.resblocks.4.attn.out_proj",
53
+ "transformer.visual.transformer.resblocks.34.attn.in_proj",
54
+ "transformer.h.17.mlp.c_proj",
55
+ "transformer.visual.transformer.resblocks.1.mlp.c_proj",
56
+ "transformer.visual.transformer.resblocks.4.mlp.c_proj",
57
+ "transformer.visual.transformer.resblocks.39.attn.out_proj",
58
+ "transformer.h.13.mlp.c_proj",
59
+ "transformer.visual.transformer.resblocks.3.mlp.c_proj",
60
+ "transformer.visual.transformer.resblocks.22.attn.out_proj",
61
+ "transformer.h.27.attn.c_attn",
62
+ "transformer.visual.transformer.resblocks.47.attn.in_proj",
63
+ "transformer.h.1.mlp.c_proj",
64
+ "transformer.h.21.attn.c_attn",
65
+ "transformer.visual.transformer.resblocks.18.attn.out_proj",
66
+ "transformer.visual.transformer.resblocks.5.attn.in_proj",
67
+ "transformer.h.6.attn.c_proj",
68
+ "transformer.visual.transformer.resblocks.17.mlp.c_fc",
69
+ "transformer.h.16.attn.c_attn",
70
+ "transformer.visual.transformer.resblocks.0.attn.in_proj",
71
+ "transformer.visual.transformer.resblocks.22.mlp.c_fc",
72
+ "transformer.visual.transformer.resblocks.27.mlp.c_proj",
73
+ "transformer.h.11.attn.c_attn",
74
+ "transformer.h.22.mlp.w2",
75
+ "transformer.h.8.mlp.w1",
76
+ "transformer.visual.transformer.resblocks.29.mlp.c_proj",
77
+ "transformer.visual.transformer.resblocks.44.attn.out_proj",
78
+ "transformer.h.13.mlp.w2",
79
+ "transformer.visual.transformer.resblocks.19.mlp.c_proj",
80
+ "transformer.visual.transformer.resblocks.27.attn.in_proj",
81
+ "transformer.h.29.mlp.w1",
82
+ "transformer.h.24.mlp.c_proj",
83
+ "transformer.visual.transformer.resblocks.25.mlp.c_proj",
84
+ "transformer.visual.transformer.resblocks.4.attn.in_proj",
85
+ "transformer.visual.transformer.resblocks.42.mlp.c_fc",
86
+ "transformer.h.28.mlp.w2",
87
+ "transformer.visual.transformer.resblocks.20.attn.in_proj",
88
+ "transformer.visual.transformer.resblocks.9.attn.in_proj",
89
+ "transformer.h.10.attn.c_proj",
90
+ "transformer.h.13.attn.c_proj",
91
+ "transformer.visual.transformer.resblocks.26.attn.in_proj",
92
+ "transformer.h.17.attn.c_attn",
93
+ "transformer.visual.transformer.resblocks.7.attn.in_proj",
94
+ "transformer.h.23.attn.c_proj",
95
+ "transformer.visual.transformer.resblocks.38.attn.in_proj",
96
+ "transformer.h.19.attn.c_attn",
97
+ "transformer.h.1.attn.c_proj",
98
+ "transformer.visual.transformer.resblocks.2.attn.in_proj",
99
+ "transformer.h.4.mlp.w2",
100
+ "transformer.h.15.mlp.c_proj",
101
+ "transformer.h.4.mlp.c_proj",
102
+ "transformer.h.19.mlp.w2",
103
+ "transformer.h.12.mlp.w1",
104
+ "transformer.visual.transformer.resblocks.30.attn.in_proj",
105
+ "transformer.visual.transformer.resblocks.9.attn.out_proj",
106
+ "transformer.h.28.mlp.c_proj",
107
+ "transformer.h.1.attn.c_attn",
108
+ "transformer.h.8.attn.c_proj",
109
+ "transformer.visual.transformer.resblocks.40.mlp.c_fc",
110
+ "transformer.visual.transformer.resblocks.25.attn.out_proj",
111
+ "transformer.h.4.mlp.w1",
112
+ "transformer.visual.transformer.resblocks.31.mlp.c_fc",
113
+ "transformer.visual.transformer.resblocks.7.mlp.c_proj",
114
+ "transformer.visual.transformer.resblocks.33.attn.in_proj",
115
+ "transformer.h.0.mlp.w1",
116
+ "transformer.visual.transformer.resblocks.8.mlp.c_proj",
117
+ "transformer.h.20.mlp.w2",
118
+ "transformer.visual.transformer.resblocks.37.mlp.c_fc",
119
+ "transformer.visual.transformer.resblocks.43.attn.in_proj",
120
+ "transformer.visual.transformer.resblocks.37.attn.out_proj",
121
+ "transformer.h.25.mlp.w1",
122
+ "transformer.visual.transformer.resblocks.2.mlp.c_proj",
123
+ "transformer.h.27.mlp.w2",
124
+ "transformer.visual.transformer.resblocks.47.attn.out_proj",
125
+ "transformer.visual.transformer.resblocks.18.attn.in_proj",
126
+ "transformer.visual.transformer.resblocks.12.mlp.c_fc",
127
+ "transformer.visual.transformer.resblocks.11.attn.in_proj",
128
+ "transformer.h.14.mlp.c_proj",
129
+ "transformer.h.7.attn.c_attn",
130
+ "transformer.h.10.mlp.w2",
131
+ "transformer.h.11.mlp.w2",
132
+ "transformer.visual.transformer.resblocks.13.attn.out_proj",
133
+ "transformer.h.24.mlp.w1",
134
+ "transformer.h.0.mlp.c_proj",
135
+ "transformer.h.24.attn.c_proj",
136
+ "transformer.visual.transformer.resblocks.10.mlp.c_fc",
137
+ "transformer.visual.transformer.resblocks.33.mlp.c_proj",
138
+ "transformer.visual.transformer.resblocks.35.attn.in_proj",
139
+ "transformer.visual.transformer.resblocks.38.attn.out_proj",
140
+ "transformer.h.2.mlp.w2",
141
+ "transformer.visual.transformer.resblocks.40.attn.in_proj",
142
+ "transformer.h.25.mlp.w2",
143
+ "transformer.visual.transformer.resblocks.17.mlp.c_proj",
144
+ "transformer.visual.transformer.resblocks.18.mlp.c_proj",
145
+ "transformer.h.2.attn.c_proj",
146
+ "transformer.visual.transformer.resblocks.0.mlp.c_proj",
147
+ "transformer.visual.transformer.resblocks.23.mlp.c_proj",
148
+ "transformer.visual.transformer.resblocks.2.attn.out_proj",
149
+ "transformer.h.16.mlp.w2",
150
+ "transformer.h.29.mlp.c_proj",
151
+ "transformer.visual.transformer.resblocks.21.attn.in_proj",
152
+ "transformer.visual.transformer.resblocks.7.mlp.c_fc",
153
+ "transformer.visual.transformer.resblocks.8.attn.out_proj",
154
+ "transformer.visual.transformer.resblocks.28.attn.in_proj",
155
+ "transformer.h.11.mlp.w1",
156
+ "transformer.visual.transformer.resblocks.41.mlp.c_fc",
157
+ "transformer.h.18.attn.c_proj",
158
+ "transformer.visual.transformer.resblocks.29.mlp.c_fc",
159
+ "transformer.visual.transformer.resblocks.32.mlp.c_proj",
160
+ "transformer.visual.transformer.resblocks.28.attn.out_proj",
161
+ "transformer.h.6.attn.c_attn",
162
+ "transformer.visual.transformer.resblocks.3.attn.out_proj",
163
+ "transformer.h.25.attn.c_attn",
164
+ "transformer.h.28.attn.c_proj",
165
+ "transformer.visual.transformer.resblocks.16.attn.out_proj",
166
+ "transformer.visual.transformer.resblocks.20.mlp.c_proj",
167
+ "transformer.visual.transformer.resblocks.45.attn.in_proj",
168
+ "transformer.h.8.mlp.c_proj",
169
+ "transformer.visual.transformer.resblocks.29.attn.out_proj",
170
+ "transformer.h.27.mlp.c_proj",
171
+ "transformer.visual.transformer.resblocks.11.mlp.c_fc",
172
+ "transformer.visual.transformer.resblocks.11.attn.out_proj",
173
+ "transformer.visual.transformer.resblocks.10.attn.in_proj",
174
+ "transformer.visual.transformer.resblocks.27.mlp.c_fc",
175
+ "transformer.h.4.attn.c_proj",
176
+ "transformer.visual.transformer.resblocks.43.mlp.c_fc",
177
+ "transformer.visual.transformer.resblocks.22.mlp.c_proj",
178
+ "transformer.h.22.attn.c_proj",
179
+ "transformer.visual.transformer.resblocks.5.mlp.c_proj",
180
+ "transformer.h.22.mlp.c_proj",
181
+ "transformer.visual.transformer.resblocks.7.attn.out_proj",
182
+ "transformer.visual.transformer.resblocks.14.attn.out_proj",
183
+ "transformer.h.30.mlp.w1",
184
+ "transformer.h.14.attn.c_attn",
185
+ "transformer.h.4.attn.c_attn",
186
+ "transformer.visual.transformer.resblocks.46.mlp.c_fc",
187
+ "transformer.visual.transformer.resblocks.27.attn.out_proj",
188
+ "transformer.visual.transformer.resblocks.41.attn.in_proj",
189
+ "transformer.visual.transformer.resblocks.13.mlp.c_fc",
190
+ "transformer.h.12.attn.c_proj",
191
+ "transformer.visual.transformer.resblocks.32.attn.out_proj",
192
+ "transformer.visual.transformer.resblocks.18.mlp.c_fc",
193
+ "transformer.visual.transformer.resblocks.45.attn.out_proj",
194
+ "transformer.visual.transformer.resblocks.44.mlp.c_proj",
195
+ "transformer.visual.transformer.resblocks.35.mlp.c_proj",
196
+ "transformer.visual.transformer.resblocks.20.attn.out_proj",
197
+ "transformer.visual.transformer.resblocks.19.attn.in_proj",
198
+ "transformer.h.17.mlp.w2",
199
+ "transformer.h.10.attn.c_attn",
200
+ "transformer.visual.transformer.resblocks.45.mlp.c_proj",
201
+ "transformer.visual.transformer.resblocks.39.mlp.c_proj",
202
+ "transformer.h.7.mlp.w1",
203
+ "transformer.h.14.mlp.w1",
204
+ "transformer.visual.transformer.resblocks.2.mlp.c_fc",
205
+ "transformer.h.21.mlp.w1",
206
+ "transformer.visual.transformer.resblocks.30.attn.out_proj",
207
+ "transformer.h.19.mlp.c_proj",
208
+ "transformer.h.5.mlp.w1",
209
+ "transformer.visual.transformer.resblocks.47.mlp.c_fc",
210
+ "transformer.visual.transformer.resblocks.38.mlp.c_fc",
211
+ "transformer.h.10.mlp.c_proj",
212
+ "transformer.visual.transformer.resblocks.28.mlp.c_proj",
213
+ "transformer.visual.transformer.resblocks.25.attn.in_proj",
214
+ "transformer.h.9.attn.c_attn",
215
+ "transformer.visual.transformer.resblocks.0.mlp.c_fc",
216
+ "transformer.h.29.attn.c_proj",
217
+ "transformer.h.5.mlp.w2",
218
+ "transformer.h.30.attn.c_attn",
219
+ "transformer.h.1.mlp.w1",
220
+ "transformer.visual.transformer.resblocks.8.mlp.c_fc",
221
+ "transformer.h.19.mlp.w1",
222
+ "transformer.h.18.attn.c_attn",
223
+ "transformer.h.11.attn.c_proj",
224
+ "transformer.h.5.mlp.c_proj",
225
+ "transformer.visual.transformer.resblocks.45.mlp.c_fc",
226
+ "transformer.visual.transformer.resblocks.47.mlp.c_proj",
227
+ "transformer.h.9.attn.c_proj",
228
+ "transformer.visual.transformer.resblocks.6.attn.out_proj",
229
+ "transformer.h.26.mlp.c_proj",
230
+ "transformer.visual.transformer.resblocks.24.mlp.c_proj",
231
+ "transformer.h.31.attn.c_attn",
232
+ "transformer.h.13.mlp.w1",
233
+ "transformer.visual.transformer.resblocks.39.attn.in_proj",
234
+ "transformer.h.20.mlp.w1",
235
+ "transformer.visual.transformer.resblocks.22.attn.in_proj",
236
+ "transformer.visual.transformer.resblocks.35.mlp.c_fc",
237
+ "transformer.visual.transformer.resblocks.24.attn.in_proj",
238
+ "transformer.h.16.mlp.w1",
239
+ "transformer.visual.transformer.resblocks.6.mlp.c_proj",
240
+ "transformer.h.6.mlp.c_proj",
241
+ "transformer.visual.transformer.resblocks.26.mlp.c_proj",
242
+ "transformer.visual.transformer.resblocks.35.attn.out_proj",
243
+ "transformer.h.24.mlp.w2",
244
+ "transformer.visual.transformer.resblocks.31.attn.in_proj",
245
+ "transformer.visual.transformer.resblocks.15.mlp.c_fc",
246
+ "transformer.h.2.mlp.w1",
247
+ "transformer.h.31.mlp.c_proj",
248
+ "transformer.h.13.attn.c_attn",
249
+ "transformer.visual.transformer.resblocks.14.attn.in_proj",
250
+ "transformer.h.12.mlp.w2",
251
+ "transformer.visual.transformer.resblocks.34.mlp.c_proj",
252
+ "transformer.h.26.mlp.w2",
253
+ "transformer.visual.transformer.resblocks.5.mlp.c_fc",
254
+ "transformer.h.5.attn.c_proj",
255
+ "transformer.h.9.mlp.w2",
256
+ "transformer.h.15.mlp.w2",
257
+ "transformer.h.12.attn.c_attn",
258
+ "transformer.visual.transformer.resblocks.10.mlp.c_proj",
259
+ "transformer.h.28.mlp.w1",
260
+ "transformer.visual.transformer.resblocks.46.attn.out_proj",
261
+ "transformer.h.18.mlp.c_proj",
262
+ "transformer.h.15.attn.c_proj",
263
+ "transformer.visual.transformer.resblocks.44.attn.in_proj",
264
+ "transformer.visual.transformer.resblocks.12.attn.out_proj",
265
+ "transformer.visual.transformer.resblocks.1.mlp.c_fc",
266
+ "transformer.visual.transformer.resblocks.15.attn.out_proj",
267
+ "transformer.h.17.mlp.w1",
268
+ "transformer.visual.transformer.resblocks.46.attn.in_proj",
269
+ "transformer.h.2.attn.c_attn",
270
+ "transformer.h.25.attn.c_proj",
271
+ "transformer.h.14.attn.c_proj",
272
+ "transformer.h.26.attn.c_proj",
273
+ "transformer.h.31.mlp.w1",
274
+ "transformer.h.23.mlp.w2",
275
+ "transformer.visual.transformer.resblocks.26.attn.out_proj",
276
+ "transformer.h.20.mlp.c_proj",
277
+ "transformer.visual.transformer.resblocks.6.mlp.c_fc",
278
+ "transformer.h.27.mlp.w1",
279
+ "transformer.h.7.attn.c_proj",
280
