Commit
•
7139a11
1
Parent(s):
0f751d3
Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/config.json +47 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/generation_config.json +7 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/model.safetensors +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/optimizer.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/rng_state.pth +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/scheduler.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/special_tokens_map.json +30 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/tokenizer.json +0 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/tokenizer_config.json +235 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/trainer_state.json +495 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/training_args.bin +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/config.json +47 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/generation_config.json +7 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/model.safetensors +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/optimizer.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/rng_state.pth +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/scheduler.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/special_tokens_map.json +30 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/tokenizer.json +0 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/tokenizer_config.json +235 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/trainer_state.json +733 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/training_args.bin +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/config.json +47 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/generation_config.json +7 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/model.safetensors +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/optimizer.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/rng_state.pth +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/scheduler.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/special_tokens_map.json +30 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/tokenizer.json +0 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/tokenizer_config.json +235 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/trainer_state.json +964 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/training_args.bin +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/config.json +47 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/generation_config.json +7 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/model.safetensors +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/optimizer.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/rng_state.pth +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/scheduler.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/special_tokens_map.json +30 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/tokenizer.json +0 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/tokenizer_config.json +235 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/trainer_state.json +1195 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/training_args.bin +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/config.json +47 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/generation_config.json +7 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/model.safetensors +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/optimizer.pt +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/rng_state.pth +3 -0
- data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/scheduler.pt +3 -0
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "jonathanjordan21/mos-mamba-6x130m-trainer-sft",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MoSMambaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "jonathanjordan21/mos-mamba-6x130m-hf--configuration_mos_mamba.MoSMambaConfig",
|
| 8 |
+
"AutoModelForCausalLM": "jonathanjordan21/mos-mamba-6x130m-hf--modeling_mos_mamba.MoSMambaForCausalLM"
|
| 9 |
+
},
|
| 10 |
+
"bos_token_id": 0,
|
| 11 |
+
"conv_kernel": 4,
|
| 12 |
+
"d_inner": 1536,
|
| 13 |
+
"d_model": 768,
|
| 14 |
+
"eos_token_id": 50278,
|
| 15 |
+
"expand": 2,
|
| 16 |
+
"fused_add_norm": true,
|
| 17 |
+
"hidden_act": "silu",
|
| 18 |
+
"hidden_size": 768,
|
| 19 |
+
"initializer_range": 0.1,
|
| 20 |
+
"intermediate_size": 1536,
|
| 21 |
+
"layer_norm_epsilon": 1e-05,
|
| 22 |
+
"model_type": "MoSMamba",
|
| 23 |
+
"n_layer": 24,
|
| 24 |
+
"num_hidden_layers": 24,
|
| 25 |
+
"num_selectivities": 6,
|
| 26 |
+
"num_selectivities_per_tok": 2,
|
| 27 |
+
"output_router_logits": true,
|
| 28 |
+
"pad_token_id": 0,
|
| 29 |
+
"pad_vocab_size_multiple": 8,
|
| 30 |
+
"rescale_prenorm_residual": false,
|
| 31 |
+
"residual_in_fp32": true,
|
| 32 |
+
"rms_norm": true,
|
| 33 |
+
"ssm_cfg": {},
|
| 34 |
+
"state_size": 16,
|
| 35 |
+
"time_step_floor": 0.0001,
|
| 36 |
+
"time_step_init_scheme": "random",
|
| 37 |
+
"time_step_max": 0.1,
|
| 38 |
+
"time_step_min": 0.001,
|
| 39 |
+
"time_step_rank": 48,
|
| 40 |
+
"time_step_scale": 1.0,
|
| 41 |
+
"torch_dtype": "float32",
|
| 42 |
+
"transformers_version": "4.42.4",
|
| 43 |
+
"use_bias": false,
|
| 44 |
+
"use_cache": true,
|
| 45 |
+
"use_conv_bias": true,
|
| 46 |
+
"vocab_size": 50280
|
| 47 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 50278,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.42.4"
|
| 7 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6937142212651a678a33c4e1f2719cf3be1031c8f8b3c111fd55ba7f6b9ad214
|
| 3 |
+
size 576008736
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01dfeaff61c12b37c87ce89cb53985959d2e7263c19133140cb531854ced789c
|
| 3 |
+
size 1152256984
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b947f2bfa134bed4999be08a91af4e579cf020d254a594fd4caa119ddb94028b
|
| 3 |
+
size 14244
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08901cbd97b64e6a5392b2f8bc40dd6a03473274fe44442064f83101235e7ffc
|
| 3 |
+
size 1064
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/special_tokens_map.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<|endoftext|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "<|im_end|>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": false,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": {
|
| 17 |
+
"content": "<|endoftext|>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": false,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": false
|
| 22 |
+
},
|
| 23 |
+
"unk_token": {
|
| 24 |
+
"content": "<|endoftext|>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": false,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": false
|
| 29 |
+
}
|
| 30 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/tokenizer_config.json
ADDED
|
@@ -0,0 +1,235 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"add_prefix_space": false,
|
| 5 |
+
"added_tokens_decoder": {
|
| 6 |
+
"0": {
|
| 7 |
+
"content": "<|endoftext|>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false,
|
| 12 |
+
"special": true
|
| 13 |
+
},
|
| 14 |
+
"1": {
|
| 15 |
+
"content": "<|padding|>",
|
| 16 |
+
"lstrip": false,
|
| 17 |
+
"normalized": false,
|
| 18 |
+
"rstrip": false,
|
| 19 |
+
"single_word": false,
|
| 20 |
+
"special": true
|
| 21 |
+
},
|
| 22 |
+
"50254": {
|
| 23 |
+
"content": " ",
|
| 24 |
+
"lstrip": false,
|
| 25 |
+
"normalized": true,
|
| 26 |
+
"rstrip": false,
|
| 27 |
+
"single_word": false,
|
| 28 |
+
"special": false
|
| 29 |
+
},
|
| 30 |
+
"50255": {
|
| 31 |
+
"content": " ",
|
| 32 |
+
"lstrip": false,
|
| 33 |
+
"normalized": true,
|
| 34 |
+
"rstrip": false,
|
| 35 |
+
"single_word": false,
|
| 36 |
+
"special": false
|
| 37 |
+
},
|
| 38 |
+
"50256": {
|
| 39 |
+
"content": " ",
|
| 40 |
+
"lstrip": false,
|
| 41 |
+
"normalized": true,
|
| 42 |
+
"rstrip": false,
|
| 43 |
+
"single_word": false,
|
| 44 |
+
"special": false
|
| 45 |
+
},
|
| 46 |
+
"50257": {
|
| 47 |
+
"content": " ",
|
| 48 |
+
"lstrip": false,
|
| 49 |
+
"normalized": true,
|
| 50 |
+
"rstrip": false,
|
| 51 |
+
"single_word": false,
|
| 52 |
+
"special": false
|
| 53 |
+
},
|
| 54 |
+
"50258": {
|
| 55 |
+
"content": " ",
|
| 56 |
+
"lstrip": false,
|
| 57 |
+
"normalized": true,
|
| 58 |
+
"rstrip": false,
|
| 59 |
+
"single_word": false,
|
| 60 |
+
"special": false
|
| 61 |
+
},
|
| 62 |
+
"50259": {
|
| 63 |
+
"content": " ",
|
| 64 |
+
"lstrip": false,
|
| 65 |
+
"normalized": true,
|
| 66 |
+
"rstrip": false,
|
| 67 |
+
"single_word": false,
|
| 68 |
+
"special": false
|
| 69 |
+
},
|
| 70 |
+
"50260": {
|
| 71 |
+
"content": " ",
|
| 72 |
+
"lstrip": false,
|
| 73 |
+
"normalized": true,
|
| 74 |
+
"rstrip": false,