+ "transformer.visual.transformer.resblocks.15.mlp.c_proj",
281
+ "transformer.h.16.mlp.c_proj",
282
+ "transformer.visual.transformer.resblocks.12.mlp.c_proj",
283
+ "transformer.h.29.mlp.w2",
284
+ "transformer.h.15.mlp.w1",
285
+ "transformer.h.6.mlp.w2",
286
+ "transformer.h.3.attn.c_attn",
287
+ "transformer.h.21.mlp.w2",
288
+ "transformer.visual.transformer.resblocks.16.mlp.c_fc",
289
+ "transformer.visual.transformer.resblocks.23.mlp.c_fc",
290
+ "transformer.visual.transformer.resblocks.34.attn.out_proj",
291
+ "transformer.h.8.attn.c_attn",
292
+ "transformer.visual.transformer.resblocks.43.mlp.c_proj",
293
+ "transformer.visual.transformer.resblocks.14.mlp.c_fc",
294
+ "transformer.visual.transformer.resblocks.24.attn.out_proj",
295
+ "transformer.visual.transformer.resblocks.12.attn.in_proj",
296
+ "transformer.h.25.mlp.c_proj",
297
+ "transformer.visual.transformer.resblocks.46.mlp.c_proj",
298
+ "transformer.h.7.mlp.c_proj",
299
+ "transformer.h.15.attn.c_attn",
300
+ "transformer.visual.transformer.resblocks.36.attn.in_proj",
301
+ "transformer.h.26.attn.c_attn",
302
+ "transformer.h.0.attn.c_proj",
303
+ "transformer.visual.transformer.resblocks.9.mlp.c_fc",
304
+ "transformer.h.19.attn.c_proj",
305
+ "transformer.visual.transformer.resblocks.29.attn.in_proj",
306
+ "transformer.visual.transformer.resblocks.13.attn.in_proj",
307
+ "transformer.visual.transformer.resblocks.23.attn.in_proj",
308
+ "transformer.visual.transformer.resblocks.36.attn.out_proj",
309
+ "transformer.h.3.mlp.c_proj",
310
+ "transformer.h.27.attn.c_proj",
311
+ "transformer.h.31.attn.c_proj",
312
+ "transformer.visual.transformer.resblocks.19.attn.out_proj",
313
+ "transformer.h.0.mlp.w2",
314
+ "transformer.h.17.attn.c_proj",
315
+ "transformer.h.30.mlp.w2",
316
+ "transformer.visual.transformer.resblocks.24.mlp.c_fc",
317
+ "transformer.h.28.attn.c_attn",
318
+ "transformer.visual.transformer.resblocks.16.mlp.c_proj",
319
+ "transformer.visual.transformer.resblocks.36.mlp.c_fc",
320
+ "transformer.h.30.mlp.c_proj",
321
+ "transformer.visual.transformer.resblocks.37.attn.in_proj",
322
+ "transformer.visual.transformer.resblocks.20.mlp.c_fc",
323
+ "transformer.h.9.mlp.c_proj",
324
+ "transformer.visual.transformer.resblocks.43.attn.out_proj",
325
+ "transformer.visual.transformer.resblocks.42.attn.in_proj",
326
+ "transformer.h.1.mlp.w2",
327
+ "transformer.h.6.mlp.w1",
328
+ "transformer.visual.transformer.resblocks.17.attn.out_proj",
329
+ "transformer.visual.transformer.resblocks.33.mlp.c_fc",
330
+ "transformer.h.5.attn.c_attn",
331
+ "transformer.h.8.mlp.w2",
332
+ "transformer.h.23.mlp.c_proj",
333
+ "transformer.h.20.attn.c_attn",
334
+ "transformer.visual.transformer.resblocks.4.mlp.c_fc",
335
+ "transformer.h.31.mlp.w2",
336
+ "transformer.h.9.mlp.w1",
337
+ "transformer.h.12.mlp.c_proj",
338
+ "transformer.visual.transformer.resblocks.1.attn.out_proj",
339
+ "transformer.visual.transformer.resblocks.30.mlp.c_fc",
340
+ "transformer.visual.transformer.resblocks.1.attn.in_proj",
341
+ "transformer.visual.transformer.resblocks.32.attn.in_proj",
342
+ "transformer.h.16.attn.c_proj",
343
+ "transformer.visual.transformer.resblocks.42.mlp.c_proj",
344
+ "transformer.h.3.mlp.w1",
345
+ "transformer.visual.transformer.resblocks.9.mlp.c_proj",
346
+ "transformer.h.18.mlp.w1",
347
+ "transformer.visual.transformer.resblocks.44.mlp.c_fc",
348
+ "transformer.h.21.mlp.c_proj",
349
+ "transformer.visual.transformer.resblocks.5.attn.out_proj",
350
+ "transformer.visual.transformer.resblocks.34.mlp.c_fc",
351
+ "transformer.visual.transformer.resblocks.42.attn.out_proj",
352
+ "transformer.visual.transformer.resblocks.41.attn.out_proj",
353
+ "transformer.h.10.mlp.w1",
354
+ "transformer.visual.transformer.resblocks.19.mlp.c_fc",
355
+ "transformer.visual.transformer.resblocks.13.mlp.c_proj",
356
+ "transformer.h.21.attn.c_proj",
357
+ "transformer.visual.transformer.resblocks.6.attn.in_proj",
358
+ "transformer.visual.transformer.resblocks.23.attn.out_proj",
359
+ "transformer.visual.transformer.resblocks.40.attn.out_proj",
360
+ "transformer.visual.transformer.resblocks.36.mlp.c_proj",
361
+ "transformer.h.2.mlp.c_proj",
362
+ "transformer.visual.transformer.resblocks.25.mlp.c_fc",
363
+ "transformer.h.22.attn.c_attn",
364
+ "transformer.h.23.mlp.w1",
365
+ "transformer.h.20.attn.c_proj",
366
+ "transformer.h.23.attn.c_attn",
367
+ "transformer.visual.transformer.resblocks.37.mlp.c_proj",
368
+ "transformer.h.26.mlp.w1",
369
+ "transformer.h.18.mlp.w2",
370
+ "transformer.visual.transformer.resblocks.21.mlp.c_proj",
371
+ "transformer.visual.transformer.resblocks.31.attn.out_proj",
372
+ "transformer.h.24.attn.c_attn",
373
+ "transformer.visual.transformer.resblocks.10.attn.out_proj",
374
+ "transformer.visual.transformer.resblocks.30.mlp.c_proj",
375
+ "transformer.visual.transformer.resblocks.38.mlp.c_proj"
376
+ ],
377
+ "task_type": "CAUSAL_LM",
378
+ "use_dora": false,
379
+ "use_rslora": false
380
+ }
checkpoint-2000/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57751af01e12bd284ebbcc91731f4702389a0a71f96126bbd38615dc92b5dacf
3
+ size 469105640
checkpoint-2000/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step2000
checkpoint-2000/qwen.tiktoken ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-2000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0715a8783a5a13f7dbeaf0a2c771ae5d1e63d85e4ab7590f4639ca34ca9d44ee
3
+ size 14960
checkpoint-2000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0db04156eec797d71f821e1a8ef1067502546f617400ccfd7ebb79a443772c00
3
+ size 14960
checkpoint-2000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b25c100629b37e7df637936b46c6e551b007d14236fa4263774eee275860dfb2
3
+ size 14960
checkpoint-2000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04c0464b9610bcd72269c77ebee7c9f59d3e72ab66b76b333f8c0965f3b65010
3
+ size 14960
checkpoint-2000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a92e67779de6b66020b0d0f30877f89cdcf7536938bb2812dd4b249f5bdb5ac
3
+ size 1064
checkpoint-2000/special_tokens_map.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "pad_token": "<|endoftext|>"
3
+ }
checkpoint-2000/tokenizer_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {},
3
+ "auto_map": {
4
+ "AutoTokenizer": [
5
+ "Qwen/Qwen-VL-Chat--tokenization_qwen.QWenTokenizer",
6
+ null
7
+ ]
8
+ },
9
+ "clean_up_tokenization_spaces": true,
10
+ "model_max_length": 768,
11
+ "pad_token": "<|endoftext|>",
12
+ "padding_side": "right",
13
+ "tokenizer_class": "QWenTokenizer"
14
+ }
checkpoint-2000/trainer_state.json ADDED
@@ -0,0 +1,1433 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.26676448030944677,
5
+ "eval_steps": 500,
6
+ "global_step": 2000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.001333822401547234,
13
+ "grad_norm": 5.80256772259428,
14
+ "learning_rate": 4e-06,
15
+ "loss": 1.0498,
16
+ "step": 10
17
+ },
18
+ {
19
+ "epoch": 0.002667644803094468,
20
+ "grad_norm": 33.895696082107904,
21
+ "learning_rate": 8e-06,
22
+ "loss": 1.0653,
23
+ "step": 20
24
+ },
25
+ {
26
+ "epoch": 0.004001467204641702,
27
+ "grad_norm": 5.523348234283539,
28
+ "learning_rate": 1.2e-05,
29
+ "loss": 1.0341,
30
+ "step": 30
31
+ },
32
+ {
33
+ "epoch": 0.005335289606188936,
34
+ "grad_norm": 11.1556403156453,
35
+ "learning_rate": 1.6e-05,
36
+ "loss": 0.9692,
37
+ "step": 40
38
+ },
39
+ {
40
+ "epoch": 0.00666911200773617,
41
+ "grad_norm": 3.7375231126561825,
42
+ "learning_rate": 1.9999999999999998e-05,
43
+ "loss": 0.9554,
44
+ "step": 50
45
+ },
46
+ {
47
+ "epoch": 0.008002934409283404,
48
+ "grad_norm": 8.43538339698909,
49
+ "learning_rate": 2.4e-05,
50
+ "loss": 0.8965,
51
+ "step": 60
52
+ },
53
+ {
54
+ "epoch": 0.009336756810830639,
55
+ "grad_norm": 13.403454896011478,
56
+ "learning_rate": 2.8e-05,
57
+ "loss": 0.8273,
58
+ "step": 70
59
+ },
60
+ {
61
+ "epoch": 0.010670579212377872,
62
+ "grad_norm": 3.95522050766088,
63
+ "learning_rate": 2.9999966406213696e-05,
64
+ "loss": 0.7837,
65
+ "step": 80
66
+ },
67
+ {
68
+ "epoch": 0.012004401613925107,
69
+ "grad_norm": 36.799552052300854,
70
+ "learning_rate": 2.9999697656826056e-05,
71
+ "loss": 0.8288,
72
+ "step": 90
73
+ },
74
+ {
75
+ "epoch": 0.01333822401547234,
76
+ "grad_norm": 1.6305479563258536,
77
+ "learning_rate": 2.9999160162865885e-05,
78
+ "loss": 0.7778,
79
+ "step": 100
80
+ },
81
+ {
82
+ "epoch": 0.014672046417019574,
83
+ "grad_norm": 2.159536648784889,
84
+ "learning_rate": 2.9998353933963273e-05,
85
+ "loss": 0.7616,
86
+ "step": 110
87
+ },
88
+ {
89
+ "epoch": 0.016005868818566808,
90
+ "grad_norm": 3.397321425707004,
91
+ "learning_rate": 2.999727898456315e-05,
92
+ "loss": 0.7594,
93
+ "step": 120
94
+ },
95
+ {
96
+ "epoch": 0.017339691220114042,
97
+ "grad_norm": 4.772220837365037,
98
+ "learning_rate": 2.999593533392503e-05,
99
+ "loss": 0.756,
100
+ "step": 130
101
+ },
102
+ {
103
+ "epoch": 0.018673513621661277,
104
+ "grad_norm": 2.4845945633126885,
105
+ "learning_rate": 2.9994323006122654e-05,
106
+ "loss": 0.7601,
107
+ "step": 140
108
+ },
109
+ {
110
+ "epoch": 0.02000733602320851,
111
+ "grad_norm": 3.591682569169127,
112
+ "learning_rate": 2.9992442030043557e-05,
113
+ "loss": 0.7894,
114
+ "step": 150
115
+ },
116
+ {
117
+ "epoch": 0.021341158424755743,
118
+ "grad_norm": 2.5679458807474416,
119
+ "learning_rate": 2.9990292439388565e-05,
120
+ "loss": 0.7093,
121
+ "step": 160
122
+ },
123
+ {
124
+ "epoch": 0.022674980826302978,
125
+ "grad_norm": 1.9412569107551652,
126
+ "learning_rate": 2.9987874272671168e-05,
127
+ "loss": 0.706,
128
+ "step": 170
129
+ },
130
+ {
131
+ "epoch": 0.024008803227850213,
132
+ "grad_norm": 3.2667097270489,
133
+ "learning_rate": 2.9985187573216855e-05,
134
+ "loss": 0.7586,
135
+ "step": 180
136
+ },
137
+ {
138
+ "epoch": 0.025342625629397444,
139
+ "grad_norm": 4.4208737375400675,
140
+ "learning_rate": 2.998223238916232e-05,
141
+ "loss": 0.6985,
142
+ "step": 190
143
+ },
144
+ {
145
+ "epoch": 0.02667644803094468,
146
+ "grad_norm": 5.515966302183704,
147
+ "learning_rate": 2.9979008773454618e-05,
148
+ "loss": 0.7323,
149
+ "step": 200
150
+ },
151
+ {
152
+ "epoch": 0.028010270432491914,
153
+ "grad_norm": 2.964165450396077,
154
+ "learning_rate": 2.997551678385019e-05,
155
+ "loss": 0.7603,
156
+ "step": 210
157
+ },
158
+ {
159
+ "epoch": 0.02934409283403915,
160
+ "grad_norm": 3.0952916783456197,
161
+ "learning_rate": 2.997175648291384e-05,
162
+ "loss": 0.7421,
163
+ "step": 220
164
+ },
165
+ {
166
+ "epoch": 0.03067791523558638,
167
+ "grad_norm": 4.213588693904103,
168
+ "learning_rate": 2.996772793801763e-05,
169
+ "loss": 0.7322,
170
+ "step": 230
171
+ },
172
+ {
173
+ "epoch": 0.032011737637133615,
174
+ "grad_norm": 1.8568586103139084,
175
+ "learning_rate": 2.996343122133965e-05,
176
+ "loss": 0.6922,
177
+ "step": 240
178
+ },
179
+ {
180
+ "epoch": 0.033345560038680847,
181
+ "grad_norm": 4.494146778909846,
182
+ "learning_rate": 2.9958866409862745e-05,
183
+ "loss": 0.7244,
184
+ "step": 250
185
+ },
186
+ {
187
+ "epoch": 0.034679382440228085,
188
+ "grad_norm": 7.438170074282725,
189
+ "learning_rate": 2.9954033585373108e-05,
190
+ "loss": 0.7093,
191
+ "step": 260
192
+ },
193
+ {
194
+ "epoch": 0.036013204841775316,
195
+ "grad_norm": 2.3744787346857015,
196
+ "learning_rate": 2.994893283445885e-05,
197
+ "loss": 0.6983,
198
+ "step": 270
199
+ },
200
+ {
201
+ "epoch": 0.037347027243322554,
202
+ "grad_norm": 1.4722011682616383,
203
+ "learning_rate": 2.9943564248508415e-05,
204
+ "loss": 0.6781,
205
+ "step": 280
206
+ },
207
+ {
208
+ "epoch": 0.038680849644869786,
209
+ "grad_norm": 3.3397620832486075,
210
+ "learning_rate": 2.9937927923708966e-05,
211
+ "loss": 0.7399,
212
+ "step": 290
213
+ },
214
+ {
215
+ "epoch": 0.04001467204641702,
216
+ "grad_norm": 5.05063397044549,
217
+ "learning_rate": 2.993202396104465e-05,
218
+ "loss": 0.7671,
219
+ "step": 300
220
+ },
221
+ {
222
+ "epoch": 0.041348494447964255,
223
+ "grad_norm": 3.0128431385936767,
224
+ "learning_rate": 2.9925852466294795e-05,
225
+ "loss": 0.7015,
226
+ "step": 310
227
+ },
228
+ {
229
+ "epoch": 0.04268231684951149,
230
+ "grad_norm": 2.0161342716764237,
231
+ "learning_rate": 2.9919413550032014e-05,
232
+ "loss": 0.7009,
233
+ "step": 320
234
+ },
235
+ {
236
+ "epoch": 0.04401613925105872,
237
+ "grad_norm": 1.3114004070324985,
238
+ "learning_rate": 2.991270732762022e-05,
239
+ "loss": 0.7153,
240
+ "step": 330
241
+ },
242
+ {
243
+ "epoch": 0.045349961652605957,
244
+ "grad_norm": 18.493625676806268,
245
+ "learning_rate": 2.990573391921255e-05,
246
+ "loss": 0.7518,
247
+ "step": 340
248
+ },
249
+ {
250
+ "epoch": 0.04668378405415319,
251
+ "grad_norm": 2.9526764059703567,
252
+ "learning_rate": 2.989849344974924e-05,
253
+ "loss": 0.7133,
254
+ "step": 350
255
+ },
256
+ {
257
+ "epoch": 0.048017606455700426,
258
+ "grad_norm": 5.26274958582726,
259
+ "learning_rate": 2.9890986048955368e-05,
260
+ "loss": 0.7139,
261
+ "step": 360
262
+ },
263
+ {
264
+ "epoch": 0.04935142885724766,
265
+ "grad_norm": 3.5319788357887933,
266
+ "learning_rate": 2.9883211851338516e-05,
267
+ "loss": 0.7084,
268
+ "step": 370
269
+ },
270
+ {
271
+ "epoch": 0.05068525125879489,
272
+ "grad_norm": 7.607269935902469,
273
+ "learning_rate": 2.9875170996186392e-05,
274
+ "loss": 0.7309,
275
+ "step": 380
276
+ },
277
+ {
278
+ "epoch": 0.05201907366034213,
279
+ "grad_norm": 2.3456663308287253,
280
+ "learning_rate": 2.986686362756431e-05,
281
+ "loss": 0.6827,
282
+ "step": 390
283
+ },
284
+ {
285
+ "epoch": 0.05335289606188936,
286
+ "grad_norm": 2.176182050789012,
287
+ "learning_rate": 2.9858289894312617e-05,
288
+ "loss": 0.6995,
289
+ "step": 400
290
+ },
291
+ {
292
+ "epoch": 0.0546867184634366,
293
+ "grad_norm": 11.171630173781537,
294
+ "learning_rate": 2.9849449950044036e-05,
295
+ "loss": 0.7335,
296
+ "step": 410
297
+ },
298
+ {
299
+ "epoch": 0.05602054086498383,
300
+ "grad_norm": 6.63441431767892,
301
+ "learning_rate": 2.984034395314088e-05,
302
+ "loss": 0.7031,
303
+ "step": 420
304
+ },
305
+ {
306
+ "epoch": 0.05735436326653106,
307
+ "grad_norm": 2.861620412225736,
308
+ "learning_rate": 2.983097206675227e-05,
309
+ "loss": 0.6559,
310
+ "step": 430
311
+ },
312
+ {
313
+ "epoch": 0.0586881856680783,
314
+ "grad_norm": 5.523165036486206,
315
+ "learning_rate": 2.9821334458791156e-05,
316
+ "loss": 0.726,
317
+ "step": 440
318
+ },
319
+ {
320
+ "epoch": 0.06002200806962553,
321
+ "grad_norm": 3.5602243751368197,
322
+ "learning_rate": 2.9811431301931344e-05,
323
+ "loss": 0.7202,
324
+ "step": 450
325
+ },
326
+ {
327
+ "epoch": 0.06135583047117276,
328
+ "grad_norm": 11.333380381168622,
329
+ "learning_rate": 2.9801262773604377e-05,
330
+ "loss": 0.7189,
331
+ "step": 460
332
+ },
333
+ {
334
+ "epoch": 0.06268965287271999,
335
+ "grad_norm": 14.159758615106613,
336
+ "learning_rate": 2.9790829055996398e-05,
337
+ "loss": 0.7267,
338
+ "step": 470
339
+ },
340
+ {
341
+ "epoch": 0.06402347527426723,
342
+ "grad_norm": 9.009079485918289,
343
+ "learning_rate": 2.978013033604483e-05,
344
+ "loss": 0.748,
345
+ "step": 480
346
+ },
347
+ {
348
+ "epoch": 0.06535729767581447,
349
+ "grad_norm": 1.9682648681675994,
350
+ "learning_rate": 2.976916680543506e-05,
351
+ "loss": 0.7369,
352
+ "step": 490
353
+ },
354
+ {
355
+ "epoch": 0.06669112007736169,
356
+ "grad_norm": 2.9278164598232777,
357
+ "learning_rate": 2.975793866059701e-05,
358
+ "loss": 0.7037,
359
+ "step": 500
360
+ },
361
+ {
362
+ "epoch": 0.06802494247890893,
363
+ "grad_norm": 5.5563562303649885,
364
+ "learning_rate": 2.9746446102701606e-05,
365
+ "loss": 0.6986,
366
+ "step": 510
367
+ },
368
+ {
369
+ "epoch": 0.06935876488045617,
370
+ "grad_norm": 4.036767303783137,
371
+ "learning_rate": 2.9734689337657157e-05,
372
+ "loss": 0.7119,
373
+ "step": 520
374
+ },
375
+ {
376
+ "epoch": 0.07069258728200341,
377
+ "grad_norm": 1.9856990692088847,
378
+ "learning_rate": 2.9722668576105703e-05,
379
+ "loss": 0.7205,
380
+ "step": 530
381
+ },
382
+ {
383
+ "epoch": 0.07202640968355063,
384
+ "grad_norm": 5.200308739226583,
385
+ "learning_rate": 2.971038403341921e-05,
386
+ "loss": 0.6918,
387
+ "step": 540
388
+ },
389
+ {
390
+ "epoch": 0.07336023208509787,
391
+ "grad_norm": 2.237349124701919,
392
+ "learning_rate": 2.9697835929695727e-05,
393
+ "loss": 0.7339,
394
+ "step": 550
395
+ },
396
+ {
397
+ "epoch": 0.07469405448664511,
398
+ "grad_norm": 1.6388680632753365,
399
+ "learning_rate": 2.968502448975544e-05,
400
+ "loss": 0.7086,
401
+ "step": 560
402
+ },
403
+ {
404
+ "epoch": 0.07602787688819233,
405
+ "grad_norm": 2.8545575025135244,
406
+ "learning_rate": 2.967194994313663e-05,
407
+ "loss": 0.678,
408
+ "step": 570
409
+ },
410
+ {
411
+ "epoch": 0.07736169928973957,
412
+ "grad_norm": 2.674647983669599,
413
+ "learning_rate": 2.9658612524091594e-05,
414
+ "loss": 0.7119,
415
+ "step": 580
416
+ },
417
+ {
418
+ "epoch": 0.07869552169128681,
419
+ "grad_norm": 2.489047760330112,
420
+ "learning_rate": 2.9645012471582406e-05,
421
+ "loss": 0.7382,
422
+ "step": 590
423
+ },
424
+ {
425
+ "epoch": 0.08002934409283403,
426
+ "grad_norm": 5.509352102248308,
427
+ "learning_rate": 2.9631150029276662e-05,
428
+ "loss": 0.738,
429
+ "step": 600
430
+ },
431
+ {
432
+ "epoch": 0.08136316649438127,
433
+ "grad_norm": 3.6489235270404015,
434
+ "learning_rate": 2.9617025445543114e-05,
435
+ "loss": 0.7018,
436
+ "step": 610
437
+ },
438
+ {
439
+ "epoch": 0.08269698889592851,
440
+ "grad_norm": 2.7813651243235697,
441
+ "learning_rate": 2.9602638973447218e-05,
442
+ "loss": 0.7381,
443
+ "step": 620
444
+ },
445
+ {
446
+ "epoch": 0.08403081129747574,
447
+ "grad_norm": 8.271390523006518,
448
+ "learning_rate": 2.9587990870746574e-05,
449
+ "loss": 0.7168,
450
+ "step": 630
451
+ },
452
+ {
453
+ "epoch": 0.08536463369902297,
454
+ "grad_norm": 1.2460611751687307,
455
+ "learning_rate": 2.9573081399886356e-05,
456
+ "loss": 0.7004,
457
+ "step": 640
458
+ },
459
+ {
460
+ "epoch": 0.08669845610057021,
461
+ "grad_norm": 1.704626418994062,
462
+ "learning_rate": 2.9557910827994568e-05,
463
+ "loss": 0.738,
464
+ "step": 650
465
+ },
466
+ {
467
+ "epoch": 0.08803227850211744,
468
+ "grad_norm": 3.275051693107957,
469
+ "learning_rate": 2.9542479426877283e-05,
470
+ "loss": 0.7017,
471
+ "step": 660
472
+ },
473
+ {
474
+ "epoch": 0.08936610090366467,
475
+ "grad_norm": 11.389990685570503,
476
+ "learning_rate": 2.9526787473013753e-05,
477
+ "loss": 0.7107,
478
+ "step": 670
479
+ },
480
+ {
481
+ "epoch": 0.09069992330521191,
482
+ "grad_norm": 5.591277359184055,
483
+ "learning_rate": 2.9510835247551485e-05,
484
+ "loss": 0.7141,
485
+ "step": 680
486
+ },
487
+ {
488
+ "epoch": 0.09203374570675915,
489
+ "grad_norm": 3.180111568581053,
490
+ "learning_rate": 2.949462303630116e-05,
491
+ "loss": 0.6987,
492
+ "step": 690
493
+ },
494
+ {
495
+ "epoch": 0.09336756810830638,
496
+ "grad_norm": 3.8428068166831753,
497
+ "learning_rate": 2.9478151129731567e-05,
498
+ "loss": 0.7373,
499
+ "step": 700
500
+ },
501
+ {
502
+ "epoch": 0.09470139050985361,
503
+ "grad_norm": 2.231397231771392,
504
+ "learning_rate": 2.9461419822964348e-05,
505
+ "loss": 0.6962,
506
+ "step": 710
507
+ },
508
+ {
509
+ "epoch": 0.09603521291140085,
510
+ "grad_norm": 18.287201889017563,
511
+ "learning_rate": 2.9444429415768726e-05,
512
+ "loss": 0.6723,
513
+ "step": 720
514
+ },
515
+ {
516
+ "epoch": 0.09736903531294808,
517
+ "grad_norm": 4.340932687135137,
518
+ "learning_rate": 2.942718021255617e-05,
519
+ "loss": 0.7151,
520
+ "step": 730
521
+ },
522
+ {
523
+ "epoch": 0.09870285771449532,
524
+ "grad_norm": 2.7813821825484446,
525
+ "learning_rate": 2.940967252237488e-05,
526
+ "loss": 0.7332,
527
+ "step": 740
528
+ },
529
+ {
530
+ "epoch": 0.10003668011604255,
531
+ "grad_norm": 2.3251782912937475,
532
+ "learning_rate": 2.9391906658904296e-05,
533
+ "loss": 0.6751,
534
+ "step": 750
535
+ },
536
+ {
537
+ "epoch": 0.10137050251758978,
538
+ "grad_norm": 8.123799866292751,
539
+ "learning_rate": 2.937388294044946e-05,
540
+ "loss": 0.6886,
541
+ "step": 760
542
+ },
543
+ {
544
+ "epoch": 0.10270432491913702,
545
+ "grad_norm": 1.528579329214318,
546
+ "learning_rate": 2.9355601689935315e-05,
547
+ "loss": 0.7146,
548
+ "step": 770
549
+ },
550
+ {
551
+ "epoch": 0.10403814732068425,
552
+ "grad_norm": 2.0278953433974825,
553
+ "learning_rate": 2.933706323490092e-05,
554
+ "loss": 0.7453,
555
+ "step": 780
556
+ },
557
+ {
558
+ "epoch": 0.10537196972223148,
559
+ "grad_norm": 1.4306270659678864,
560
+ "learning_rate": 2.9318267907493583e-05,
561
+ "loss": 0.6702,
562
+ "step": 790
563
+ },
564
+ {
565
+ "epoch": 0.10670579212377872,
566
+ "grad_norm": 1.5178081087799355,
567
+ "learning_rate": 2.9299216044462903e-05,
568
+ "loss": 0.7346,
569
+ "step": 800
570
+ },
571
+ {
572
+ "epoch": 0.10803961452532596,
573
+ "grad_norm": 9.506616797760028,
574
+ "learning_rate": 2.927990798715475e-05,
575
+ "loss": 0.6558,
576
+ "step": 810
577
+ },
578
+ {
579
+ "epoch": 0.1093734369268732,
580
+ "grad_norm": 2.4597311302505767,
581
+ "learning_rate": 2.926034408150513e-05,
582
+ "loss": 0.726,
583
+ "step": 820
584
+ },
585
+ {
586
+ "epoch": 0.11070725932842042,
587
+ "grad_norm": 12.372180964422007,
588
+ "learning_rate": 2.9240524678034016e-05,
589
+ "loss": 0.7308,
590
+ "step": 830
591
+ },
592
+ {
593
+ "epoch": 0.11204108172996766,
594
+ "grad_norm": 1.4488469801164658,
595
+ "learning_rate": 2.9220450131839037e-05,
596
+ "loss": 0.7072,
597
+ "step": 840
598
+ },
599
+ {
600
+ "epoch": 0.1133749041315149,
601
+ "grad_norm": 8.602946960846197,
602
+ "learning_rate": 2.920012080258912e-05,
603
+ "loss": 0.7234,
604
+ "step": 850
605
+ },
606
+ {
607
+ "epoch": 0.11470872653306212,
608
+ "grad_norm": 1.441195423452674,
609
+ "learning_rate": 2.9179537054518085e-05,
610
+ "loss": 0.6934,
611
+ "step": 860
612
+ },
613
+ {
614
+ "epoch": 0.11604254893460936,
615
+ "grad_norm": 4.318952956999577,
616
+ "learning_rate": 2.9158699256418056e-05,
617
+ "loss": 0.6534,
618
+ "step": 870
619
+ },
620
+ {
621
+ "epoch": 0.1173763713361566,
622
+ "grad_norm": 9.733179695623866,
623
+ "learning_rate": 2.9137607781632913e-05,
624
+ "loss": 0.71,
625
+ "step": 880
626
+ },
627
+ {
628
+ "epoch": 0.11871019373770382,
629
+ "grad_norm": 7.397049093836735,
630
+ "learning_rate": 2.911626300805155e-05,
631
+ "loss": 0.7386,
632
+ "step": 890
633
+ },
634
+ {
635
+ "epoch": 0.12004401613925106,
636
+ "grad_norm": 2.920812240139869,
637
+ "learning_rate": 2.9094665318101155e-05,
638
+ "loss": 0.6789,
639
+ "step": 900
640
+ },
641
+ {
642
+ "epoch": 0.1213778385407983,
643
+ "grad_norm": 1.7031296196271206,
644
+ "learning_rate": 2.9072815098740326e-05,
645
+ "loss": 0.715,
646
+ "step": 910
647
+ },
648
+ {
649
+ "epoch": 0.12271166094234552,
650
+ "grad_norm": 1.5630656172291801,
651
+ "learning_rate": 2.9050712741452136e-05,
652
+ "loss": 0.7136,
653
+ "step": 920
654
+ },
655
+ {
656
+ "epoch": 0.12404548334389276,
657
+ "grad_norm": 7.870543414771234,
658
+ "learning_rate": 2.902835864223715e-05,
659
+ "loss": 0.6669,
660
+ "step": 930
661
+ },
662
+ {
663