|
| 75 |
+
"single_word": false,
|
| 76 |
+
"special": false
|
| 77 |
+
},
|
| 78 |
+
"50261": {
|
| 79 |
+
"content": " ",
|
| 80 |
+
"lstrip": false,
|
| 81 |
+
"normalized": true,
|
| 82 |
+
"rstrip": false,
|
| 83 |
+
"single_word": false,
|
| 84 |
+
"special": false
|
| 85 |
+
},
|
| 86 |
+
"50262": {
|
| 87 |
+
"content": " ",
|
| 88 |
+
"lstrip": false,
|
| 89 |
+
"normalized": true,
|
| 90 |
+
"rstrip": false,
|
| 91 |
+
"single_word": false,
|
| 92 |
+
"special": false
|
| 93 |
+
},
|
| 94 |
+
"50263": {
|
| 95 |
+
"content": " ",
|
| 96 |
+
"lstrip": false,
|
| 97 |
+
"normalized": true,
|
| 98 |
+
"rstrip": false,
|
| 99 |
+
"single_word": false,
|
| 100 |
+
"special": false
|
| 101 |
+
},
|
| 102 |
+
"50264": {
|
| 103 |
+
"content": " ",
|
| 104 |
+
"lstrip": false,
|
| 105 |
+
"normalized": true,
|
| 106 |
+
"rstrip": false,
|
| 107 |
+
"single_word": false,
|
| 108 |
+
"special": false
|
| 109 |
+
},
|
| 110 |
+
"50265": {
|
| 111 |
+
"content": " ",
|
| 112 |
+
"lstrip": false,
|
| 113 |
+
"normalized": true,
|
| 114 |
+
"rstrip": false,
|
| 115 |
+
"single_word": false,
|
| 116 |
+
"special": false
|
| 117 |
+
},
|
| 118 |
+
"50266": {
|
| 119 |
+
"content": " ",
|
| 120 |
+
"lstrip": false,
|
| 121 |
+
"normalized": true,
|
| 122 |
+
"rstrip": false,
|
| 123 |
+
"single_word": false,
|
| 124 |
+
"special": false
|
| 125 |
+
},
|
| 126 |
+
"50267": {
|
| 127 |
+
"content": " ",
|
| 128 |
+
"lstrip": false,
|
| 129 |
+
"normalized": true,
|
| 130 |
+
"rstrip": false,
|
| 131 |
+
"single_word": false,
|
| 132 |
+
"special": false
|
| 133 |
+
},
|
| 134 |
+
"50268": {
|
| 135 |
+
"content": " ",
|
| 136 |
+
"lstrip": false,
|
| 137 |
+
"normalized": true,
|
| 138 |
+
"rstrip": false,
|
| 139 |
+
"single_word": false,
|
| 140 |
+
"special": false
|
| 141 |
+
},
|
| 142 |
+
"50269": {
|
| 143 |
+
"content": " ",
|
| 144 |
+
"lstrip": false,
|
| 145 |
+
"normalized": true,
|
| 146 |
+
"rstrip": false,
|
| 147 |
+
"single_word": false,
|
| 148 |
+
"special": false
|
| 149 |
+
},
|
| 150 |
+
"50270": {
|
| 151 |
+
"content": " ",
|
| 152 |
+
"lstrip": false,
|
| 153 |
+
"normalized": true,
|
| 154 |
+
"rstrip": false,
|
| 155 |
+
"single_word": false,
|
| 156 |
+
"special": false
|
| 157 |
+
},
|
| 158 |
+
"50271": {
|
| 159 |
+
"content": " ",
|
| 160 |
+
"lstrip": false,
|
| 161 |
+
"normalized": true,
|
| 162 |
+
"rstrip": false,
|
| 163 |
+
"single_word": false,
|
| 164 |
+
"special": false
|
| 165 |
+
},
|
| 166 |
+
"50272": {
|
| 167 |
+
"content": " ",
|
| 168 |
+
"lstrip": false,
|
| 169 |
+
"normalized": true,
|
| 170 |
+
"rstrip": false,
|
| 171 |
+
"single_word": false,
|
| 172 |
+
"special": false
|
| 173 |
+
},
|
| 174 |
+
"50273": {
|
| 175 |
+
"content": " ",
|
| 176 |
+
"lstrip": false,
|
| 177 |
+
"normalized": true,
|
| 178 |
+
"rstrip": false,
|
| 179 |
+
"single_word": false,
|
| 180 |
+
"special": false
|
| 181 |
+
},
|
| 182 |
+
"50274": {
|
| 183 |
+
"content": " ",
|
| 184 |
+
"lstrip": false,
|
| 185 |
+
"normalized": true,
|
| 186 |
+
"rstrip": false,
|
| 187 |
+
"single_word": false,
|
| 188 |
+
"special": false
|
| 189 |
+
},
|
| 190 |
+
"50275": {
|
| 191 |
+
"content": " ",
|
| 192 |
+
"lstrip": false,
|
| 193 |
+
"normalized": true,
|
| 194 |
+
"rstrip": false,
|
| 195 |
+
"single_word": false,
|
| 196 |
+
"special": false
|
| 197 |
+
},
|
| 198 |
+
"50276": {
|
| 199 |
+
"content": " ",
|
| 200 |
+
"lstrip": false,
|
| 201 |
+
"normalized": true,
|
| 202 |
+
"rstrip": false,
|
| 203 |
+
"single_word": false,
|
| 204 |
+
"special": false
|
| 205 |
+
},
|
| 206 |
+
"50277": {
|
| 207 |
+
"content": "<|im_start|>",
|
| 208 |
+
"lstrip": false,
|
| 209 |
+
"normalized": false,
|
| 210 |
+
"rstrip": false,
|
| 211 |
+
"single_word": false,
|
| 212 |
+
"special": true
|
| 213 |
+
},
|
| 214 |
+
"50278": {
|
| 215 |
+
"content": "<|im_end|>",
|
| 216 |
+
"lstrip": false,
|
| 217 |
+
"normalized": false,
|
| 218 |
+
"rstrip": false,
|
| 219 |
+
"single_word": false,
|
| 220 |
+
"special": true
|
| 221 |
+
}
|
| 222 |
+
},
|
| 223 |
+
"bos_token": "<|endoftext|>",
|
| 224 |
+
"chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
|
| 225 |
+
"clean_up_tokenization_spaces": true,
|
| 226 |
+
"eos_token": "<|im_end|>",
|
| 227 |
+
"max_length": 1024,
|
| 228 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 229 |
+
"pad_token": "<|endoftext|>",
|
| 230 |
+
"stride": 0,
|
| 231 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
| 232 |
+
"truncation_side": "right",
|
| 233 |
+
"truncation_strategy": "longest_first",
|
| 234 |
+
"unk_token": "<|endoftext|>"
|
| 235 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/trainer_state.json
ADDED
|
@@ -0,0 +1,495 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.0019473514073508622,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 1000,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 2.9210271110262932e-05,
|
| 13 |
+
"grad_norm": 5.397985935211182,
|
| 14 |
+
"learning_rate": 2.9999999936841446e-05,
|
| 15 |
+
"loss": 2.1156,
|
| 16 |
+
"step": 15
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 5.8420542220525864e-05,
|
| 20 |
+
"grad_norm": 3.999190092086792,
|
| 21 |
+
"learning_rate": 2.999999974736578e-05,
|
| 22 |
+
"loss": 2.5268,
|
| 23 |
+
"step": 30
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 8.76308133307888e-05,
|
| 27 |
+
"grad_norm": 5.580333709716797,
|
| 28 |
+
"learning_rate": 2.999999943157301e-05,
|
| 29 |
+
"loss": 2.3955,
|
| 30 |
+
"step": 45
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.00011684108444105173,
|
| 34 |
+
"grad_norm": 3.6752560138702393,
|
| 35 |
+
"learning_rate": 2.9999998989463132e-05,
|
| 36 |
+
"loss": 2.171,
|
| 37 |
+
"step": 60
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.00014605135555131467,
|
| 41 |
+
"grad_norm": 2.655461072921753,
|
| 42 |
+
"learning_rate": 2.9999998421036153e-05,
|
| 43 |
+
"loss": 2.188,
|
| 44 |
+
"step": 75
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.0001752616266615776,
|
| 48 |
+
"grad_norm": 2.572371006011963,
|
| 49 |
+
"learning_rate": 2.9999997726292083e-05,
|
| 50 |
+
"loss": 2.1977,
|
| 51 |
+
"step": 90
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.00020447189777184053,
|
| 55 |
+
"grad_norm": 8.544811248779297,
|
| 56 |
+
"learning_rate": 2.999999690523092e-05,
|
| 57 |
+
"loss": 2.1871,
|
| 58 |
+
"step": 105
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.00023368216888210345,
|
| 62 |
+
"grad_norm": 3.530529737472534,
|
| 63 |
+
"learning_rate": 2.999999595785267e-05,
|
| 64 |
+
"loss": 2.2238,
|
| 65 |
+
"step": 120
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.00026289243999236636,
|
| 69 |
+
"grad_norm": 3.671182155609131,
|
| 70 |
+
"learning_rate": 2.9999994884157345e-05,
|
| 71 |
+
"loss": 2.1338,
|
| 72 |
+
"step": 135
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.00029210271110262934,
|
| 76 |
+
"grad_norm": 3.9643137454986572,
|
| 77 |
+
"learning_rate": 2.9999993684144956e-05,
|
| 78 |
+
"loss": 2.0788,
|
| 79 |
+
"step": 150
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.00032131298221289227,
|
| 83 |
+
"grad_norm": 2.62164044380188,
|
| 84 |
+
"learning_rate": 2.9999992357815508e-05,
|
| 85 |
+
"loss": 2.117,
|
| 86 |
+
"step": 165
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.0003505232533231552,
|
| 90 |
+
"grad_norm": 4.189517021179199,
|
| 91 |
+
"learning_rate": 2.999999090516902e-05,
|
| 92 |
+
"loss": 2.262,
|
| 93 |
+
"step": 180
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.0003797335244334181,
|
| 97 |
+
"grad_norm": 2.880622625350952,
|
| 98 |
+
"learning_rate": 2.9999989326205494e-05,
|
| 99 |
+
"loss": 2.1102,
|
| 100 |
+
"step": 195
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.00040894379554368105,
|
| 104 |
+
"grad_norm": 5.852839469909668,
|
| 105 |
+
"learning_rate": 2.9999987620924948e-05,
|
| 106 |
+
"loss": 2.0684,
|
| 107 |
+
"step": 210
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.000438154066653944,
|
| 111 |
+
"grad_norm": 3.6339924335479736,
|
| 112 |
+
"learning_rate": 2.9999985789327394e-05,
|
| 113 |
+
"loss": 2.2518,
|
| 114 |
+
"step": 225
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.0004673643377642069,
|
| 118 |
+
"grad_norm": 2.743267297744751,
|
| 119 |
+
"learning_rate": 2.9999983831412858e-05,
|
| 120 |
+
"loss": 2.1236,
|
| 121 |
+
"step": 240
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.0004965746088744698,
|
| 125 |
+
"grad_norm": 3.9490363597869873,
|
| 126 |
+
"learning_rate": 2.9999981747181345e-05,
|
| 127 |
+
"loss": 2.0878,
|
| 128 |
+
"step": 255
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.0005257848799847327,
|
| 132 |
+
"grad_norm": 3.0342047214508057,
|
| 133 |
+
"learning_rate": 2.9999979536632872e-05,
|
| 134 |
+
"loss": 2.1415,
|
| 135 |
+
"step": 270
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.0005549951510949957,
|
| 139 |
+
"grad_norm": 4.045855522155762,
|
| 140 |
+
"learning_rate": 2.9999977199767467e-05,
|
| 141 |
+
"loss": 2.0558,
|
| 142 |
+
"step": 285
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"epoch": 0.0005842054222052587,
|
| 146 |
+
"grad_norm": 4.325068950653076,
|
| 147 |
+
"learning_rate": 2.999997473658514e-05,
|
| 148 |
+
"loss": 2.0617,
|
| 149 |
+
"step": 300
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.0006134156933155216,
|
| 153 |
+
"grad_norm": 4.770871162414551,
|
| 154 |
+
"learning_rate": 2.999997214708592e-05,
|
| 155 |
+
"loss": 2.0922,
|
| 156 |
+
"step": 315
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 0.0006426259644257845,
|
| 160 |
+
"grad_norm": 4.1802754402160645,
|
| 161 |
+
"learning_rate": 2.999996943126982e-05,
|
| 162 |
+
"loss": 2.217,
|
| 163 |
+
"step": 330
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.0006718362355360475,
|
| 167 |
+
"grad_norm": 3.044260263442993,
|
| 168 |
+
"learning_rate": 2.999996658913687e-05,
|
| 169 |
+
"loss": 2.0508,
|
| 170 |
+
"step": 345
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.0007010465066463104,
|
| 174 |
+
"grad_norm": 4.1525187492370605,
|
| 175 |
+
"learning_rate": 2.9999963620687095e-05,
|
| 176 |
+
"loss": 2.1678,
|
| 177 |
+
"step": 360
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.0007302567777565733,
|
| 181 |
+
"grad_norm": 2.5676231384277344,
|
| 182 |
+
"learning_rate": 2.9999960525920515e-05,
|
| 183 |
+
"loss": 2.1693,
|
| 184 |
+
"step": 375
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 0.0007594670488668362,
|
| 188 |
+
"grad_norm": 3.880307674407959,
|
| 189 |
+
"learning_rate": 2.9999957304837156e-05,
|
| 190 |
+
"loss": 1.9008,
|
| 191 |
+
"step": 390
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.0007886773199770992,
|
| 195 |
+
"grad_norm": 3.3433754444122314,
|
| 196 |
+
"learning_rate": 2.999995395743705e-05,
|
| 197 |
+
"loss": 2.0734,
|
| 198 |
+
"step": 405
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.0008178875910873621,
|
| 202 |
+
"grad_norm": 3.451378583908081,
|
| 203 |
+
"learning_rate": 2.999995048372022e-05,
|
| 204 |
+
"loss": 1.9901,
|
| 205 |
+
"step": 420
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.000847097862197625,
|
| 209 |
+
"grad_norm": 3.4411580562591553,
|
| 210 |
+
"learning_rate": 2.9999946883686695e-05,
|
| 211 |
+
"loss": 1.9734,
|
| 212 |
+
"step": 435
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.000876308133307888,
|
| 216 |
+
"grad_norm": 3.514651298522949,
|
| 217 |
+
"learning_rate": 2.999994315733651e-05,
|
| 218 |
+
"loss": 2.1632,
|
| 219 |
+
"step": 450
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.0009055184044181509,
|
| 223 |
+
"grad_norm": 5.380824565887451,
|
| 224 |
+
"learning_rate": 2.999993930466969e-05,
|
| 225 |
+
"loss": 2.1534,
|
| 226 |
+
"step": 465
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.0009347286755284138,
|
| 230 |
+
"grad_norm": 3.768618106842041,
|
| 231 |
+
"learning_rate": 2.999993532568628e-05,
|
| 232 |
+
"loss": 2.2182,
|
| 233 |
+
"step": 480
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.0009639389466386767,
|
| 237 |
+
"grad_norm": 2.9490509033203125,
|
| 238 |
+
"learning_rate": 2.99999312203863e-05,
|
| 239 |
+
"loss": 1.9677,
|
| 240 |
+
"step": 495
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 0.0009931492177489396,
|
| 244 |
+
"grad_norm": 3.4881184101104736,
|
| 245 |
+
"learning_rate": 2.9999926988769788e-05,
|
| 246 |
+
"loss": 2.0786,
|
| 247 |
+
"step": 510
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 0.0010223594888592025,
|
| 251 |
+
"grad_norm": 3.788675546646118,
|
| 252 |
+
"learning_rate": 2.9999922630836784e-05,
|
| 253 |
+
"loss": 1.9696,
|
| 254 |
+
"step": 525
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 0.0010515697599694654,
|
| 258 |
+
"grad_norm": 3.070878744125366,
|
| 259 |
+
"learning_rate": 2.999991814658732e-05,
|
| 260 |
+
"loss": 1.9053,
|
| 261 |
+
"step": 540
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 0.0010807800310797286,
|
| 265 |
+
"grad_norm": 2.456150531768799,
|
| 266 |
+
"learning_rate": 2.9999913536021436e-05,
|
| 267 |
+
"loss": 1.9813,
|
| 268 |
+
"step": 555
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 0.0011099903021899915,
|
| 272 |
+
"grad_norm": 4.81670618057251,
|
| 273 |
+
"learning_rate": 2.999990879913917e-05,
|
| 274 |
+
"loss": 1.9317,
|
| 275 |
+
"step": 570
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 0.0011392005733002544,
|
| 279 |
+
"grad_norm": 4.014492988586426,
|
| 280 |
+
"learning_rate": 2.9999903935940567e-05,
|
| 281 |
+
"loss": 1.9709,
|
| 282 |
+
"step": 585
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.0011684108444105174,
|
| 286 |
+
"grad_norm": 4.765859603881836,
|
| 287 |
+
"learning_rate": 2.999989894642566e-05,
|
| 288 |
+
"loss": 1.9252,
|
| 289 |
+
"step": 600
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 0.0011976211155207803,
|
| 293 |
+
"grad_norm": 3.5599727630615234,
|
| 294 |
+
"learning_rate": 2.9999893830594492e-05,
|
| 295 |
+
"loss": 2.0175,
|
| 296 |
+
"step": 615
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 0.0012268313866310432,
|
| 300 |
+
"grad_norm": 4.441170692443848,
|
| 301 |
+
"learning_rate": 2.999988858844711e-05,
|
| 302 |
+
"loss": 2.1166,
|
| 303 |
+
"step": 630
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"epoch": 0.0012560416577413061,
|
| 307 |
+
"grad_norm": 2.691765308380127,
|
| 308 |
+
"learning_rate": 2.9999883219983558e-05,
|
| 309 |
+
"loss": 1.991,
|
| 310 |
+
"step": 645
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.001285251928851569,
|
| 314 |
+
"grad_norm": 4.039743423461914,
|
| 315 |
+
"learning_rate": 2.9999877725203878e-05,
|
| 316 |
+
"loss": 1.9917,
|
| 317 |
+
"step": 660
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 0.001314462199961832,
|
| 321 |
+
"grad_norm": 3.0612881183624268,
|
| 322 |
+
"learning_rate": 2.9999872104108115e-05,
|
| 323 |
+
"loss": 2.0436,
|
| 324 |
+
"step": 675
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.001343672471072095,
|
| 328 |
+
"grad_norm": 3.3579518795013428,
|
| 329 |
+
"learning_rate": 2.9999866356696326e-05,
|
| 330 |
+
"loss": 1.9659,
|
| 331 |
+
"step": 690
|
| 332 |
+
},
|
| 333 |
+
{
|
| 334 |
+
"epoch": 0.0013728827421823579,
|
| 335 |
+
"grad_norm": 4.145946502685547,
|
| 336 |
+
"learning_rate": 2.9999860482968552e-05,
|
| 337 |
+
"loss": 2.0267,
|
| 338 |
+
"step": 705
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 0.0014020930132926208,
|
| 342 |
+
"grad_norm": 5.213028430938721,
|
| 343 |
+
"learning_rate": 2.999985448292484e-05,
|
| 344 |
+
"loss": 2.0953,
|
| 345 |
+
"step": 720
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.0014313032844028837,
|
| 349 |
+
"grad_norm": 2.440793752670288,
|
| 350 |
+
"learning_rate": 2.9999848356565246e-05,
|
| 351 |
+
"loss": 2.0218,
|
| 352 |
+
"step": 735
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.0014605135555131466,
|
| 356 |
+
"grad_norm": 2.6167280673980713,
|
| 357 |
+
"learning_rate": 2.9999842103889813e-05,
|
| 358 |
+
"loss": 2.0638,
|
| 359 |
+
"step": 750
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 0.0014897238266234096,
|
| 363 |
+
"grad_norm": 2.2680811882019043,
|
| 364 |
+
"learning_rate": 2.999983572489861e-05,
|
| 365 |
+
"loss": 1.973,
|
| 366 |
+
"step": 765
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.0015189340977336725,
|
| 370 |
+
"grad_norm": 2.905947208404541,
|
| 371 |
+
"learning_rate": 2.999982921959167e-05,
|
| 372 |
+
"loss": 1.9346,
|
| 373 |
+
"step": 780
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.0015481443688439354,
|
| 377 |
+
"grad_norm": 4.673079967498779,
|
| 378 |
+
"learning_rate": 2.999982258796907e-05,
|
| 379 |
+
"loss": 2.1666,
|
| 380 |
+
"step": 795
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 0.0015773546399541984,
|
| 384 |
+
"grad_norm": 2.5963408946990967,
|
| 385 |
+
"learning_rate": 2.9999815830030846e-05,
|
| 386 |
+
"loss": 2.1063,
|
| 387 |
+
"step": 810
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 0.0016065649110644613,
|
| 391 |
+
"grad_norm": 3.656632661819458,
|
| 392 |
+
"learning_rate": 2.9999808945777066e-05,
|
| 393 |
+
"loss": 2.1328,
|
| 394 |
+
"step": 825
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 0.0016357751821747242,
|
| 398 |
+
"grad_norm": 4.4109368324279785,
|
| 399 |
+
"learning_rate": 2.9999801935207786e-05,
|
| 400 |
+
"loss": 1.9602,
|
| 401 |
+
"step": 840
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.0016649854532849871,
|
| 405 |
+
"grad_norm": 2.10768985748291,
|
| 406 |
+
"learning_rate": 2.9999794798323065e-05,
|
| 407 |
+
"loss": 1.9766,
|