+ "epoch": 0.12537930574543998,
664
+ "grad_norm": 4.843671834991794,
665
+ "learning_rate": 2.9005753201606287e-05,
666
+ "loss": 0.7281,
667
+ "step": 940
668
+ },
669
+ {
670
+ "epoch": 0.12671312814698724,
671
+ "grad_norm": 3.010503818258016,
672
+ "learning_rate": 2.8982896824573678e-05,
673
+ "loss": 0.7018,
674
+ "step": 950
675
+ },
676
+ {
677
+ "epoch": 0.12804695054853446,
678
+ "grad_norm": 2.5552186559589654,
679
+ "learning_rate": 2.8959789920649394e-05,
680
+ "loss": 0.7338,
681
+ "step": 960
682
+ },
683
+ {
684
+ "epoch": 0.12938077295008168,
685
+ "grad_norm": 12.306055851495117,
686
+ "learning_rate": 2.893643290383212e-05,
687
+ "loss": 0.6732,
688
+ "step": 970
689
+ },
690
+ {
691
+ "epoch": 0.13071459535162894,
692
+ "grad_norm": 2.16185926525944,
693
+ "learning_rate": 2.891282619260172e-05,
694
+ "loss": 0.7108,
695
+ "step": 980
696
+ },
697
+ {
698
+ "epoch": 0.13204841775317616,
699
+ "grad_norm": 5.992378798792086,
700
+ "learning_rate": 2.8888970209911754e-05,
701
+ "loss": 0.6525,
702
+ "step": 990
703
+ },
704
+ {
705
+ "epoch": 0.13338224015472339,
706
+ "grad_norm": 2.986272238787896,
707
+ "learning_rate": 2.8864865383181893e-05,
708
+ "loss": 0.6655,
709
+ "step": 1000
710
+ },
711
+ {
712
+ "epoch": 0.13471606255627064,
713
+ "grad_norm": 12.855377354582437,
714
+ "learning_rate": 2.8840512144290273e-05,
715
+ "loss": 0.6826,
716
+ "step": 1010
717
+ },
718
+ {
719
+ "epoch": 0.13604988495781786,
720
+ "grad_norm": 2.045979893776702,
721
+ "learning_rate": 2.8815910929565734e-05,
722
+ "loss": 0.6616,
723
+ "step": 1020
724
+ },
725
+ {
726
+ "epoch": 0.1373837073593651,
727
+ "grad_norm": 6.623264301300591,
728
+ "learning_rate": 2.879106217978002e-05,
729
+ "loss": 0.6935,
730
+ "step": 1030
731
+ },
732
+ {
733
+ "epoch": 0.13871752976091234,
734
+ "grad_norm": 2.67990218211766,
735
+ "learning_rate": 2.8765966340139892e-05,
736
+ "loss": 0.6671,
737
+ "step": 1040
738
+ },
739
+ {
740
+ "epoch": 0.14005135216245956,
741
+ "grad_norm": 2.699521523924172,
742
+ "learning_rate": 2.8740623860279116e-05,
743
+ "loss": 0.6763,
744
+ "step": 1050
745
+ },
746
+ {
747
+ "epoch": 0.14138517456400682,
748
+ "grad_norm": 4.1129898011507535,
749
+ "learning_rate": 2.871503519425044e-05,
750
+ "loss": 0.7159,
751
+ "step": 1060
752
+ },
753
+ {
754
+ "epoch": 0.14271899696555404,
755
+ "grad_norm": 2.4592021333659146,
756
+ "learning_rate": 2.8689200800517448e-05,
757
+ "loss": 0.6551,
758
+ "step": 1070
759
+ },
760
+ {
761
+ "epoch": 0.14405281936710126,
762
+ "grad_norm": 5.138500389099849,
763
+ "learning_rate": 2.866312114194634e-05,
764
+ "loss": 0.7214,
765
+ "step": 1080
766
+ },
767
+ {
768
+ "epoch": 0.14538664176864852,
769
+ "grad_norm": 2.822433730666048,
770
+ "learning_rate": 2.8636796685797657e-05,
771
+ "loss": 0.6862,
772
+ "step": 1090
773
+ },
774
+ {
775
+ "epoch": 0.14672046417019574,
776
+ "grad_norm": 3.086468537427806,
777
+ "learning_rate": 2.8610227903717876e-05,
778
+ "loss": 0.6784,
779
+ "step": 1100
780
+ },
781
+ {
782
+ "epoch": 0.14805428657174297,
783
+ "grad_norm": 2.079766793749202,
784
+ "learning_rate": 2.8583415271730994e-05,
785
+ "loss": 0.7065,
786
+ "step": 1110
787
+ },
788
+ {
789
+ "epoch": 0.14938810897329022,
790
+ "grad_norm": 1.659870509072264,
791
+ "learning_rate": 2.855635927022998e-05,
792
+ "loss": 0.7197,
793
+ "step": 1120
794
+ },
795
+ {
796
+ "epoch": 0.15072193137483744,
797
+ "grad_norm": 7.870626779339635,
798
+ "learning_rate": 2.8529060383968175e-05,
799
+ "loss": 0.7305,
800
+ "step": 1130
801
+ },
802
+ {
803
+ "epoch": 0.15205575377638467,
804
+ "grad_norm": 3.0600340899893537,
805
+ "learning_rate": 2.850151910205061e-05,
806
+ "loss": 0.6922,
807
+ "step": 1140
808
+ },
809
+ {
810
+ "epoch": 0.15338957617793192,
811
+ "grad_norm": 3.6147451373702806,
812
+ "learning_rate": 2.847373591792523e-05,
813
+ "loss": 0.7044,
814
+ "step": 1150
815
+ },
816
+ {
817
+ "epoch": 0.15472339857947914,
818
+ "grad_norm": 4.740777951553679,
819
+ "learning_rate": 2.844571132937407e-05,
820
+ "loss": 0.6794,
821
+ "step": 1160
822
+ },
823
+ {
824
+ "epoch": 0.15605722098102637,
825
+ "grad_norm": 3.377522973717319,
826
+ "learning_rate": 2.841744583850431e-05,
827
+ "loss": 0.673,
828
+ "step": 1170
829
+ },
830
+ {
831
+ "epoch": 0.15739104338257362,
832
+ "grad_norm": 4.250656077289992,
833
+ "learning_rate": 2.838893995173932e-05,
834
+ "loss": 0.6975,
835
+ "step": 1180
836
+ },
837
+ {
838
+ "epoch": 0.15872486578412084,
839
+ "grad_norm": 11.73693900915769,
840
+ "learning_rate": 2.836019417980955e-05,
841
+ "loss": 0.6572,
842
+ "step": 1190
843
+ },
844
+ {
845
+ "epoch": 0.16005868818566807,
846
+ "grad_norm": 2.729291714043308,
847
+ "learning_rate": 2.8331209037743387e-05,
848
+ "loss": 0.7247,
849
+ "step": 1200
850
+ },
851
+ {
852
+ "epoch": 0.16139251058721532,
853
+ "grad_norm": 2.347985877636318,
854
+ "learning_rate": 2.8301985044857947e-05,
855
+ "loss": 0.7199,
856
+ "step": 1210
857
+ },
858
+ {
859
+ "epoch": 0.16272633298876255,
860
+ "grad_norm": 2.2534314586033113,
861
+ "learning_rate": 2.8272522724749743e-05,
862
+ "loss": 0.6835,
863
+ "step": 1220
864
+ },
865
+ {
866
+ "epoch": 0.16406015539030977,
867
+ "grad_norm": 3.159583116387406,
868
+ "learning_rate": 2.8242822605285323e-05,
869
+ "loss": 0.7122,
870
+ "step": 1230
871
+ },
872
+ {
873
+ "epoch": 0.16539397779185702,
874
+ "grad_norm": 2.086588782887239,
875
+ "learning_rate": 2.8212885218591812e-05,
876
+ "loss": 0.6949,
877
+ "step": 1240
878
+ },
879
+ {
880
+ "epoch": 0.16672780019340425,
881
+ "grad_norm": 7.284236966547317,
882
+ "learning_rate": 2.8182711101047362e-05,
883
+ "loss": 0.6641,
884
+ "step": 1250
885
+ },
886
+ {
887
+ "epoch": 0.16806162259495147,
888
+ "grad_norm": 3.0369619450249594,
889
+ "learning_rate": 2.815230079327156e-05,
890
+ "loss": 0.6731,
891
+ "step": 1260
892
+ },
893
+ {
894
+ "epoch": 0.16939544499649872,
895
+ "grad_norm": 1.4144726574636068,
896
+ "learning_rate": 2.8121654840115734e-05,
897
+ "loss": 0.6898,
898
+ "step": 1270
899
+ },
900
+ {
901
+ "epoch": 0.17072926739804595,
902
+ "grad_norm": 3.66202356670303,
903
+ "learning_rate": 2.809077379065319e-05,
904
+ "loss": 0.7174,
905
+ "step": 1280
906
+ },
907
+ {
908
+ "epoch": 0.17206308979959317,
909
+ "grad_norm": 4.778073521019285,
910
+ "learning_rate": 2.805965819816937e-05,
911
+ "loss": 0.6186,
912
+ "step": 1290
913
+ },
914
+ {
915
+ "epoch": 0.17339691220114042,
916
+ "grad_norm": 3.9620427201734576,
917
+ "learning_rate": 2.802830862015196e-05,
918
+ "loss": 0.684,
919
+ "step": 1300
920
+ },
921
+ {
922
+ "epoch": 0.17473073460268765,
923
+ "grad_norm": 4.170199740083487,
924
+ "learning_rate": 2.799672561828087e-05,
925
+ "loss": 0.7102,
926
+ "step": 1310
927
+ },
928
+ {
929
+ "epoch": 0.17606455700423487,
930
+ "grad_norm": 2.2612205048804714,
931
+ "learning_rate": 2.79649097584182e-05,
932
+ "loss": 0.7451,
933
+ "step": 1320
934
+ },
935
+ {
936
+ "epoch": 0.17739837940578213,
937
+ "grad_norm": 1.7156828128822517,
938
+ "learning_rate": 2.7932861610598077e-05,
939
+ "loss": 0.6641,
940
+ "step": 1330
941
+ },
942
+ {
943
+ "epoch": 0.17873220180732935,
944
+ "grad_norm": 7.960733847217257,
945
+ "learning_rate": 2.7900581749016466e-05,
946
+ "loss": 0.7365,
947
+ "step": 1340
948
+ },
949
+ {
950
+ "epoch": 0.1800660242088766,
951
+ "grad_norm": 2.5364939682563756,
952
+ "learning_rate": 2.7868070752020865e-05,
953
+ "loss": 0.7078,
954
+ "step": 1350
955
+ },
956
+ {
957
+ "epoch": 0.18139984661042383,
958
+ "grad_norm": 2.7446281678776137,
959
+ "learning_rate": 2.7835329202099944e-05,
960
+ "loss": 0.7214,
961
+ "step": 1360
962
+ },
963
+ {
964
+ "epoch": 0.18273366901197105,
965
+ "grad_norm": 3.2416602016145886,
966
+ "learning_rate": 2.7802357685873117e-05,
967
+ "loss": 0.6757,
968
+ "step": 1370
969
+ },
970
+ {
971
+ "epoch": 0.1840674914135183,
972
+ "grad_norm": 5.225459736579946,
973
+ "learning_rate": 2.7769156794080033e-05,
974
+ "loss": 0.7381,
975
+ "step": 1380
976
+ },
977
+ {
978
+ "epoch": 0.18540131381506553,
979
+ "grad_norm": 5.176692689501482,
980
+ "learning_rate": 2.7735727121569967e-05,
981
+ "loss": 0.7354,
982
+ "step": 1390
983
+ },
984
+ {
985
+ "epoch": 0.18673513621661275,
986
+ "grad_norm": 2.7441883232342574,
987
+ "learning_rate": 2.770206926729121e-05,
988
+ "loss": 0.6937,
989
+ "step": 1400
990
+ },
991
+ {
992
+ "epoch": 0.18806895861816,
993
+ "grad_norm": 2.9792116246243525,
994
+ "learning_rate": 2.7668183834280284e-05,
995
+ "loss": 0.6641,
996
+ "step": 1410
997
+ },
998
+ {
999
+ "epoch": 0.18940278101970723,
1000
+ "grad_norm": 2.4645298487410723,
1001
+ "learning_rate": 2.763407142965117e-05,
1002
+ "loss": 0.6274,
1003
+ "step": 1420
1004
+ },
1005
+ {
1006
+ "epoch": 0.19073660342125445,
1007
+ "grad_norm": 7.245032878035033,
1008
+ "learning_rate": 2.759973266458444e-05,
1009
+ "loss": 0.6962,
1010
+ "step": 1430
1011
+ },
1012
+ {
1013
+ "epoch": 0.1920704258228017,
1014
+ "grad_norm": 5.642209662597534,
1015
+ "learning_rate": 2.756516815431627e-05,
1016
+ "loss": 0.7016,
1017
+ "step": 1440
1018
+ },
1019
+ {
1020
+ "epoch": 0.19340424822434893,
1021
+ "grad_norm": 2.9804981875184526,
1022
+ "learning_rate": 2.7530378518127445e-05,
1023
+ "loss": 0.7331,
1024
+ "step": 1450
1025
+ },
1026
+ {
1027
+ "epoch": 0.19473807062589615,
1028
+ "grad_norm": 7.496561660992361,
1029
+ "learning_rate": 2.7495364379332256e-05,
1030
+ "loss": 0.7234,
1031
+ "step": 1460
1032
+ },
1033
+ {
1034
+ "epoch": 0.1960718930274434,
1035
+ "grad_norm": 1.6139389803246291,
1036
+ "learning_rate": 2.7460126365267335e-05,
1037
+ "loss": 0.7013,
1038
+ "step": 1470
1039
+ },
1040
+ {
1041
+ "epoch": 0.19740571542899063,
1042
+ "grad_norm": 4.618678334755141,
1043
+ "learning_rate": 2.7424665107280402e-05,
1044
+ "loss": 0.6892,
1045
+ "step": 1480
1046
+ },
1047
+ {
1048
+ "epoch": 0.19873953783053785,
1049
+ "grad_norm": 15.494190234738744,
1050
+ "learning_rate": 2.738898124071898e-05,
1051
+ "loss": 0.6785,
1052
+ "step": 1490
1053
+ },
1054
+ {
1055
+ "epoch": 0.2000733602320851,
1056
+ "grad_norm": 3.1680363319798954,
1057
+ "learning_rate": 2.735307540491898e-05,
1058
+ "loss": 0.669,
1059
+ "step": 1500
1060
+ },
1061
+ {
1062
+ "epoch": 0.20140718263363233,
1063
+ "grad_norm": 2.5397562341036224,
1064
+ "learning_rate": 2.7316948243193273e-05,
1065
+ "loss": 0.6726,
1066
+ "step": 1510
1067
+ },
1068
+ {
1069
+ "epoch": 0.20274100503517956,
1070
+ "grad_norm": 4.139021422606072,
1071
+ "learning_rate": 2.7280600402820146e-05,
1072
+ "loss": 0.6706,
1073
+ "step": 1520
1074
+ },
1075
+ {
1076
+ "epoch": 0.2040748274367268,
1077
+ "grad_norm": 2.7422468825646065,
1078
+ "learning_rate": 2.724403253503171e-05,
1079
+ "loss": 0.7078,
1080
+ "step": 1530
1081
+ },
1082
+ {
1083
+ "epoch": 0.20540864983827403,
1084
+ "grad_norm": 2.744225768808104,
1085
+ "learning_rate": 2.7207245295002242e-05,
1086
+ "loss": 0.6821,
1087
+ "step": 1540
1088
+ },
1089
+ {
1090
+ "epoch": 0.20674247223982126,
1091
+ "grad_norm": 2.234040668790152,
1092
+ "learning_rate": 2.7170239341836436e-05,
1093
+ "loss": 0.7451,
1094
+ "step": 1550
1095
+ },
1096
+ {
1097
+ "epoch": 0.2080762946413685,
1098
+ "grad_norm": 2.531733996425376,
1099
+ "learning_rate": 2.7133015338557585e-05,
1100
+ "loss": 0.7205,
1101
+ "step": 1560
1102
+ },
1103
+ {
1104
+ "epoch": 0.20941011704291573,
1105
+ "grad_norm": 2.9772483856455616,
1106
+ "learning_rate": 2.7095573952095727e-05,
1107
+ "loss": 0.7274,
1108
+ "step": 1570
1109
+ },
1110
+ {
1111
+ "epoch": 0.21074393944446296,