| 408 |
+
"step": 855
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"epoch": 0.00169419572439525,
|
| 412 |
+
"grad_norm": 4.301672458648682,
|
| 413 |
+
"learning_rate": 2.999978753512296e-05,
|
| 414 |
+
"loss": 1.9386,
|
| 415 |
+
"step": 870
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 0.001723405995505513,
|
| 419 |
+
"grad_norm": 8.171302795410156,
|
| 420 |
+
"learning_rate": 2.9999780145607538e-05,
|
| 421 |
+
"loss": 2.1288,
|
| 422 |
+
"step": 885
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 0.001752616266615776,
|
| 426 |
+
"grad_norm": 3.4784069061279297,
|
| 427 |
+
"learning_rate": 2.999977262977685e-05,
|
| 428 |
+
"loss": 2.1346,
|
| 429 |
+
"step": 900
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.0017818265377260389,
|
| 433 |
+
"grad_norm": 2.9945054054260254,
|
| 434 |
+
"learning_rate": 2.9999764987630976e-05,
|
| 435 |
+
"loss": 2.0432,
|
| 436 |
+
"step": 915
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"epoch": 0.0018110368088363018,
|
| 440 |
+
"grad_norm": 3.632039785385132,
|
| 441 |
+
"learning_rate": 2.9999757219169964e-05,
|
| 442 |
+
"loss": 2.0642,
|
| 443 |
+
"step": 930
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.0018402470799465647,
|
| 447 |
+
"grad_norm": 3.9847047328948975,
|
| 448 |
+
"learning_rate": 2.999974932439389e-05,
|
| 449 |
+
"loss": 2.016,
|
| 450 |
+
"step": 945
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.0018694573510568276,
|
| 454 |
+
"grad_norm": 1.89451003074646,
|
| 455 |
+
"learning_rate": 2.9999741303302816e-05,
|
| 456 |
+
"loss": 2.0355,
|
| 457 |
+
"step": 960
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.0018986676221670906,
|
| 461 |
+
"grad_norm": 6.945189952850342,
|
| 462 |
+
"learning_rate": 2.9999733155896814e-05,
|
| 463 |
+
"loss": 1.97,
|
| 464 |
+
"step": 975
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.0019278778932773535,
|
| 468 |
+
"grad_norm": 3.347123384475708,
|
| 469 |
+
"learning_rate": 2.9999724882175947e-05,
|
| 470 |
+
"loss": 2.0421,
|
| 471 |
+
"step": 990
|
| 472 |
+
}
|
| 473 |
+
],
|
| 474 |
+
"logging_steps": 15,
|
| 475 |
+
"max_steps": 513518,
|
| 476 |
+
"num_input_tokens_seen": 0,
|
| 477 |
+
"num_train_epochs": 1,
|
| 478 |
+
"save_steps": 500,
|
| 479 |
+
"stateful_callbacks": {
|
| 480 |
+
"TrainerControl": {
|
| 481 |
+
"args": {
|
| 482 |
+
"should_epoch_stop": false,
|
| 483 |
+
"should_evaluate": false,
|
| 484 |
+
"should_log": false,
|
| 485 |
+
"should_save": true,
|
| 486 |
+
"should_training_stop": false
|
| 487 |
+
},
|
| 488 |
+
"attributes": {}
|
| 489 |
+
}
|
| 490 |
+
},
|
| 491 |
+
"total_flos": 1579021509666816.0,
|
| 492 |
+
"train_batch_size": 4,
|
| 493 |
+
"trial_name": null,
|
| 494 |
+
"trial_params": null
|
| 495 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1000/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0287fbd96bea9652b04a1bb7ed4b6b1e8de92cccbc4b448c2b5d263a08664d48
|
| 3 |
+
size 5304
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "jonathanjordan21/mos-mamba-6x130m-trainer-sft",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MoSMambaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "jonathanjordan21/mos-mamba-6x130m-hf--configuration_mos_mamba.MoSMambaConfig",
|
| 8 |
+
"AutoModelForCausalLM": "jonathanjordan21/mos-mamba-6x130m-hf--modeling_mos_mamba.MoSMambaForCausalLM"
|
| 9 |
+
},
|
| 10 |
+
"bos_token_id": 0,
|
| 11 |
+
"conv_kernel": 4,
|
| 12 |
+
"d_inner": 1536,
|
| 13 |
+
"d_model": 768,
|
| 14 |
+
"eos_token_id": 50278,
|
| 15 |
+
"expand": 2,
|
| 16 |
+
"fused_add_norm": true,
|
| 17 |
+
"hidden_act": "silu",
|
| 18 |
+
"hidden_size": 768,
|
| 19 |
+
"initializer_range": 0.1,
|
| 20 |
+
"intermediate_size": 1536,
|
| 21 |
+
"layer_norm_epsilon": 1e-05,
|
| 22 |
+
"model_type": "MoSMamba",
|
| 23 |
+
"n_layer": 24,
|
| 24 |
+
"num_hidden_layers": 24,
|
| 25 |
+
"num_selectivities": 6,
|
| 26 |
+
"num_selectivities_per_tok": 2,
|
| 27 |
+
"output_router_logits": true,
|
| 28 |
+
"pad_token_id": 0,
|
| 29 |
+
"pad_vocab_size_multiple": 8,
|
| 30 |
+
"rescale_prenorm_residual": false,
|
| 31 |
+
"residual_in_fp32": true,
|
| 32 |
+
"rms_norm": true,
|
| 33 |
+
"ssm_cfg": {},
|
| 34 |
+
"state_size": 16,
|
| 35 |
+
"time_step_floor": 0.0001,
|
| 36 |
+
"time_step_init_scheme": "random",
|
| 37 |
+
"time_step_max": 0.1,
|
| 38 |
+
"time_step_min": 0.001,
|
| 39 |
+
"time_step_rank": 48,
|
| 40 |
+
"time_step_scale": 1.0,
|
| 41 |
+
"torch_dtype": "float32",
|
| 42 |
+
"transformers_version": "4.42.4",
|
| 43 |
+
"use_bias": false,
|
| 44 |
+
"use_cache": true,
|
| 45 |
+
"use_conv_bias": true,
|
| 46 |
+
"vocab_size": 50280
|
| 47 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 50278,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.42.4"
|
| 7 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f80b6d472e59ae7fec8088664ce242349a60d9bb131ecf6726c8f55974154a30
|
| 3 |
+
size 576008736
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8f682bdfee7b01afaf9fbbfc05bc61155368f01558bf2b4d7cd1ba746f1a2c86
|
| 3 |
+
size 1152256984
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f78dd4f19f673b42f5cf8774841c12deb3b6cb36f4dd01a4434396355d5759d5
|
| 3 |
+
size 14244
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:09d7d7feb9068a099d497fccb58fb0d86eeebc6da532371e14c643d3f077639f
|
| 3 |
+
size 1064
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/special_tokens_map.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<|endoftext|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "<|im_end|>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": false,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": {
|
| 17 |
+
"content": "<|endoftext|>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": false,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": false
|
| 22 |
+
},
|
| 23 |
+
"unk_token": {
|
| 24 |
+
"content": "<|endoftext|>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": false,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": false
|
| 29 |
+
}
|
| 30 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/tokenizer_config.json
ADDED
|
@@ -0,0 +1,235 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"add_prefix_space": false,
|
| 5 |
+
"added_tokens_decoder": {
|
| 6 |
+
"0": {
|
| 7 |
+
"content": "<|endoftext|>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false,
|
| 12 |
+
"special": true
|
| 13 |
+
},
|
| 14 |
+
"1": {
|
| 15 |
+
"content": "<|padding|>",
|
| 16 |
+
"lstrip": false,
|
| 17 |
+
"normalized": false,
|
| 18 |
+
"rstrip": false,
|
| 19 |
+
"single_word": false,
|
| 20 |
+
"special": true
|
| 21 |
+
},
|
| 22 |
+
"50254": {
|
| 23 |
+
"content": " ",
|
| 24 |
+
"lstrip": false,
|
| 25 |
+
"normalized": true,
|
| 26 |
+
"rstrip": false,
|
| 27 |
+
"single_word": false,
|
| 28 |
+
"special": false
|
| 29 |
+
},
|
| 30 |
+
"50255": {
|
| 31 |
+
"content": " ",
|
| 32 |
+
"lstrip": false,
|
| 33 |
+
"normalized": true,
|
| 34 |
+
"rstrip": false,
|
| 35 |
+
"single_word": false,
|
| 36 |
+
"special": false
|
| 37 |
+
},
|
| 38 |
+
"50256": {
|
| 39 |
+
"content": " ",
|
| 40 |
+
"lstrip": false,
|
| 41 |
+
"normalized": true,
|
| 42 |
+
"rstrip": false,
|
| 43 |
+
"single_word": false,
|
| 44 |
+
"special": false
|
| 45 |
+
},
|
| 46 |
+
"50257": {
|
| 47 |
+
"content": " ",
|
| 48 |
+
"lstrip": false,
|
| 49 |
+
"normalized": true,
|
| 50 |
+
"rstrip": false,
|
| 51 |
+
"single_word": false,
|
| 52 |
+
"special": false
|
| 53 |
+
},
|
| 54 |
+
"50258": {
|
| 55 |
+
"content": " ",
|
| 56 |
+
"lstrip": false,
|
| 57 |
+
"normalized": true,
|
| 58 |
+
"rstrip": false,
|
| 59 |
+
"single_word": false,
|
| 60 |
+
"special": false
|
| 61 |
+
},
|
| 62 |
+
"50259": {
|
| 63 |
+
"content": " ",
|
| 64 |
+
"lstrip": false,
|
| 65 |
+
"normalized": true,
|
| 66 |
+
"rstrip": false,
|
| 67 |
+
"single_word": false,
|
| 68 |
+
"special": false
|
| 69 |
+
},
|
| 70 |
+
"50260": {
|
| 71 |
+
"content": " ",
|
| 72 |
+
"lstrip": false,
|
| 73 |
+
"normalized": true,
|
| 74 |
+
"rstrip": false,
|
| 75 |
+
"single_word": false,
|
| 76 |
+
"special": false
|
| 77 |
+
},
|
| 78 |
+
"50261": {
|
| 79 |
+
"content": " ",
|
| 80 |
+
"lstrip": false,
|
| 81 |
+
"normalized": true,
|
| 82 |
+
"rstrip": false,
|
| 83 |
+
"single_word": false,
|
| 84 |
+
"special": false
|
| 85 |
+
},
|
| 86 |
+
"50262": {
|
| 87 |
+
"content": " ",
|
| 88 |
+
"lstrip": false,
|
| 89 |
+
"normalized": true,
|
| 90 |
+
"rstrip": false,
|
| 91 |
+
"single_word": false,
|
| 92 |
+
"special": false
|
| 93 |
+
},
|
| 94 |
+
"50263": {
|
| 95 |
+
"content": " ",
|
| 96 |
+
"lstrip": false,
|
| 97 |
+
"normalized": true,
|
| 98 |
+
"rstrip": false,
|
| 99 |
+
"single_word": false,
|
| 100 |
+
"special": false
|
| 101 |
+
},
|
| 102 |
+
"50264": {
|
| 103 |
+