1112
+ "grad_norm": 3.317235333047955,
1113
+ "learning_rate": 2.705791585327568e-05,
1114
+ "loss": 0.7309,
1115
+ "step": 1580
1116
+ },
1117
+ {
1118
+ "epoch": 0.2120777618460102,
1119
+ "grad_norm": 1.9652386793628944,
1120
+ "learning_rate": 2.7020041716805014e-05,
1121
+ "loss": 0.7157,
1122
+ "step": 1590
1123
+ },
1124
+ {
1125
+ "epoch": 0.21341158424755743,
1126
+ "grad_norm": 2.93724058913164,
1127
+ "learning_rate": 2.6981952221261986e-05,
1128
+ "loss": 0.7123,
1129
+ "step": 1600
1130
+ },
1131
+ {
1132
+ "epoch": 0.21474540664910466,
1133
+ "grad_norm": 6.395577225750395,
1134
+ "learning_rate": 2.6943648049083366e-05,
1135
+ "loss": 0.6991,
1136
+ "step": 1610
1137
+ },
1138
+ {
1139
+ "epoch": 0.2160792290506519,
1140
+ "grad_norm": 2.4292347967714973,
1141
+ "learning_rate": 2.6905129886552208e-05,
1142
+ "loss": 0.7004,
1143
+ "step": 1620
1144
+ },
1145
+ {
1146
+ "epoch": 0.21741305145219914,
1147
+ "grad_norm": 1.8304810950546353,
1148
+ "learning_rate": 2.6866398423785568e-05,
1149
+ "loss": 0.6941,
1150
+ "step": 1630
1151
+ },
1152
+ {
1153
+ "epoch": 0.2187468738537464,
1154
+ "grad_norm": 2.762870839632077,
1155
+ "learning_rate": 2.682745435472212e-05,
1156
+ "loss": 0.6928,
1157
+ "step": 1640
1158
+ },
1159
+ {
1160
+ "epoch": 0.2200806962552936,
1161
+ "grad_norm": 3.4172019229090917,
1162
+ "learning_rate": 2.6788298377109748e-05,
1163
+ "loss": 0.7344,
1164
+ "step": 1650
1165
+ },
1166
+ {
1167
+ "epoch": 0.22141451865684084,
1168
+ "grad_norm": 2.7483538989548175,
1169
+ "learning_rate": 2.6748931192493017e-05,
1170
+ "loss": 0.7367,
1171
+ "step": 1660
1172
+ },
1173
+ {
1174
+ "epoch": 0.2227483410583881,
1175
+ "grad_norm": 7.314729269236597,
1176
+ "learning_rate": 2.670935350620063e-05,
1177
+ "loss": 0.6849,
1178
+ "step": 1670
1179
+ },
1180
+ {
1181
+ "epoch": 0.2240821634599353,
1182
+ "grad_norm": 3.8688065039432527,
1183
+ "learning_rate": 2.6669566027332767e-05,
1184
+ "loss": 0.6812,
1185
+ "step": 1680
1186
+ },
1187
+ {
1188
+ "epoch": 0.22541598586148254,
1189
+ "grad_norm": 7.10517346658295,
1190
+ "learning_rate": 2.6629569468748404e-05,
1191
+ "loss": 0.6089,
1192
+ "step": 1690
1193
+ },
1194
+ {
1195
+ "epoch": 0.2267498082630298,
1196
+ "grad_norm": 2.4198822683275147,
1197
+ "learning_rate": 2.658936454705251e-05,
1198
+ "loss": 0.6666,
1199
+ "step": 1700
1200
+ },
1201
+ {
1202
+ "epoch": 0.22808363066457701,
1203
+ "grad_norm": 2.4915285584652054,
1204
+ "learning_rate": 2.6548951982583246e-05,
1205
+ "loss": 0.7088,
1206
+ "step": 1710
1207
+ },
1208
+ {
1209
+ "epoch": 0.22941745306612424,
1210
+ "grad_norm": 2.2849831540010537,
1211
+ "learning_rate": 2.650833249939903e-05,
1212
+ "loss": 0.7149,
1213
+ "step": 1720
1214
+ },
1215
+ {
1216
+ "epoch": 0.2307512754676715,
1217
+ "grad_norm": 1.5098088938051029,
1218
+ "learning_rate": 2.6467506825265573e-05,
1219
+ "loss": 0.7254,
1220
+ "step": 1730
1221
+ },
1222
+ {
1223
+ "epoch": 0.23208509786921871,
1224
+ "grad_norm": 3.4800248296443814,
1225
+ "learning_rate": 2.642647569164284e-05,
1226
+ "loss": 0.6916,
1227
+ "step": 1740
1228
+ },
1229
+ {
1230
+ "epoch": 0.23341892027076594,
1231
+ "grad_norm": 7.281500947090542,
1232
+ "learning_rate": 2.638523983367194e-05,
1233
+ "loss": 0.6831,
1234
+ "step": 1750
1235
+ },
1236
+ {
1237
+ "epoch": 0.2347527426723132,
1238
+ "grad_norm": 3.0161864395495446,
1239
+ "learning_rate": 2.634379999016198e-05,
1240
+ "loss": 0.6999,
1241
+ "step": 1760
1242
+ },
1243
+ {
1244
+ "epoch": 0.23608656507386042,
1245
+ "grad_norm": 2.0917745352156762,
1246
+ "learning_rate": 2.6302156903576784e-05,
1247
+ "loss": 0.7112,
1248
+ "step": 1770
1249
+ },
1250
+ {
1251
+ "epoch": 0.23742038747540764,
1252
+ "grad_norm": 1.918811185774526,
1253
+ "learning_rate": 2.6260311320021628e-05,
1254
+ "loss": 0.6725,
1255
+ "step": 1780
1256
+ },
1257
+ {
1258
+ "epoch": 0.2387542098769549,
1259
+ "grad_norm": 3.0697413876733695,
1260
+ "learning_rate": 2.6218263989229855e-05,
1261
+ "loss": 0.7133,
1262
+ "step": 1790
1263
+ },
1264
+ {
1265
+ "epoch": 0.24008803227850212,
1266
+ "grad_norm": 6.14274393655379,
1267
+ "learning_rate": 2.617601566454944e-05,
1268
+ "loss": 0.6678,
1269
+ "step": 1800
1270
+ },
1271
+ {
1272
+ "epoch": 0.24142185468004934,
1273
+ "grad_norm": 4.259979200715344,
1274
+ "learning_rate": 2.613356710292951e-05,
1275
+ "loss": 0.7013,
1276
+ "step": 1810
1277
+ },
1278
+ {
1279
+ "epoch": 0.2427556770815966,
1280
+ "grad_norm": 3.1011058557692808,
1281
+ "learning_rate": 2.6090919064906766e-05,
1282
+ "loss": 0.7027,
1283
+ "step": 1820
1284
+ },
1285
+ {
1286
+ "epoch": 0.24408949948314382,
1287
+ "grad_norm": 3.677900978078831,
1288
+ "learning_rate": 2.6048072314591854e-05,
1289
+ "loss": 0.711,
1290
+ "step": 1830
1291
+ },
1292
+ {
1293
+ "epoch": 0.24542332188469104,
1294
+ "grad_norm": 2.368576699713982,
1295
+ "learning_rate": 2.600502761965569e-05,
1296
+ "loss": 0.6917,
1297
+ "step": 1840
1298
+ },
1299
+ {
1300
+ "epoch": 0.2467571442862383,
1301
+ "grad_norm": 3.0346306894457,
1302
+ "learning_rate": 2.59617857513157e-05,
1303
+ "loss": 0.69,
1304
+ "step": 1850
1305
+ },
1306
+ {
1307
+ "epoch": 0.24809096668778552,
1308
+ "grad_norm": 3.1228131080916204,
1309
+ "learning_rate": 2.591834748432198e-05,
1310
+ "loss": 0.695,
1311
+ "step": 1860
1312
+ },
1313
+ {
1314
+ "epoch": 0.24942478908933274,
1315
+ "grad_norm": 2.6886660685401034,
1316
+ "learning_rate": 2.5874713596943465e-05,
1317
+ "loss": 0.6681,
1318
+ "step": 1870
1319
+ },
1320
+ {
1321
+ "epoch": 0.25075861149087997,
1322
+ "grad_norm": 1.7244460999561722,
1323
+ "learning_rate": 2.5830884870953933e-05,
1324
+ "loss": 0.6737,
1325
+ "step": 1880
1326
+ },
1327
+ {
1328
+ "epoch": 0.25209243389242725,
1329
+ "grad_norm": 2.4283725332509842,
1330
+ "learning_rate": 2.578686209161803e-05,
1331
+ "loss": 0.6598,
1332
+ "step": 1890
1333
+ },
1334
+ {
1335
+ "epoch": 0.2534262562939745,
1336
+ "grad_norm": 5.496556851547161,
1337
+ "learning_rate": 2.5742646047677186e-05,
1338
+ "loss": 0.6931,
1339
+ "step": 1900
1340
+ },
1341
+ {
1342
+ "epoch": 0.2547600786955217,
1343
+ "grad_norm": 1.2751270156124934,
1344
+ "learning_rate": 2.5698237531335493e-05,
1345
+ "loss": 0.7043,
1346
+ "step": 1910
1347
+ },
1348
+ {
1349
+ "epoch": 0.2560939010970689,
1350
+ "grad_norm": 8.807017683974516,
1351
+ "learning_rate": 2.56536373382455e-05,
1352
+ "loss": 0.6234,
1353
+ "step": 1920
1354
+ },
1355
+ {
1356
+ "epoch": 0.25742772349861615,
1357
+ "grad_norm": 3.6331868296726277,
1358
+ "learning_rate": 2.5608846267493974e-05,
1359
+ "loss": 0.6763,
1360
+ "step": 1930
1361
+ },
1362
+ {
1363
+ "epoch": 0.25876154590016337,
1364
+ "grad_norm": 5.094905230807839,
1365
+ "learning_rate": 2.5563865121587563e-05,
1366
+ "loss": 0.6692,
1367
+ "step": 1940
1368
+ },
1369
+ {
1370
+ "epoch": 0.26009536830171065,
1371
+ "grad_norm": 2.0520732769663237,
1372
+ "learning_rate": 2.5518694706438445e-05,
1373
+ "loss": 0.7008,
1374
+ "step": 1950
1375
+ },
1376
+ {
1377
+ "epoch": 0.2614291907032579,
1378
+ "grad_norm": 2.1265138955486336,
1379
+ "learning_rate": 2.5473335831349842e-05,
1380
+ "loss": 0.6623,
1381
+ "step": 1960
1382
+ },
1383
+ {
1384
+ "epoch": 0.2627630131048051,
1385
+ "grad_norm": 4.532469697105077,
1386
+ "learning_rate": 2.5427789309001577e-05,
1387
+ "loss": 0.7099,
1388
+ "step": 1970
1389
+ },
1390
+ {
1391
+ "epoch": 0.2640968355063523,
1392
+ "grad_norm": 1.8912900905557881,
1393
+ "learning_rate": 2.538205595543548e-05,
1394
+ "loss": 0.712,
1395
+ "step": 1980
1396
+ },
1397
+ {
1398
+ "epoch": 0.26543065790789955,
1399
+ "grad_norm": 9.714825687307293,
1400
+ "learning_rate": 2.5336136590040767e-05,
1401
+ "loss": 0.6418,
1402
+ "step": 1990
1403
+ },
1404
+ {
1405
+ "epoch": 0.26676448030944677,
1406
+ "grad_norm": 4.375615975749738,
1407
+ "learning_rate": 2.529003203553937e-05,
1408
+ "loss": 0.6933,
1409
+ "step": 2000
1410
+ }
1411
+ ],
1412
+ "logging_steps": 10,
1413
+ "max_steps": 7497,
1414
+ "num_input_tokens_seen": 0,
1415
+ "num_train_epochs": 1,
1416
+ "save_steps": 400,
1417
+ "stateful_callbacks": {
1418
+ "TrainerControl": {
1419
+ "args": {
1420
+ "should_epoch_stop": false,
1421
+ "should_evaluate": false,
1422
+ "should_log": false,
1423
+ "should_save": true,
1424
+ "should_training_stop": false
1425
+ },
1426
+ "attributes": {}
1427
+ }
1428
+ },
1429
+ "total_flos": 5.467141180489728e+18,
1430
+ "train_batch_size": 4,
1431
+ "trial_name": null,
1432
+ "trial_params": null
1433
+ }
checkpoint-2000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81e448587a2aaa72701b15796f660da2d8aedd670c2a2eaae01d5488a2bf1cfe
3
+ size 6520
checkpoint-2000/zero_to_fp32.py ADDED
@@ -0,0 +1,587 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
215
+ elif zero_stage == 3:
216
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
217
+
218
+
219
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
220
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
221
+ return
222
+
223
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
224
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
225
+
226
+ if debug:
227
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
228
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
229
+
230
+ wanted_params = len(frozen_param_shapes)
231
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
232
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
233
+ print(f'Frozen params: Have {avail_numel} numels to process.')
234
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
235
+
236
+ total_params = 0
237
+ total_numel = 0
238
+ for name, shape in frozen_param_shapes.items():
239
+ total_params += 1
240
+ unpartitioned_numel = shape.numel()
241
+ total_numel += unpartitioned_numel
242
+
243
+ state_dict[name] = frozen_param_fragments[name]
244
+
245
+ if debug:
246
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
247
+
248
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
249
+
250
+
251
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
252
+ param_shapes = zero_model_states[0].param_shapes
253
+
254
+ # Reconstruction protocol:
255
+ #
256
+ # XXX: document this
257
+
258
+ if debug:
259
+ for i in range(world_size):
260
+ for j in range(len(fp32_flat_groups[0])):
261
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
262
+
263
+ # XXX: memory usage doubles here (zero2)
264
+ num_param_groups = len(fp32_flat_groups[0])
265
+ merged_single_partition_of_fp32_groups = []
266
+ for i in range(num_param_groups):
267
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
268
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
269
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
270
+ avail_numel = sum(
271
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
272
+
273
+ if debug:
274
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
275
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
276
+ # not asserting if there is a mismatch due to possible padding
277
+ print(f"Have {avail_numel} numels to process.")
278
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
279
+
280
+ # params
281
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
282
+ # out-of-core computing solution
283
+ total_numel = 0
284
+ total_params = 0
285
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
286
+ offset = 0
287
+ avail_numel = full_single_fp32_vector.numel()
288
+ for name, shape in shapes.items():
289
+
290
+ unpartitioned_numel = shape.numel()
291
+ total_numel += unpartitioned_numel
292
+ total_params += 1
293
+
294
+ if debug:
295
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
296
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
297
+ offset += unpartitioned_numel
298
+
299
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
300
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
301
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
302
+ # live optimizer object, so we are checking that the numbers are within the right range
303
+ align_to = 2 * world_size
304
+
305
+ def zero2_align(x):
306
+ return align_to * math.ceil(x / align_to)
307
+
308
+ if debug:
309
+ print(f"original offset={offset}, avail_numel={avail_numel}")
310
+
311
+ offset = zero2_align(offset)
312
+ avail_numel = zero2_align(avail_numel)
313
+
314
+ if debug:
315
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
316
+
317
+ # Sanity check
318
+ if offset != avail_numel:
319
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
320
+
321
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
322
+
323
+
324
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
325
+ state_dict = OrderedDict()
326
+
327
+ # buffers
328
+ buffers = zero_model_states[0].buffers
329
+ state_dict.update(buffers)
330
+ if debug:
331
+ print(f"added {len(buffers)} buffers")
332
+
333
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
334
+
335
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
336
+
337
+ # recover shared parameters
338
+ for pair in zero_model_states[0].shared_params:
339
+ if pair[1] in state_dict:
340
+ state_dict[pair[0]] = state_dict[pair[1]]
341
+
342
+ return state_dict
343
+
344
+
345
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
346
+ remainder = unpartitioned_numel % world_size
347
+ padding_numel = (world_size - remainder) if remainder else 0
348
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
349
+ return partitioned_numel, padding_numel
350
+
351
+
352
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
353
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
354
+ return
355
+
356
+ if debug:
357
+ for i in range(world_size):
358
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
359
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
360
+
361
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
362
+ wanted_params = len(frozen_param_shapes)
363
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
364
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
365
+ print(f'Frozen params: Have {avail_numel} numels to process.')
366
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
367
+
368
+ total_params = 0
369
+ total_numel = 0
370
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
371
+ total_params += 1
372
+ unpartitioned_numel = shape.numel()
373
+ total_numel += unpartitioned_numel
374
+
375
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
376
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
377
+
378
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
379
+
380
+ if debug:
381
+ print(
382
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
383
+ )
384
+
385
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
386
+
387
+
388
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
389
+ param_shapes = zero_model_states[0].param_shapes
390
+ avail_numel = fp32_flat_groups[0].numel() * world_size
391
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
392
+ # param, re-consolidating each param, while dealing with padding if any
393
+
394
+ # merge list of dicts, preserving order
395
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
396
+
397
+ if debug:
398
+ for i in range(world_size):
399
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
400
+
401
+ wanted_params = len(param_shapes)
402
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
403
+ # not asserting if there is a mismatch due to possible padding
404
+ avail_numel = fp32_flat_groups[0].numel() * world_size
405
+ print(f"Trainable params: Have {avail_numel} numels to process.")
406
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
407
+
408
+ # params
409
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
410
+ # out-of-core computing solution
411
+ offset = 0
412
+ total_numel = 0
413
+ total_params = 0
414
+ for name, shape in param_shapes.items():
415
+
416
+ unpartitioned_numel = shape.numel()
417
+ total_numel += unpartitioned_numel
418
+ total_params += 1
419
+
420
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
421
+
422
+ if debug:
423
+ print(
424
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
425
+ )
426
+
427
+ # XXX: memory usage doubles here
428
+ state_dict[name] = torch.cat(
429
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
430
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
431
+ offset += partitioned_numel
432
+
433
+ offset *= world_size
434
+
435
+ # Sanity check
436
+ if offset != avail_numel:
437
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
438
+
439
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
440
+
441
+
442
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
443
+ state_dict = OrderedDict()
444
+
445
+ # buffers
446
+ buffers = zero_model_states[0].buffers
447
+ state_dict.update(buffers)
448
+ if debug:
449
+ print(f"added {len(buffers)} buffers")
450
+
451
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
452
+
453
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
454
+
455
+ # recover shared parameters
456
+ for pair in zero_model_states[0].shared_params:
457
+ if pair[1] in state_dict:
458
+ state_dict[pair[0]] = state_dict[pair[1]]
459
+
460
+ return state_dict
461
+
462
+
463
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
464
+ """
465
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
466
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
467
+ via a model hub.
468
+
469
+ Args:
470
+ - ``checkpoint_dir``: path to the desired checkpoint folder
471
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
472
+
473
+ Returns:
474
+ - pytorch ``state_dict``
475
+
476
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
477
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
478
+ the checkpoint.
479
+
480
+ A typical usage might be ::
481
+
482
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
483
+ # do the training and checkpoint saving
484
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
485
+ model = model.cpu() # move to cpu
486
+ model.load_state_dict(state_dict)
487
+ # submit to model hub or save the model to share with others
488
+
489
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
490
+ application. i.e. you will need to re-initialize the deepspeed engine, since
491
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
492
+
493
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
494
+
495
+ """
496
+ if tag is None:
497
+ latest_path = os.path.join(checkpoint_dir, 'latest')
498
+ if os.path.isfile(latest_path):
499
+ with open(latest_path, 'r') as fd:
500
+ tag = fd.read().strip()
501
+ else:
502
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
503
+
504
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
505
+
506
+ if not os.path.isdir(ds_checkpoint_dir):
507
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
508
+
509
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
510
+
511
+
512
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
513
+ """
514
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
515
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
516
+
517
+ Args:
518
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
519
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
520
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
521
+ """
522
+
523
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
524
+ print(f"Saving fp32 state dict to {output_file}")
525
+ torch.save(state_dict, output_file)
526
+
527
+
528
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
529
+ """
530
+ 1. Put the provided model to cpu
531
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
532
+ 3. Load it into the provided model
533
+
534
+ Args:
535
+ - ``model``: the model object to update
536
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
537
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
538
+
539
+ Returns:
540
+ - ``model`: modified model
541
+
542
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
543
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
544
+ conveniently placed for you in the checkpoint folder.
545
+
546
+ A typical usage might be ::
547
+
548
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
549
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
550
+ # submit to model hub or save the model to share with others
551
+
552
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
553
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
554
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
555
+
556
+ """
557
+ logger.info(f"Extracting fp32 weights")
558
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
559
+
560
+ logger.info(f"Overwriting model with fp32 weights")
561
+ model = model.cpu()
562
+ model.load_state_dict(state_dict, strict=False)
563
+
564
+ return model
565
+
566
+
567
+ if __name__ == "__main__":
568
+
569
+ parser = argparse.ArgumentParser()
570
+ parser.add_argument("checkpoint_dir",
571
+ type=str,
572
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
573
+ parser.add_argument(
574
+ "output_file",
575
+ type=str,
576
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
577
+ parser.add_argument("-t",
578
+ "--tag",
579
+ type=str,
580
+ default=None,
581
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
582
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
583
+ args = parser.parse_args()
584
+
585
+ debug = args.debug
586
+
587
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)
checkpoint-2400/README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ base_model: Qwen/Qwen-VL-Chat
4
+ ---
5
+
6
+ # Model Card for Model ID
7
+
8
+ <!-- Provide a quick summary of what the model is/does. -->
9
+
10
+
11
+
12
+ ## Model Details
13
+
14
+ ### Model Description
15
+
16
+ <!-- Provide a longer summary of what this model is. -->
17
+
18
+
19
+
20
+ - **Developed by:** [More Information Needed]
21
+ - **Funded by [optional]:** [More Information Needed]
22
+ - **Shared by [optional]:** [More Information Needed]
23
+ - **Model type:** [More Information Needed]
24
+ - **Language(s) (NLP):** [More Information Needed]
25
+ - **License:** [More Information Needed]
26
+ - **Finetuned from model [optional]:** [More Information Needed]
27
+
28
+ ### Model Sources [optional]
29
+
30
+ <!-- Provide the basic links for the model. -->
31
+
32
+ - **Repository:** [More Information Needed]
33
+ - **Paper [optional]:** [More Information Needed]
34
+ - **Demo [optional]:** [More Information Needed]
35
+
36
+ ## Uses
37
+
38
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
+
40
+ ### Direct Use
41
+
42
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
+
44
+ [More Information Needed]
45
+
46
+ ### Downstream Use [optional]
47
+
48
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Out-of-Scope Use
53
+
54
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
+
56
+ [More Information Needed]
57
+
58
+ ## Bias, Risks, and Limitations
59
+
60
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ### Recommendations
65
+
66
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
+
68
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
+
70
+ ## How to Get Started with the Model
71
+
72
+ Use the code below to get started with the model.