"content": " ",
|
| 104 |
+
"lstrip": false,
|
| 105 |
+
"normalized": true,
|
| 106 |
+
"rstrip": false,
|
| 107 |
+
"single_word": false,
|
| 108 |
+
"special": false
|
| 109 |
+
},
|
| 110 |
+
"50265": {
|
| 111 |
+
"content": " ",
|
| 112 |
+
"lstrip": false,
|
| 113 |
+
"normalized": true,
|
| 114 |
+
"rstrip": false,
|
| 115 |
+
"single_word": false,
|
| 116 |
+
"special": false
|
| 117 |
+
},
|
| 118 |
+
"50266": {
|
| 119 |
+
"content": " ",
|
| 120 |
+
"lstrip": false,
|
| 121 |
+
"normalized": true,
|
| 122 |
+
"rstrip": false,
|
| 123 |
+
"single_word": false,
|
| 124 |
+
"special": false
|
| 125 |
+
},
|
| 126 |
+
"50267": {
|
| 127 |
+
"content": " ",
|
| 128 |
+
"lstrip": false,
|
| 129 |
+
"normalized": true,
|
| 130 |
+
"rstrip": false,
|
| 131 |
+
"single_word": false,
|
| 132 |
+
"special": false
|
| 133 |
+
},
|
| 134 |
+
"50268": {
|
| 135 |
+
"content": " ",
|
| 136 |
+
"lstrip": false,
|
| 137 |
+
"normalized": true,
|
| 138 |
+
"rstrip": false,
|
| 139 |
+
"single_word": false,
|
| 140 |
+
"special": false
|
| 141 |
+
},
|
| 142 |
+
"50269": {
|
| 143 |
+
"content": " ",
|
| 144 |
+
"lstrip": false,
|
| 145 |
+
"normalized": true,
|
| 146 |
+
"rstrip": false,
|
| 147 |
+
"single_word": false,
|
| 148 |
+
"special": false
|
| 149 |
+
},
|
| 150 |
+
"50270": {
|
| 151 |
+
"content": " ",
|
| 152 |
+
"lstrip": false,
|
| 153 |
+
"normalized": true,
|
| 154 |
+
"rstrip": false,
|
| 155 |
+
"single_word": false,
|
| 156 |
+
"special": false
|
| 157 |
+
},
|
| 158 |
+
"50271": {
|
| 159 |
+
"content": " ",
|
| 160 |
+
"lstrip": false,
|
| 161 |
+
"normalized": true,
|
| 162 |
+
"rstrip": false,
|
| 163 |
+
"single_word": false,
|
| 164 |
+
"special": false
|
| 165 |
+
},
|
| 166 |
+
"50272": {
|
| 167 |
+
"content": " ",
|
| 168 |
+
"lstrip": false,
|
| 169 |
+
"normalized": true,
|
| 170 |
+
"rstrip": false,
|
| 171 |
+
"single_word": false,
|
| 172 |
+
"special": false
|
| 173 |
+
},
|
| 174 |
+
"50273": {
|
| 175 |
+
"content": " ",
|
| 176 |
+
"lstrip": false,
|
| 177 |
+
"normalized": true,
|
| 178 |
+
"rstrip": false,
|
| 179 |
+
"single_word": false,
|
| 180 |
+
"special": false
|
| 181 |
+
},
|
| 182 |
+
"50274": {
|
| 183 |
+
"content": " ",
|
| 184 |
+
"lstrip": false,
|
| 185 |
+
"normalized": true,
|
| 186 |
+
"rstrip": false,
|
| 187 |
+
"single_word": false,
|
| 188 |
+
"special": false
|
| 189 |
+
},
|
| 190 |
+
"50275": {
|
| 191 |
+
"content": " ",
|
| 192 |
+
"lstrip": false,
|
| 193 |
+
"normalized": true,
|
| 194 |
+
"rstrip": false,
|
| 195 |
+
"single_word": false,
|
| 196 |
+
"special": false
|
| 197 |
+
},
|
| 198 |
+
"50276": {
|
| 199 |
+
"content": " ",
|
| 200 |
+
"lstrip": false,
|
| 201 |
+
"normalized": true,
|
| 202 |
+
"rstrip": false,
|
| 203 |
+
"single_word": false,
|
| 204 |
+
"special": false
|
| 205 |
+
},
|
| 206 |
+
"50277": {
|
| 207 |
+
"content": "<|im_start|>",
|
| 208 |
+
"lstrip": false,
|
| 209 |
+
"normalized": false,
|
| 210 |
+
"rstrip": false,
|
| 211 |
+
"single_word": false,
|
| 212 |
+
"special": true
|
| 213 |
+
},
|
| 214 |
+
"50278": {
|
| 215 |
+
"content": "<|im_end|>",
|
| 216 |
+
"lstrip": false,
|
| 217 |
+
"normalized": false,
|
| 218 |
+
"rstrip": false,
|
| 219 |
+
"single_word": false,
|
| 220 |
+
"special": true
|
| 221 |
+
}
|
| 222 |
+
},
|
| 223 |
+
"bos_token": "<|endoftext|>",
|
| 224 |
+
"chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
|
| 225 |
+
"clean_up_tokenization_spaces": true,
|
| 226 |
+
"eos_token": "<|im_end|>",
|
| 227 |
+
"max_length": 1024,
|
| 228 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 229 |
+
"pad_token": "<|endoftext|>",
|
| 230 |
+
"stride": 0,
|
| 231 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
| 232 |
+
"truncation_side": "right",
|
| 233 |
+
"truncation_strategy": "longest_first",
|
| 234 |
+
"unk_token": "<|endoftext|>"
|
| 235 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/trainer_state.json
ADDED
|
@@ -0,0 +1,733 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.0029210271110262933,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 1500,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 2.9210271110262932e-05,
|
| 13 |
+
"grad_norm": 5.397985935211182,
|
| 14 |
+
"learning_rate": 2.9999999936841446e-05,
|
| 15 |
+
"loss": 2.1156,
|
| 16 |
+
"step": 15
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 5.8420542220525864e-05,
|
| 20 |
+
"grad_norm": 3.999190092086792,
|
| 21 |
+
"learning_rate": 2.999999974736578e-05,
|
| 22 |
+
"loss": 2.5268,
|
| 23 |
+
"step": 30
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 8.76308133307888e-05,
|
| 27 |
+
"grad_norm": 5.580333709716797,
|
| 28 |
+
"learning_rate": 2.999999943157301e-05,
|
| 29 |
+
"loss": 2.3955,
|
| 30 |
+
"step": 45
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.00011684108444105173,
|
| 34 |
+
"grad_norm": 3.6752560138702393,
|
| 35 |
+
"learning_rate": 2.9999998989463132e-05,
|
| 36 |
+
"loss": 2.171,
|
| 37 |
+
"step": 60
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.00014605135555131467,
|
| 41 |
+
"grad_norm": 2.655461072921753,
|
| 42 |
+
"learning_rate": 2.9999998421036153e-05,
|
| 43 |
+
"loss": 2.188,
|
| 44 |
+
"step": 75
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.0001752616266615776,
|
| 48 |
+
"grad_norm": 2.572371006011963,
|
| 49 |
+
"learning_rate": 2.9999997726292083e-05,
|
| 50 |
+
"loss": 2.1977,
|
| 51 |
+
"step": 90
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.00020447189777184053,
|
| 55 |
+
"grad_norm": 8.544811248779297,
|
| 56 |
+
"learning_rate": 2.999999690523092e-05,
|
| 57 |
+
"loss": 2.1871,
|
| 58 |
+
"step": 105
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.00023368216888210345,
|
| 62 |
+
"grad_norm": 3.530529737472534,
|
| 63 |
+
"learning_rate": 2.999999595785267e-05,
|
| 64 |
+
"loss": 2.2238,
|
| 65 |
+
"step": 120
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.00026289243999236636,
|
| 69 |
+
"grad_norm": 3.671182155609131,
|
| 70 |
+
"learning_rate": 2.9999994884157345e-05,
|
| 71 |
+
"loss": 2.1338,
|
| 72 |
+
"step": 135
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.00029210271110262934,
|
| 76 |
+
"grad_norm": 3.9643137454986572,
|
| 77 |
+
"learning_rate": 2.9999993684144956e-05,
|
| 78 |
+
"loss": 2.0788,
|
| 79 |
+
"step": 150
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.00032131298221289227,
|
| 83 |
+
"grad_norm": 2.62164044380188,
|
| 84 |
+
"learning_rate": 2.9999992357815508e-05,
|
| 85 |
+
"loss": 2.117,
|
| 86 |
+
"step": 165
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.0003505232533231552,
|
| 90 |
+
"grad_norm": 4.189517021179199,
|
| 91 |
+
"learning_rate": 2.999999090516902e-05,
|
| 92 |
+
"loss": 2.262,
|
| 93 |
+
"step": 180
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.0003797335244334181,
|
| 97 |
+
"grad_norm": 2.880622625350952,
|
| 98 |
+
"learning_rate": 2.9999989326205494e-05,
|
| 99 |
+
"loss": 2.1102,
|
| 100 |
+
"step": 195
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.00040894379554368105,
|
| 104 |
+
"grad_norm": 5.852839469909668,
|
| 105 |
+
"learning_rate": 2.9999987620924948e-05,
|
| 106 |
+
"loss": 2.0684,
|
| 107 |
+
"step": 210
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.000438154066653944,
|
| 111 |
+
"grad_norm": 3.6339924335479736,
|
| 112 |
+
"learning_rate": 2.9999985789327394e-05,
|
| 113 |
+
"loss": 2.2518,
|
| 114 |
+
"step": 225
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.0004673643377642069,
|
| 118 |
+
"grad_norm": 2.743267297744751,
|
| 119 |
+
"learning_rate": 2.9999983831412858e-05,
|
| 120 |
+
"loss": 2.1236,
|
| 121 |
+
"step": 240
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.0004965746088744698,
|
| 125 |
+
"grad_norm": 3.9490363597869873,
|
| 126 |
+
"learning_rate": 2.9999981747181345e-05,
|
| 127 |
+
"loss": 2.0878,
|
| 128 |
+
"step": 255
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.0005257848799847327,
|
| 132 |
+
"grad_norm": 3.0342047214508057,
|
| 133 |
+
"learning_rate": 2.9999979536632872e-05,
|
| 134 |
+
"loss": 2.1415,
|
| 135 |
+
"step": 270
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.0005549951510949957,
|
| 139 |
+
"grad_norm": 4.045855522155762,
|
| 140 |
+
"learning_rate": 2.9999977199767467e-05,
|
| 141 |
+
"loss": 2.0558,
|
| 142 |
+
"step": 285
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"epoch": 0.0005842054222052587,
|
| 146 |
+
"grad_norm": 4.325068950653076,
|
| 147 |
+
"learning_rate": 2.999997473658514e-05,
|
| 148 |
+
"loss": 2.0617,
|
| 149 |
+
"step": 300
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.0006134156933155216,
|
| 153 |
+
"grad_norm": 4.770871162414551,
|
| 154 |
+
"learning_rate": 2.999997214708592e-05,
|
| 155 |
+
"loss": 2.0922,
|
| 156 |
+
"step": 315
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 0.0006426259644257845,