73
+
74
+ [More Information Needed]
75
+
76
+ ## Training Details
77
+
78
+ ### Training Data
79
+
80
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
+
82
+ [More Information Needed]
83
+
84
+ ### Training Procedure
85
+
86
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
+
88
+ #### Preprocessing [optional]
89
+
90
+ [More Information Needed]
91
+
92
+
93
+ #### Training Hyperparameters
94
+
95
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
+
97
+ #### Speeds, Sizes, Times [optional]
98
+
99
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
+
101
+ [More Information Needed]
102
+
103
+ ## Evaluation
104
+
105
+ <!-- This section describes the evaluation protocols and provides the results. -->
106
+
107
+ ### Testing Data, Factors & Metrics
108
+
109
+ #### Testing Data
110
+
111
+ <!-- This should link to a Dataset Card if possible. -->
112
+
113
+ [More Information Needed]
114
+
115
+ #### Factors
116
+
117
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Metrics
122
+
123
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
+
125
+ [More Information Needed]
126
+
127
+ ### Results
128
+
129
+ [More Information Needed]
130
+
131
+ #### Summary
132
+
133
+
134
+
135
+ ## Model Examination [optional]
136
+
137
+ <!-- Relevant interpretability work for the model goes here -->
138
+
139
+ [More Information Needed]
140
+
141
+ ## Environmental Impact
142
+
143
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
+
145
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
+
147
+ - **Hardware Type:** [More Information Needed]
148
+ - **Hours used:** [More Information Needed]
149
+ - **Cloud Provider:** [More Information Needed]
150
+ - **Compute Region:** [More Information Needed]
151
+ - **Carbon Emitted:** [More Information Needed]
152
+
153
+ ## Technical Specifications [optional]
154
+
155
+ ### Model Architecture and Objective
156
+
157
+ [More Information Needed]
158
+
159
+ ### Compute Infrastructure
160
+
161
+ [More Information Needed]
162
+
163
+ #### Hardware
164
+
165
+ [More Information Needed]
166
+
167
+ #### Software
168
+
169
+ [More Information Needed]
170
+
171
+ ## Citation [optional]
172
+
173
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
+
175
+ **BibTeX:**
176
+
177
+ [More Information Needed]
178
+
179
+ **APA:**
180
+
181
+ [More Information Needed]
182
+
183
+ ## Glossary [optional]
184
+
185
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
+
187
+ [More Information Needed]
188
+
189
+ ## More Information [optional]
190
+
191
+ [More Information Needed]
192
+
193
+ ## Model Card Authors [optional]
194
+
195
+ [More Information Needed]
196
+
197
+ ## Model Card Contact
198
+
199
+ [More Information Needed]
200
+ ### Framework versions
201
+
202
+ - PEFT 0.10.0
checkpoint-2400/adapter_config.json ADDED
@@ -0,0 +1,380 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "Qwen/Qwen-VL-Chat",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layer_replication": null,
10
+ "layers_pattern": null,
11
+ "layers_to_transform": null,
12
+ "loftq_config": {},
13
+ "lora_alpha": 16,
14
+ "lora_dropout": 0.05,
15
+ "megatron_config": null,
16
+ "megatron_core": "megatron.core",
17
+ "modules_to_save": null,
18
+ "peft_type": "LORA",
19
+ "r": 64,
20
+ "rank_pattern": {},
21
+ "revision": null,
22
+ "target_modules": [
23
+ "transformer.visual.transformer.resblocks.32.mlp.c_fc",
24
+ "transformer.visual.transformer.resblocks.8.attn.in_proj",
25
+ "transformer.visual.transformer.resblocks.16.attn.in_proj",
26
+ "transformer.visual.transformer.resblocks.41.mlp.c_proj",
27
+ "transformer.visual.transformer.resblocks.31.mlp.c_proj",
28
+ "transformer.visual.transformer.resblocks.14.mlp.c_proj",
29
+ "transformer.h.14.mlp.w2",
30
+ "transformer.visual.transformer.resblocks.39.mlp.c_fc",
31
+ "transformer.visual.transformer.resblocks.40.mlp.c_proj",
32
+ "transformer.h.0.attn.c_attn",
33
+ "transformer.visual.conv1",
34
+ "transformer.visual.transformer.resblocks.26.mlp.c_fc",
35
+ "transformer.h.7.mlp.w2",
36
+ "transformer.visual.transformer.resblocks.21.mlp.c_fc",
37
+ "transformer.visual.transformer.resblocks.17.attn.in_proj",
38
+ "transformer.h.29.attn.c_attn",
39
+ "transformer.h.3.attn.c_proj",
40
+ "transformer.visual.transformer.resblocks.0.attn.out_proj",
41
+ "transformer.visual.transformer.resblocks.28.mlp.c_fc",
42
+ "transformer.h.30.attn.c_proj",
43
+ "transformer.h.3.mlp.w2",
44
+ "transformer.h.22.mlp.w1",
45
+ "transformer.visual.transformer.resblocks.11.mlp.c_proj",
46
+ "transformer.h.11.mlp.c_proj",
47
+ "transformer.visual.transformer.resblocks.3.attn.in_proj",
48
+ "transformer.visual.transformer.resblocks.21.attn.out_proj",
49
+ "transformer.visual.transformer.resblocks.33.attn.out_proj",
50
+ "transformer.visual.transformer.resblocks.3.mlp.c_fc",
51
+ "transformer.visual.transformer.resblocks.15.attn.in_proj",
52
+ "transformer.visual.transformer.resblocks.4.attn.out_proj",
53
+ "transformer.visual.transformer.resblocks.34.attn.in_proj",
54
+ "transformer.h.17.mlp.c_proj",
55
+ "transformer.visual.transformer.resblocks.1.mlp.c_proj",
56
+ "transformer.visual.transformer.resblocks.4.mlp.c_proj",
57
+ "transformer.visual.transformer.resblocks.39.attn.out_proj",
58
+ "transformer.h.13.mlp.c_proj",
59
+ "transformer.visual.transformer.resblocks.3.mlp.c_proj",
60
+ "transformer.visual.transformer.resblocks.22.attn.out_proj",
61
+ "transformer.h.27.attn.c_attn",
62
+ "transformer.visual.transformer.resblocks.47.attn.in_proj",
63
+ "transformer.h.1.mlp.c_proj",
64
+ "transformer.h.21.attn.c_attn",
65
+ "transformer.visual.transformer.resblocks.18.attn.out_proj",
66
+ "transformer.visual.transformer.resblocks.5.attn.in_proj",
67
+ "transformer.h.6.attn.c_proj",
68
+ "transformer.visual.transformer.resblocks.17.mlp.c_fc",
69
+ "transformer.h.16.attn.c_attn",
70
+ "transformer.visual.transformer.resblocks.0.attn.in_proj",
71
+ "transformer.visual.transformer.resblocks.22.mlp.c_fc",
72
+ "transformer.visual.transformer.resblocks.27.mlp.c_proj",
73
+ "transformer.h.11.attn.c_attn",
74
+ "transformer.h.22.mlp.w2",
75
+ "transformer.h.8.mlp.w1",
76
+ "transformer.visual.transformer.resblocks.29.mlp.c_proj",
77
+ "transformer.visual.transformer.resblocks.44.attn.out_proj",
78
+ "transformer.h.13.mlp.w2",
79
+ "transformer.visual.transformer.resblocks.19.mlp.c_proj",
80
+ "transformer.visual.transformer.resblocks.27.attn.in_proj",
81
+ "transformer.h.29.mlp.w1",
82
+ "transformer.h.24.mlp.c_proj",
83
+ "transformer.visual.transformer.resblocks.25.mlp.c_proj",
84
+ "transformer.visual.transformer.resblocks.4.attn.in_proj",
85
+ "transformer.visual.transformer.resblocks.42.mlp.c_fc",
86
+ "transformer.h.28.mlp.w2",
87
+ "transformer.visual.transformer.resblocks.20.attn.in_proj",
88
+ "transformer.visual.transformer.resblocks.9.attn.in_proj",
89
+ "transformer.h.10.attn.c_proj",
90
+ "transformer.h.13.attn.c_proj",
91
+ "transformer.visual.transformer.resblocks.26.attn.in_proj",
92
+ "transformer.h.17.attn.c_attn",
93
+ "transformer.visual.transformer.resblocks.7.attn.in_proj",
94
+ "transformer.h.23.attn.c_proj",
95
+ "transformer.visual.transformer.resblocks.38.attn.in_proj",
96
+ "transformer.h.19.attn.c_attn",
97
+ "transformer.h.1.attn.c_proj",
98
+ "transformer.visual.transformer.resblocks.2.attn.in_proj",
99
+ "transformer.h.4.mlp.w2",
100
+ "transformer.h.15.mlp.c_proj",
101
+ "transformer.h.4.mlp.c_proj",
102
+ "transformer.h.19.mlp.w2",
103
+ "transformer.h.12.mlp.w1",
104
+ "transformer.visual.transformer.resblocks.30.attn.in_proj",
105
+ "transformer.visual.transformer.resblocks.9.attn.out_proj",
106
+ "transformer.h.28.mlp.c_proj",
107
+ "transformer.h.1.attn.c_attn",
108
+ "transformer.h.8.attn.c_proj",
109
+ "transformer.visual.transformer.resblocks.40.mlp.c_fc",
110
+ "transformer.visual.transformer.resblocks.25.attn.out_proj",
111
+ "transformer.h.4.mlp.w1",
112
+ "transformer.visual.transformer.resblocks.31.mlp.c_fc",
113
+ "transformer.visual.transformer.resblocks.7.mlp.c_proj",
114
+ "transformer.visual.transformer.resblocks.33.attn.in_proj",
115
+ "transformer.h.0.mlp.w1",
116
+ "transformer.visual.transformer.resblocks.8.mlp.c_proj",
117
+ "transformer.h.20.mlp.w2",
118
+ "transformer.visual.transformer.resblocks.37.mlp.c_fc",
119
+ "transformer.visual.transformer.resblocks.43.attn.in_proj",
120
+ "transformer.visual.transformer.resblocks.37.attn.out_proj",
121
+ "transformer.h.25.mlp.w1",
122
+ "transformer.visual.transformer.resblocks.2.mlp.c_proj",
123
+ "transformer.h.27.mlp.w2",
124
+ "transformer.visual.transformer.resblocks.47.attn.out_proj",
125
+ "transformer.visual.transformer.resblocks.18.attn.in_proj",
126
+ "transformer.visual.transformer.resblocks.12.mlp.c_fc",
127
+ "transformer.visual.transformer.resblocks.11.attn.in_proj",
128
+ "transformer.h.14.mlp.c_proj",
129
+ "transformer.h.7.attn.c_attn",
130
+ "transformer.h.10.mlp.w2",
131
+ "transformer.h.11.mlp.w2",
132
+ "transformer.visual.transformer.resblocks.13.attn.out_proj",
133
+ "transformer.h.24.mlp.w1",
134
+ "transformer.h.0.mlp.c_proj",
135
+ "transformer.h.24.attn.c_proj",
136
+ "transformer.visual.transformer.resblocks.10.mlp.c_fc",
137
+ "transformer.visual.transformer.resblocks.33.mlp.c_proj",
138
+ "transformer.visual.transformer.resblocks.35.attn.in_proj",
139
+ "transformer.visual.transformer.resblocks.38.attn.out_proj",
140
+ "transformer.h.2.mlp.w2",
141
+ "transformer.visual.transformer.resblocks.40.attn.in_proj",
142
+ "transformer.h.25.mlp.w2",
143
+ "transformer.visual.transformer.resblocks.17.mlp.c_proj",
144
+ "transformer.visual.transformer.resblocks.18.mlp.c_proj",
145
+ "transformer.h.2.attn.c_proj",
146
+ "transformer.visual.transformer.resblocks.0.mlp.c_proj",
147
+ "transformer.visual.transformer.resblocks.23.mlp.c_proj",
148
+ "transformer.visual.transformer.resblocks.2.attn.out_proj",
149
+ "transformer.h.16.mlp.w2",
150
+ "transformer.h.29.mlp.c_proj",
151
+ "transformer.visual.transformer.resblocks.21.attn.in_proj",
152
+ "transformer.visual.transformer.resblocks.7.mlp.c_fc",
153
+ "transformer.visual.transformer.resblocks.8.attn.out_proj",
154
+ "transformer.visual.transformer.resblocks.28.attn.in_proj",
155
+ "transformer.h.11.mlp.w1",
156
+ "transformer.visual.transformer.resblocks.41.mlp.c_fc",
157
+ "transformer.h.18.attn.c_proj",
158
+ "transformer.visual.transformer.resblocks.29.mlp.c_fc",
159
+ "transformer.visual.transformer.resblocks.32.mlp.c_proj",
160
+ "transformer.visual.transformer.resblocks.28.attn.out_proj",
161
+ "transformer.h.6.attn.c_attn",
162
+ "transformer.visual.transformer.resblocks.3.attn.out_proj",
163
+ "transformer.h.25.attn.c_attn",
164
+ "transformer.h.28.attn.c_proj",
165
+ "transformer.visual.transformer.resblocks.16.attn.out_proj",
166
+ "transformer.visual.transformer.resblocks.20.mlp.c_proj",
167
+ "transformer.visual.transformer.resblocks.45.attn.in_proj",
168