|
| 160 |
+
"grad_norm": 4.1802754402160645,
|
| 161 |
+
"learning_rate": 2.999996943126982e-05,
|
| 162 |
+
"loss": 2.217,
|
| 163 |
+
"step": 330
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.0006718362355360475,
|
| 167 |
+
"grad_norm": 3.044260263442993,
|
| 168 |
+
"learning_rate": 2.999996658913687e-05,
|
| 169 |
+
"loss": 2.0508,
|
| 170 |
+
"step": 345
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.0007010465066463104,
|
| 174 |
+
"grad_norm": 4.1525187492370605,
|
| 175 |
+
"learning_rate": 2.9999963620687095e-05,
|
| 176 |
+
"loss": 2.1678,
|
| 177 |
+
"step": 360
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.0007302567777565733,
|
| 181 |
+
"grad_norm": 2.5676231384277344,
|
| 182 |
+
"learning_rate": 2.9999960525920515e-05,
|
| 183 |
+
"loss": 2.1693,
|
| 184 |
+
"step": 375
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 0.0007594670488668362,
|
| 188 |
+
"grad_norm": 3.880307674407959,
|
| 189 |
+
"learning_rate": 2.9999957304837156e-05,
|
| 190 |
+
"loss": 1.9008,
|
| 191 |
+
"step": 390
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.0007886773199770992,
|
| 195 |
+
"grad_norm": 3.3433754444122314,
|
| 196 |
+
"learning_rate": 2.999995395743705e-05,
|
| 197 |
+
"loss": 2.0734,
|
| 198 |
+
"step": 405
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.0008178875910873621,
|
| 202 |
+
"grad_norm": 3.451378583908081,
|
| 203 |
+
"learning_rate": 2.999995048372022e-05,
|
| 204 |
+
"loss": 1.9901,
|
| 205 |
+
"step": 420
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.000847097862197625,
|
| 209 |
+
"grad_norm": 3.4411580562591553,
|
| 210 |
+
"learning_rate": 2.9999946883686695e-05,
|
| 211 |
+
"loss": 1.9734,
|
| 212 |
+
"step": 435
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.000876308133307888,
|
| 216 |
+
"grad_norm": 3.514651298522949,
|
| 217 |
+
"learning_rate": 2.999994315733651e-05,
|
| 218 |
+
"loss": 2.1632,
|
| 219 |
+
"step": 450
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.0009055184044181509,
|
| 223 |
+
"grad_norm": 5.380824565887451,
|
| 224 |
+
"learning_rate": 2.999993930466969e-05,
|
| 225 |
+
"loss": 2.1534,
|
| 226 |
+
"step": 465
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.0009347286755284138,
|
| 230 |
+
"grad_norm": 3.768618106842041,
|
| 231 |
+
"learning_rate": 2.999993532568628e-05,
|
| 232 |
+
"loss": 2.2182,
|
| 233 |
+
"step": 480
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.0009639389466386767,
|
| 237 |
+
"grad_norm": 2.9490509033203125,
|
| 238 |
+
"learning_rate": 2.99999312203863e-05,
|
| 239 |
+
"loss": 1.9677,
|
| 240 |
+
"step": 495
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 0.0009931492177489396,
|
| 244 |
+
"grad_norm": 3.4881184101104736,
|
| 245 |
+
"learning_rate": 2.9999926988769788e-05,
|
| 246 |
+
"loss": 2.0786,
|
| 247 |
+
"step": 510
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 0.0010223594888592025,
|
| 251 |
+
"grad_norm": 3.788675546646118,
|
| 252 |
+
"learning_rate": 2.9999922630836784e-05,
|
| 253 |
+
"loss": 1.9696,
|
| 254 |
+
"step": 525
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 0.0010515697599694654,
|
| 258 |
+
"grad_norm": 3.070878744125366,
|
| 259 |
+
"learning_rate": 2.999991814658732e-05,
|
| 260 |
+
"loss": 1.9053,
|
| 261 |
+
"step": 540
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 0.0010807800310797286,
|
| 265 |
+
"grad_norm": 2.456150531768799,
|
| 266 |
+
"learning_rate": 2.9999913536021436e-05,
|
| 267 |
+
"loss": 1.9813,
|
| 268 |
+
"step": 555
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 0.0011099903021899915,
|
| 272 |
+
"grad_norm": 4.81670618057251,
|
| 273 |
+
"learning_rate": 2.999990879913917e-05,
|
| 274 |
+
"loss": 1.9317,
|
| 275 |
+
"step": 570
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 0.0011392005733002544,
|
| 279 |
+
"grad_norm": 4.014492988586426,
|
| 280 |
+
"learning_rate": 2.9999903935940567e-05,
|
| 281 |
+
"loss": 1.9709,
|
| 282 |
+
"step": 585
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.0011684108444105174,
|
| 286 |
+
"grad_norm": 4.765859603881836,
|
| 287 |
+
"learning_rate": 2.999989894642566e-05,
|
| 288 |
+
"loss": 1.9252,
|
| 289 |
+
"step": 600
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 0.0011976211155207803,
|
| 293 |
+
"grad_norm": 3.5599727630615234,
|
| 294 |
+
"learning_rate": 2.9999893830594492e-05,
|
| 295 |
+
"loss": 2.0175,
|
| 296 |
+
"step": 615
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 0.0012268313866310432,
|
| 300 |
+
"grad_norm": 4.441170692443848,
|
| 301 |
+
"learning_rate": 2.999988858844711e-05,
|
| 302 |
+
"loss": 2.1166,
|
| 303 |
+
"step": 630
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"epoch": 0.0012560416577413061,
|
| 307 |
+
"grad_norm": 2.691765308380127,
|
| 308 |
+
"learning_rate": 2.9999883219983558e-05,
|
| 309 |
+
"loss": 1.991,
|
| 310 |
+
"step": 645
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.001285251928851569,
|
| 314 |
+
"grad_norm": 4.039743423461914,
|
| 315 |
+
"learning_rate": 2.9999877725203878e-05,
|
| 316 |
+
"loss": 1.9917,
|
| 317 |
+
"step": 660
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 0.001314462199961832,
|
| 321 |
+
"grad_norm": 3.0612881183624268,
|
| 322 |
+
"learning_rate": 2.9999872104108115e-05,
|
| 323 |
+
"loss": 2.0436,
|
| 324 |
+
"step": 675
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.001343672471072095,
|
| 328 |
+
"grad_norm": 3.3579518795013428,
|
| 329 |
+
"learning_rate": 2.9999866356696326e-05,
|
| 330 |
+
"loss": 1.9659,
|
| 331 |
+
"step": 690
|
| 332 |
+
},
|
| 333 |
+
{
|
| 334 |
+
"epoch": 0.0013728827421823579,
|
| 335 |
+
"grad_norm": 4.145946502685547,
|
| 336 |
+
"learning_rate": 2.9999860482968552e-05,
|
| 337 |
+
"loss": 2.0267,
|
| 338 |
+
"step": 705
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 0.0014020930132926208,
|
| 342 |
+
"grad_norm": 5.213028430938721,
|
| 343 |
+
"learning_rate": 2.999985448292484e-05,
|
| 344 |
+
"loss": 2.0953,
|
| 345 |
+
"step": 720
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.0014313032844028837,
|
| 349 |
+
"grad_norm": 2.440793752670288,
|
| 350 |
+
"learning_rate": 2.9999848356565246e-05,
|
| 351 |
+
"loss": 2.0218,
|
| 352 |
+
"step": 735
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.0014605135555131466,
|
| 356 |
+
"grad_norm": 2.6167280673980713,
|
| 357 |
+
"learning_rate": 2.9999842103889813e-05,
|
| 358 |
+
"loss": 2.0638,
|
| 359 |
+
"step": 750
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 0.0014897238266234096,
|
| 363 |
+
"grad_norm": 2.2680811882019043,
|
| 364 |
+
"learning_rate": 2.999983572489861e-05,
|
| 365 |
+
"loss": 1.973,
|
| 366 |
+
"step": 765
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.0015189340977336725,
|
| 370 |
+
"grad_norm": 2.905947208404541,
|
| 371 |
+
"learning_rate": 2.999982921959167e-05,
|
| 372 |
+
"loss": 1.9346,
|
| 373 |
+
"step": 780
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.0015481443688439354,
|
| 377 |
+
"grad_norm": 4.673079967498779,
|
| 378 |
+
"learning_rate": 2.999982258796907e-05,
|
| 379 |
+
"loss": 2.1666,
|
| 380 |
+
"step": 795
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 0.0015773546399541984,
|
| 384 |
+
"grad_norm": 2.5963408946990967,
|
| 385 |
+
"learning_rate": 2.9999815830030846e-05,
|
| 386 |
+
"loss": 2.1063,
|
| 387 |
+
"step": 810
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 0.0016065649110644613,
|
| 391 |
+
"grad_norm": 3.656632661819458,
|
| 392 |
+
"learning_rate": 2.9999808945777066e-05,
|
| 393 |
+
"loss": 2.1328,
|
| 394 |
+
"step": 825
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 0.0016357751821747242,
|
| 398 |
+
"grad_norm": 4.4109368324279785,
|
| 399 |
+
"learning_rate": 2.9999801935207786e-05,
|
| 400 |
+
"loss": 1.9602,
|
| 401 |
+
"step": 840
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.0016649854532849871,
|
| 405 |
+
"grad_norm": 2.10768985748291,
|
| 406 |
+
"learning_rate": 2.9999794798323065e-05,
|
| 407 |
+
"loss": 1.9766,
|
| 408 |
+
"step": 855
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"epoch": 0.00169419572439525,
|
| 412 |
+
"grad_norm": 4.301672458648682,
|
| 413 |
+
"learning_rate": 2.999978753512296e-05,
|
| 414 |
+
"loss": 1.9386,
|
| 415 |
+
"step": 870
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 0.001723405995505513,
|
| 419 |
+
"grad_norm": 8.171302795410156,
|
| 420 |
+
"learning_rate": 2.9999780145607538e-05,
|
| 421 |
+
"loss": 2.1288,
|
| 422 |
+
"step": 885
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 0.001752616266615776,
|
| 426 |
+
"grad_norm": 3.4784069061279297,
|
| 427 |
+
"learning_rate": 2.999977262977685e-05,
|
| 428 |
+
"loss": 2.1346,
|
| 429 |
+
"step": 900
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.0017818265377260389,
|
| 433 |
+
"grad_norm": 2.9945054054260254,
|
| 434 |
+
"learning_rate": 2.9999764987630976e-05,
|
| 435 |
+
"loss": 2.0432,
|
| 436 |
+
"step": 915
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"epoch": 0.0018110368088363018,
|
| 440 |
+
"grad_norm": 3.632039785385132,
|
| 441 |
+
"learning_rate": 2.9999757219169964e-05,
|
| 442 |
+
"loss": 2.0642,
|
| 443 |
+
"step": 930
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.0018402470799465647,
|
| 447 |
+
"grad_norm": 3.9847047328948975,
|
| 448 |
+
"learning_rate": 2.999974932439389e-05,
|
| 449 |
+
"loss": 2.016,
|
| 450 |
+
"step": 945
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.0018694573510568276,
|
| 454 |
+
"grad_norm": 1.89451003074646,
|
| 455 |
+
"learning_rate": 2.9999741303302816e-05,
|
| 456 |
+
"loss": 2.0355,
|
| 457 |
+
"step": 960
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.0018986676221670906,
|
| 461 |
+
"grad_norm": 6.945189952850342,
|
| 462 |
+
"learning_rate": 2.9999733155896814e-05,
|
| 463 |
+
"loss": 1.97,
|
| 464 |
+
"step": 975
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.0019278778932773535,
|
| 468 |
+
"grad_norm": 3.347123384475708,
|
| 469 |
+
"learning_rate": 2.9999724882175947e-05,
|
| 470 |
+
"loss": 2.0421,
|
| 471 |
+
"step": 990
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"epoch": 0.0019570881643876164,
|
| 475 |
+
"grad_norm": 4.331577777862549,
|
| 476 |
+
"learning_rate": 2.9999716482140292e-05,
|
| 477 |
+
"loss": 2.133,
|
| 478 |
+
"step": 1005
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"epoch": 0.001986298435497879,
|
| 482 |
+
"grad_norm": 4.221049785614014,
|
| 483 |
+
"learning_rate": 2.999970795578991e-05,
|
| 484 |
+
"loss": 1.9272,
|
| 485 |
+
"step": 1020
|
| 486 |
+
},
|
| 487 |
+
{
|
| 488 |
+
"epoch": 0.0020155087066081423,
|
| 489 |
+
"grad_norm": 2.820549488067627,
|
| 490 |
+
"learning_rate": 2.9999699303124876e-05,
|
| 491 |
+
"loss": 2.0116,
|
| 492 |
+
"step": 1035
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"epoch": 0.002044718977718405,
|
| 496 |
+
"grad_norm": 2.2391343116760254,
|
| 497 |
+
"learning_rate": 2.9999690524145268e-05,
|
| 498 |
+
"loss": 1.9337,
|
| 499 |
+
"step": 1050
|
| 500 |
+
},
|
| 501 |
+
{
|
| 502 |
+
"epoch": 0.002073929248828668,
|
| 503 |
+
"grad_norm": 2.285027503967285,
|
| 504 |
+
"learning_rate": 2.9999681618851156e-05,
|
| 505 |
+
"loss": 1.9773,
|
| 506 |
+
"step": 1065
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.002103139519938931,
|
| 510 |
+
"grad_norm": 3.266909599304199,
|
| 511 |
+
"learning_rate": 2.9999672587242616e-05,
|
| 512 |
+
"loss": 1.9471,
|
| 513 |
+
"step": 1080
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"epoch": 0.002132349791049194,
|
| 517 |
+
"grad_norm": 5.03222131729126,
|
| 518 |
+
"learning_rate": 2.9999663429319722e-05,
|
| 519 |
+
"loss": 2.1337,
|
| 520 |
+
"step": 1095
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 0.002161560062159457,
|
| 524 |
+
"grad_norm": 2.044246196746826,
|
| 525 |
+
"learning_rate": 2.9999654145082552e-05,
|
| 526 |
+
"loss": 2.06,
|
| 527 |
+
"step": 1110
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 0.00219077033326972,
|
| 531 |
+
"grad_norm": 3.1399905681610107,
|
| 532 |
+
"learning_rate": 2.9999644734531183e-05,
|
| 533 |
+
"loss": 1.9174,
|
| 534 |
+
"step": 1125
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"epoch": 0.002219980604379983,
|
| 538 |
+
"grad_norm": 3.2445454597473145,
|
| 539 |
+
"learning_rate": 2.99996351976657e-05,
|
| 540 |
+
"loss": 1.9728,
|
| 541 |
+
"step": 1140
|
| 542 |
+
},
|
| 543 |
+
{
|
| 544 |
+
"epoch": 0.0022491908754902457,
|
| 545 |
+
"grad_norm": 2.3682138919830322,
|
| 546 |
+
"learning_rate": 2.999962553448618e-05,
|
| 547 |
+
"loss": 1.9779,
|
| 548 |
+
"step": 1155
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 0.002278401146600509,
|
| 552 |
+
"grad_norm": 2.283808708190918,
|
| 553 |
+
"learning_rate": 2.99996157449927e-05,
|
| 554 |
+
"loss": 1.964,
|
| 555 |
+
"step": 1170
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.0023076114177107716,
|
| 559 |
+
"grad_norm": 3.048382043838501,
|
| 560 |
+
"learning_rate": 2.9999605829185353e-05,
|
| 561 |
+
"loss": 2.0908,
|
| 562 |
+
"step": 1185
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"epoch": 0.0023368216888210347,
|
| 566 |
+
"grad_norm": 6.139120101928711,
|
| 567 |
+
"learning_rate": 2.999959578706421e-05,
|
| 568 |
+
"loss": 1.9535,
|
| 569 |
+
"step": 1200
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.0023660319599312974,
|
| 573 |
+
"grad_norm": 2.104069232940674,
|
| 574 |
+
"learning_rate": 2.9999585618629363e-05,
|
| 575 |
+
"loss": 1.8775,
|
| 576 |
+
"step": 1215
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 0.0023952422310415606,
|
| 580 |
+
"grad_norm": 2.531449317932129,
|
| 581 |
+
"learning_rate": 2.9999575323880894e-05,
|
| 582 |
+
"loss": 2.0309,
|
| 583 |
+
"step": 1230
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 0.0024244525021518233,
|
| 587 |
+
"grad_norm": 2.841078996658325,
|
| 588 |
+
"learning_rate": 2.9999564902818894e-05,
|
| 589 |
+
"loss": 2.0546,
|
| 590 |
+
"step": 1245
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 0.0024536627732620864,
|
| 594 |
+
"grad_norm": 2.0627243518829346,
|
| 595 |
+
"learning_rate": 2.9999554355443445e-05,
|
| 596 |
+
"loss": 1.8833,
|
| 597 |
+
"step": 1260
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.002482873044372349,
|
| 601 |
+
"grad_norm": 3.4014151096343994,
|
| 602 |
+
"learning_rate": 2.9999543681754645e-05,
|
| 603 |
+
"loss": 1.8935,
|
| 604 |
+
"step": 1275
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"epoch": 0.0025120833154826123,
|
| 608 |
+
"grad_norm": 3.507380962371826,
|
| 609 |
+
"learning_rate": 2.999953288175257e-05,
|
| 610 |
+
"loss": 1.9512,
|
| 611 |
+
"step": 1290
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 0.002541293586592875,
|
| 615 |
+
"grad_norm": 5.4273362159729,
|
| 616 |
+
"learning_rate": 2.9999521955437325e-05,
|
| 617 |
+
"loss": 1.9794,
|
| 618 |
+
"step": 1305
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 0.002570503857703138,
|
| 622 |
+
"grad_norm": 4.466185092926025,
|
| 623 |
+
"learning_rate": 2.9999510902808993e-05,
|
| 624 |
+
"loss": 1.8266,
|
| 625 |
+
"step": 1320
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"epoch": 0.002599714128813401,
|
| 629 |
+
"grad_norm": 2.4276440143585205,
|
| 630 |
+
"learning_rate": 2.9999499723867672e-05,
|
| 631 |
+
"loss": 2.0149,
|
| 632 |
+
"step": 1335
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 0.002628924399923664,
|
| 636 |
+
"grad_norm": 3.589876413345337,
|
| 637 |
+
"learning_rate": 2.9999488418613454e-05,
|
| 638 |
+
"loss": 1.8865,
|
| 639 |
+
"step": 1350
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 0.0026581346710339267,
|
| 643 |
+
"grad_norm": 3.750687599182129,
|
| 644 |
+
"learning_rate": 2.9999476987046434e-05,
|
| 645 |
+
"loss": 2.13,
|
| 646 |
+
"step": 1365
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.00268734494214419,
|
| 650 |
+
"grad_norm": 3.638850212097168,
|
| 651 |
+
"learning_rate": 2.999946542916671e-05,
|
| 652 |
+
"loss": 1.981,
|
| 653 |
+
"step": 1380
|
| 654 |
+
},
|
| 655 |
+
{
|
| 656 |
+
"epoch": 0.0027165552132544526,
|
| 657 |
+
"grad_norm": 4.610542297363281,
|
| 658 |
+
"learning_rate": 2.9999453744974375e-05,
|
| 659 |
+
"loss": 1.8887,
|
| 660 |
+
"step": 1395
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 0.0027457654843647157,
|
| 664 |
+
"grad_norm": 2.3758933544158936,
|
| 665 |
+
"learning_rate": 2.9999441934469534e-05,
|
| 666 |
+
"loss": 1.7867,
|
| 667 |
+
"step": 1410
|
| 668 |
+
},
|
| 669 |
+
{
|
| 670 |
+
"epoch": 0.0027749757554749784,
|
| 671 |
+
"grad_norm": 1.9892805814743042,
|
| 672 |
+
"learning_rate": 2.999942999765228e-05,
|
| 673 |
+
"loss": 1.8632,
|
| 674 |
+
"step": 1425
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 0.0028041860265852416,
|
| 678 |
+
"grad_norm": 2.5921459197998047,
|
| 679 |
+
"learning_rate": 2.999941793452272e-05,
|
| 680 |
+
"loss": 1.9155,
|
| 681 |
+
"step": 1440
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.0028333962976955043,
|
| 685 |
+
"grad_norm": 6.098895072937012,
|
| 686 |
+
"learning_rate": 2.9999405745080948e-05,
|
| 687 |
+
"loss": 2.0112,
|
| 688 |
+
"step": 1455
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.0028626065688057674,
|
| 692 |
+
"grad_norm": 2.7144930362701416,
|
| 693 |
+
"learning_rate": 2.9999393429327073e-05,
|
| 694 |
+
"loss": 2.0019,
|
| 695 |
+
"step": 1470
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 0.00289181683991603,
|
| 699 |
+
"grad_norm": 2.3964602947235107,
|
| 700 |
+
"learning_rate": 2.9999380987261195e-05,
|
| 701 |
+
"loss": 1.9524,
|