+ "transformer.h.8.mlp.c_proj",
169
+ "transformer.visual.transformer.resblocks.29.attn.out_proj",
170
+ "transformer.h.27.mlp.c_proj",
171
+ "transformer.visual.transformer.resblocks.11.mlp.c_fc",
172
+ "transformer.visual.transformer.resblocks.11.attn.out_proj",
173
+ "transformer.visual.transformer.resblocks.10.attn.in_proj",
174
+ "transformer.visual.transformer.resblocks.27.mlp.c_fc",
175
+ "transformer.h.4.attn.c_proj",
176
+ "transformer.visual.transformer.resblocks.43.mlp.c_fc",
177
+ "transformer.visual.transformer.resblocks.22.mlp.c_proj",
178
+ "transformer.h.22.attn.c_proj",
179
+ "transformer.visual.transformer.resblocks.5.mlp.c_proj",
180
+ "transformer.h.22.mlp.c_proj",
181
+ "transformer.visual.transformer.resblocks.7.attn.out_proj",
182
+ "transformer.visual.transformer.resblocks.14.attn.out_proj",
183
+ "transformer.h.30.mlp.w1",
184
+ "transformer.h.14.attn.c_attn",
185
+ "transformer.h.4.attn.c_attn",
186
+ "transformer.visual.transformer.resblocks.46.mlp.c_fc",
187
+ "transformer.visual.transformer.resblocks.27.attn.out_proj",
188
+ "transformer.visual.transformer.resblocks.41.attn.in_proj",
189
+ "transformer.visual.transformer.resblocks.13.mlp.c_fc",
190
+ "transformer.h.12.attn.c_proj",
191
+ "transformer.visual.transformer.resblocks.32.attn.out_proj",
192
+ "transformer.visual.transformer.resblocks.18.mlp.c_fc",
193
+ "transformer.visual.transformer.resblocks.45.attn.out_proj",
194
+ "transformer.visual.transformer.resblocks.44.mlp.c_proj",
195
+ "transformer.visual.transformer.resblocks.35.mlp.c_proj",
196
+ "transformer.visual.transformer.resblocks.20.attn.out_proj",
197
+ "transformer.visual.transformer.resblocks.19.attn.in_proj",
198
+ "transformer.h.17.mlp.w2",
199
+ "transformer.h.10.attn.c_attn",
200
+ "transformer.visual.transformer.resblocks.45.mlp.c_proj",
201
+ "transformer.visual.transformer.resblocks.39.mlp.c_proj",
202
+ "transformer.h.7.mlp.w1",
203
+ "transformer.h.14.mlp.w1",
204
+ "transformer.visual.transformer.resblocks.2.mlp.c_fc",
205
+ "transformer.h.21.mlp.w1",
206
+ "transformer.visual.transformer.resblocks.30.attn.out_proj",
207
+ "transformer.h.19.mlp.c_proj",
208
+ "transformer.h.5.mlp.w1",
209
+ "transformer.visual.transformer.resblocks.47.mlp.c_fc",
210
+ "transformer.visual.transformer.resblocks.38.mlp.c_fc",
211
+ "transformer.h.10.mlp.c_proj",
212
+ "transformer.visual.transformer.resblocks.28.mlp.c_proj",
213
+ "transformer.visual.transformer.resblocks.25.attn.in_proj",
214
+ "transformer.h.9.attn.c_attn",
215
+ "transformer.visual.transformer.resblocks.0.mlp.c_fc",
216
+ "transformer.h.29.attn.c_proj",
217
+ "transformer.h.5.mlp.w2",
218
+ "transformer.h.30.attn.c_attn",
219
+ "transformer.h.1.mlp.w1",
220
+ "transformer.visual.transformer.resblocks.8.mlp.c_fc",
221
+ "transformer.h.19.mlp.w1",
222
+ "transformer.h.18.attn.c_attn",
223
+ "transformer.h.11.attn.c_proj",
224
+ "transformer.h.5.mlp.c_proj",
225
+ "transformer.visual.transformer.resblocks.45.mlp.c_fc",
226
+ "transformer.visual.transformer.resblocks.47.mlp.c_proj",
227
+ "transformer.h.9.attn.c_proj",
228
+ "transformer.visual.transformer.resblocks.6.attn.out_proj",
229
+ "transformer.h.26.mlp.c_proj",
230
+ "transformer.visual.transformer.resblocks.24.mlp.c_proj",
231
+ "transformer.h.31.attn.c_attn",
232
+ "transformer.h.13.mlp.w1",
233
+ "transformer.visual.transformer.resblocks.39.attn.in_proj",
234
+ "transformer.h.20.mlp.w1",
235
+ "transformer.visual.transformer.resblocks.22.attn.in_proj",
236
+ "transformer.visual.transformer.resblocks.35.mlp.c_fc",
237
+ "transformer.visual.transformer.resblocks.24.attn.in_proj",
238
+ "transformer.h.16.mlp.w1",
239
+ "transformer.visual.transformer.resblocks.6.mlp.c_proj",
240
+ "transformer.h.6.mlp.c_proj",
241
+ "transformer.visual.transformer.resblocks.26.mlp.c_proj",
242
+ "transformer.visual.transformer.resblocks.35.attn.out_proj",
243
+ "transformer.h.24.mlp.w2",
244
+ "transformer.visual.transformer.resblocks.31.attn.in_proj",
245
+ "transformer.visual.transformer.resblocks.15.mlp.c_fc",
246
+ "transformer.h.2.mlp.w1",
247
+ "transformer.h.31.mlp.c_proj",
248
+ "transformer.h.13.attn.c_attn",
249
+ "transformer.visual.transformer.resblocks.14.attn.in_proj",
250
+ "transformer.h.12.mlp.w2",
251
+ "transformer.visual.transformer.resblocks.34.mlp.c_proj",
252
+ "transformer.h.26.mlp.w2",
253
+ "transformer.visual.transformer.resblocks.5.mlp.c_fc",
254
+ "transformer.h.5.attn.c_proj",
255
+ "transformer.h.9.mlp.w2",
256
+ "transformer.h.15.mlp.w2",
257
+ "transformer.h.12.attn.c_attn",
258
+ "transformer.visual.transformer.resblocks.10.mlp.c_proj",
259
+ "transformer.h.28.mlp.w1",
260
+ "transformer.visual.transformer.resblocks.46.attn.out_proj",
261
+ "transformer.h.18.mlp.c_proj",
262
+ "transformer.h.15.attn.c_proj",
263
+ "transformer.visual.transformer.resblocks.44.attn.in_proj",
264
+ "transformer.visual.transformer.resblocks.12.attn.out_proj",
265
+ "transformer.visual.transformer.resblocks.1.mlp.c_fc",
266
+ "transformer.visual.transformer.resblocks.15.attn.out_proj",
267
+ "transformer.h.17.mlp.w1",
268
+ "transformer.visual.transformer.resblocks.46.attn.in_proj",
269
+ "transformer.h.2.attn.c_attn",
270
+ "transformer.h.25.attn.c_proj",
271
+ "transformer.h.14.attn.c_proj",
272
+ "transformer.h.26.attn.c_proj",
273
+ "transformer.h.31.mlp.w1",
274
+ "transformer.h.23.mlp.w2",
275
+ "transformer.visual.transformer.resblocks.26.attn.out_proj",
276
+ "transformer.h.20.mlp.c_proj",
277
+ "transformer.visual.transformer.resblocks.6.mlp.c_fc",
278
+ "transformer.h.27.mlp.w1",
279
+ "transformer.h.7.attn.c_proj",
280
+ "transformer.visual.transformer.resblocks.15.mlp.c_proj",
281
+ "transformer.h.16.mlp.c_proj",
282
+ "transformer.visual.transformer.resblocks.12.mlp.c_proj",
283
+ "transformer.h.29.mlp.w2",
284
+ "transformer.h.15.mlp.w1",
285
+ "transformer.h.6.mlp.w2",
286
+ "transformer.h.3.attn.c_attn",
287
+ "transformer.h.21.mlp.w2",
288
+ "transformer.visual.transformer.resblocks.16.mlp.c_fc",
289
+ "transformer.visual.transformer.resblocks.23.mlp.c_fc",
290
+ "transformer.visual.transformer.resblocks.34.attn.out_proj",
291
+ "transformer.h.8.attn.c_attn",
292
+ "transformer.visual.transformer.resblocks.43.mlp.c_proj",
293
+ "transformer.visual.transformer.resblocks.14.mlp.c_fc",
294
+ "transformer.visual.transformer.resblocks.24.attn.out_proj",
295
+ "transformer.visual.transformer.resblocks.12.attn.in_proj",
296
+ "transformer.h.25.mlp.c_proj",
297
+ "transformer.visual.transformer.resblocks.46.mlp.c_proj",
298
+ "transformer.h.7.mlp.c_proj",
299
+ "transformer.h.15.attn.c_attn",
300
+ "transformer.visual.transformer.resblocks.36.attn.in_proj",
301
+ "transformer.h.26.attn.c_attn",
302
+ "transformer.h.0.attn.c_proj",
303
+ "transformer.visual.transformer.resblocks.9.mlp.c_fc",
304
+ "transformer.h.19.attn.c_proj",
305
+ "transformer.visual.transformer.resblocks.29.attn.in_proj",
306
+ "transformer.visual.transformer.resblocks.13.attn.in_proj",
307
+ "transformer.visual.transformer.resblocks.23.attn.in_proj",
308
+ "transformer.visual.transformer.resblocks.36.attn.out_proj",
309
+ "transformer.h.3.mlp.c_proj",
310
+ "transformer.h.27.attn.c_proj",
311
+ "transformer.h.31.attn.c_proj",
312
+ "transformer.visual.transformer.resblocks.19.attn.out_proj",
313
+ "transformer.h.0.mlp.w2",
314
+ "transformer.h.17.attn.c_proj",
315
+ "transformer.h.30.mlp.w2",
316
+ "transformer.visual.transformer.resblocks.24.mlp.c_fc",
317
+ "transformer.h.28.attn.c_attn",
318
+ "transformer.visual.transformer.resblocks.16.mlp.c_proj",
319
+ "transformer.visual.transformer.resblocks.36.mlp.c_fc",
320
+ "transformer.h.30.mlp.c_proj",
321
+ "transformer.visual.transformer.resblocks.37.attn.in_proj",
322
+ "transformer.visual.transformer.resblocks.20.mlp.c_fc",
323
+ "transformer.h.9.mlp.c_proj",
324
+ "transformer.visual.transformer.resblocks.43.attn.out_proj",
325
+ "transformer.visual.transformer.resblocks.42.attn.in_proj",
326
+ "transformer.h.1.mlp.w2",
327
+ "transformer.h.6.mlp.w1",
328
+ "transformer.visual.transformer.resblocks.17.attn.out_proj",
329
+ "transformer.visual.transformer.resblocks.33.mlp.c_fc",
330
+ "transformer.h.5.attn.c_attn",
331
+ "transformer.h.8.mlp.w2",
332
+ "transformer.h.23.mlp.c_proj",
333
+ "transformer.h.20.attn.c_attn",
334
+ "transformer.visual.transformer.resblocks.4.mlp.c_fc",
335
+ "transformer.h.31.mlp.w2",
336
+ "transformer.h.9.mlp.w1",
337
+ "transformer.h.12.mlp.c_proj",
338
+ "transformer.visual.transformer.resblocks.1.attn.out_proj",
339
+ "transformer.visual.transformer.resblocks.30.mlp.c_fc",
340
+ "transformer.visual.transformer.resblocks.1.attn.in_proj",
341
+ "transformer.visual.transformer.resblocks.32.attn.in_proj",
342
+ "transformer.h.16.attn.c_proj",
343
+ "transformer.visual.transformer.resblocks.42.mlp.c_proj",
344
+ "transformer.h.3.mlp.w1",
345
+ "transformer.visual.transformer.resblocks.9.mlp.c_proj",
346
+ "transformer.h.18.mlp.w1",
347
+ "transformer.visual.transformer.resblocks.44.mlp.c_fc",
348
+ "transformer.h.21.mlp.c_proj",
349
+ "transformer.visual.transformer.resblocks.5.attn.out_proj",
350
+ "transformer.visual.transformer.resblocks.34.mlp.c_fc",
351
+ "transformer.visual.transformer.resblocks.42.attn.out_proj",
352
+ "transformer.visual.transformer.resblocks.41.attn.out_proj",
353
+ "transformer.h.10.mlp.w1",
354
+ "transformer.visual.transformer.resblocks.19.mlp.c_fc",
355
+ "transformer.visual.transformer.resblocks.13.mlp.c_proj",
356
+ "transformer.h.21.attn.c_proj",
357
+ "transformer.visual.transformer.resblocks.6.attn.in_proj",
358
+ "transformer.visual.transformer.resblocks.23.attn.out_proj",
359
+ "transformer.visual.transformer.resblocks.40.attn.out_proj",
360
+ "transformer.visual.transformer.resblocks.36.mlp.c_proj",
361
+ "transformer.h.2.mlp.c_proj",
362
+ "transformer.visual.transformer.resblocks.25.mlp.c_fc",
363
+ "transformer.h.22.attn.c_attn",
364
+ "transformer.h.23.mlp.w1",
365
+ "transformer.h.20.attn.c_proj",
366
+ "transformer.h.23.attn.c_attn",
367
+ "transformer.visual.transformer.resblocks.37.mlp.c_proj",
368
+ "transformer.h.26.mlp.w1",
369
+ "transformer.h.18.mlp.w2",
370
+ "transformer.visual.transformer.resblocks.21.mlp.c_proj",
371
+ "transformer.visual.transformer.resblocks.31.attn.out_proj",
372
+ "transformer.h.24.attn.c_attn",
373
+ "transformer.visual.transformer.resblocks.10.attn.out_proj",
374
+ "transformer.visual.transformer.resblocks.30.mlp.c_proj",
375
+ "transformer.visual.transformer.resblocks.38.mlp.c_proj"
376
+ ],
377
+ "task_type": "CAUSAL_LM",
378
+ "use_dora": false,
379
+ "use_rslora": false
380
+ }
checkpoint-2400/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d40094c4c9fa083f14d64ac21a5ff6f46bc59cf112308f9b3afc06a3c3ae11ec
3
+ size 469105640
checkpoint-2400/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step2400
checkpoint-2400/qwen.tiktoken ADDED
The diff for this file is too large to render. See raw diff