| 702 |
+
"step": 1485
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 0.0029210271110262933,
|
| 706 |
+
"grad_norm": 3.2644741535186768,
|
| 707 |
+
"learning_rate": 2.9999368418883422e-05,
|
| 708 |
+
"loss": 1.881,
|
| 709 |
+
"step": 1500
|
| 710 |
+
}
|
| 711 |
+
],
|
| 712 |
+
"logging_steps": 15,
|
| 713 |
+
"max_steps": 513518,
|
| 714 |
+
"num_input_tokens_seen": 0,
|
| 715 |
+
"num_train_epochs": 1,
|
| 716 |
+
"save_steps": 500,
|
| 717 |
+
"stateful_callbacks": {
|
| 718 |
+
"TrainerControl": {
|
| 719 |
+
"args": {
|
| 720 |
+
"should_epoch_stop": false,
|
| 721 |
+
"should_evaluate": false,
|
| 722 |
+
"should_log": false,
|
| 723 |
+
"should_save": true,
|
| 724 |
+
"should_training_stop": false
|
| 725 |
+
},
|
| 726 |
+
"attributes": {}
|
| 727 |
+
}
|
| 728 |
+
},
|
| 729 |
+
"total_flos": 2363166914015232.0,
|
| 730 |
+
"train_batch_size": 4,
|
| 731 |
+
"trial_name": null,
|
| 732 |
+
"trial_params": null
|
| 733 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-1500/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0287fbd96bea9652b04a1bb7ed4b6b1e8de92cccbc4b448c2b5d263a08664d48
|
| 3 |
+
size 5304
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "jonathanjordan21/mos-mamba-6x130m-trainer-sft",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MoSMambaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "jonathanjordan21/mos-mamba-6x130m-hf--configuration_mos_mamba.MoSMambaConfig",
|
| 8 |
+
"AutoModelForCausalLM": "jonathanjordan21/mos-mamba-6x130m-hf--modeling_mos_mamba.MoSMambaForCausalLM"
|
| 9 |
+
},
|
| 10 |
+
"bos_token_id": 0,
|
| 11 |
+
"conv_kernel": 4,
|
| 12 |
+
"d_inner": 1536,
|
| 13 |
+
"d_model": 768,
|
| 14 |
+
"eos_token_id": 50278,
|
| 15 |
+
"expand": 2,
|
| 16 |
+
"fused_add_norm": true,
|
| 17 |
+
"hidden_act": "silu",
|
| 18 |
+
"hidden_size": 768,
|
| 19 |
+
"initializer_range": 0.1,
|
| 20 |
+
"intermediate_size": 1536,
|
| 21 |
+
"layer_norm_epsilon": 1e-05,
|
| 22 |
+
"model_type": "MoSMamba",
|
| 23 |
+
"n_layer": 24,
|
| 24 |
+
"num_hidden_layers": 24,
|
| 25 |
+
"num_selectivities": 6,
|
| 26 |
+
"num_selectivities_per_tok": 2,
|
| 27 |
+
"output_router_logits": true,
|
| 28 |
+
"pad_token_id": 0,
|
| 29 |
+
"pad_vocab_size_multiple": 8,
|
| 30 |
+
"rescale_prenorm_residual": false,
|
| 31 |
+
"residual_in_fp32": true,
|
| 32 |
+
"rms_norm": true,
|
| 33 |
+
"ssm_cfg": {},
|
| 34 |
+
"state_size": 16,
|
| 35 |
+
"time_step_floor": 0.0001,
|
| 36 |
+
"time_step_init_scheme": "random",
|
| 37 |
+
"time_step_max": 0.1,
|
| 38 |
+
"time_step_min": 0.001,
|
| 39 |
+
"time_step_rank": 48,
|
| 40 |
+
"time_step_scale": 1.0,
|
| 41 |
+
"torch_dtype": "float32",
|
| 42 |
+
"transformers_version": "4.42.4",
|
| 43 |
+
"use_bias": false,
|
| 44 |
+
"use_cache": true,
|
| 45 |
+
"use_conv_bias": true,
|
| 46 |
+
"vocab_size": 50280
|
| 47 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 50278,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.42.4"
|
| 7 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e30236cfcb4deb77963ea4283e027b23ab0431d61e6e94411ad34f701480bad6
|
| 3 |
+
size 576008736
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b67fb2dd807ec41adaf01d5e09a6f54d81f3a08c2810ddf1c7d4c1f81df5d6bf
|
| 3 |
+
size 1152256984
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:571491d1edf7858876183ff428168c85d4e7d9f389bd675974be9edaef503790
|
| 3 |
+
size 14244
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:563cdca05f01d54c9f6c3cce222d522cc21645af3bd547239bd1e8cba90ef397
|
| 3 |
+
size 1064
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/special_tokens_map.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<|endoftext|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "<|im_end|>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": false,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": {
|
| 17 |
+
"content": "<|endoftext|>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": false,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": false
|
| 22 |
+
},
|
| 23 |
+
"unk_token": {
|
| 24 |
+
"content": "<|endoftext|>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": false,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": false
|
| 29 |
+
}
|
| 30 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/tokenizer_config.json
ADDED
|
@@ -0,0 +1,235 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"add_prefix_space": false,
|
| 5 |
+
"added_tokens_decoder": {
|
| 6 |
+
"0": {
|
| 7 |
+
"content": "<|endoftext|>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false,
|
| 12 |
+
"special": true
|
| 13 |
+
},
|
| 14 |
+
"1": {
|
| 15 |
+
"content": "<|padding|>",
|
| 16 |
+
"lstrip": false,
|
| 17 |
+
"normalized": false,
|
| 18 |
+
"rstrip": false,
|
| 19 |
+
"single_word": false,
|
| 20 |
+
"special": true
|
| 21 |
+
},
|
| 22 |
+
"50254": {
|
| 23 |
+
"content": " ",
|
| 24 |
+
"lstrip": false,
|
| 25 |
+
"normalized": true,
|
| 26 |
+
"rstrip": false,
|
| 27 |
+
"single_word": false,
|
| 28 |
+
"special": false
|
| 29 |
+
},
|
| 30 |
+
"50255": {
|
| 31 |
+
"content": " ",
|
| 32 |
+
"lstrip": false,
|
| 33 |
+
"normalized": true,
|
| 34 |
+
"rstrip": false,
|
| 35 |
+
"single_word": false,
|
| 36 |
+
"special": false
|
| 37 |
+
},
|
| 38 |
+
"50256": {
|
| 39 |
+
"content": " ",
|
| 40 |
+
"lstrip": false,
|
| 41 |
+
"normalized": true,
|
| 42 |
+
"rstrip": false,
|
| 43 |
+
"single_word": false,
|
| 44 |
+
"special": false
|
| 45 |
+
},
|
| 46 |
+
"50257": {
|
| 47 |
+
"content": " ",
|
| 48 |
+
"lstrip": false,
|
| 49 |
+
"normalized": true,
|
| 50 |
+
"rstrip": false,
|
| 51 |
+
"single_word": false,
|
| 52 |
+
"special": false
|
| 53 |
+
},
|
| 54 |
+
"50258": {
|
| 55 |
+
"content": " ",
|
| 56 |
+
"lstrip": false,
|
| 57 |
+
"normalized": true,
|
| 58 |
+
"rstrip": false,
|
| 59 |
+
"single_word": false,
|
| 60 |
+
"special": false
|
| 61 |
+
},
|
| 62 |
+
"50259": {
|
| 63 |
+
"content": " ",
|
| 64 |
+
"lstrip": false,
|
| 65 |
+
"normalized": true,
|
| 66 |
+
"rstrip": false,
|
| 67 |
+
"single_word": false,
|
| 68 |
+
"special": false
|
| 69 |
+
},
|
| 70 |
+
"50260": {
|
| 71 |
+
"content": " ",
|
| 72 |
+
"lstrip": false,
|
| 73 |
+
"normalized": true,
|
| 74 |
+
"rstrip": false,
|
| 75 |
+
"single_word": false,
|
| 76 |
+
"special": false
|
| 77 |
+
},
|
| 78 |
+
"50261": {
|
| 79 |
+
"content": " ",
|
| 80 |
+
"lstrip": false,
|
| 81 |
+
"normalized": true,
|
| 82 |
+
"rstrip": false,
|
| 83 |
+
"single_word": false,
|
| 84 |
+
"special": false
|
| 85 |
+
},
|
| 86 |
+
"50262": {
|
| 87 |
+
"content": " ",
|
| 88 |
+
"lstrip": false,
|
| 89 |
+
"normalized": true,
|
| 90 |
+
"rstrip": false,
|
| 91 |
+
"single_word": false,
|
| 92 |
+
"special": false
|
| 93 |
+
},
|
| 94 |
+
"50263": {
|
| 95 |
+
"content": " ",
|
| 96 |
+
"lstrip": false,
|
| 97 |
+
"normalized": true,
|
| 98 |
+
"rstrip": false,
|
| 99 |
+
"single_word": false,
|
| 100 |
+
"special": false
|
| 101 |
+
},
|
| 102 |
+
"50264": {
|
| 103 |
+
"content": " ",
|
| 104 |
+
"lstrip": false,
|
| 105 |
+
"normalized": true,
|
| 106 |
+
"rstrip": false,
|
| 107 |
+
"single_word": false,
|
| 108 |
+
"special": false
|
| 109 |
+
},
|
| 110 |
+
"50265": {
|
| 111 |
+
"content": " ",
|
| 112 |
+
"lstrip": false,
|
| 113 |
+
"normalized": true,
|
| 114 |
+
"rstrip": false,
|
| 115 |
+
"single_word": false,
|
| 116 |
+
"special": false
|
| 117 |
+
},
|
| 118 |
+
"50266": {
|
| 119 |
+
"content": " ",
|
| 120 |
+
"lstrip": false,
|
| 121 |
+
"normalized": true,
|
| 122 |
+
"rstrip": false,
|
| 123 |
+
"single_word": false,
|
| 124 |
+
"special": false
|
| 125 |
+
},
|
| 126 |
+
"50267": {
|
| 127 |
+
"content": " ",
|
| 128 |
+
"lstrip": false,
|
| 129 |
+
"normalized": true,
|
| 130 |
+
"rstrip": false,
|
| 131 |
+
"single_word": false,
|
| 132 |
+
"special": false
|
| 133 |
+
},
|
| 134 |
+
"50268": {
|
| 135 |
+
"content": " ",
|
| 136 |
+
"lstrip": false,
|
| 137 |
+
"normalized": true,
|
| 138 |
+
"rstrip": false,
|
| 139 |
+
"single_word": false,
|
| 140 |
+
"special": false
|
| 141 |
+
},
|
| 142 |
+
"50269": {
|
| 143 |
+
"content": " ",
|
| 144 |
+
"lstrip": false,
|
| 145 |
+
"normalized": true,
|
| 146 |
+
"rstrip": false,
|
| 147 |
+
"single_word": false,
|
| 148 |
+
"special": false
|
| 149 |
+
},
|
| 150 |
+
"50270": {
|
| 151 |
+
"content": " ",
|
| 152 |
+
"lstrip": false,
|
| 153 |
+
"normalized": true,
|
| 154 |
+
"rstrip": false,
|
| 155 |
+
"single_word": false,
|
| 156 |
+
"special": false
|
| 157 |
+
},
|
| 158 |
+
"50271": {
|
| 159 |
+
"content": " ",
|
| 160 |
+
"lstrip": false,
|
| 161 |
+
"normalized": true,
|
| 162 |
+
"rstrip": false,
|
| 163 |
+
"single_word": false,
|
| 164 |
+
"special": false
|
| 165 |
+
},
|
| 166 |
+
"50272": {
|
| 167 |
+
"content": " ",
|
| 168 |
+
"lstrip": false,
|
| 169 |
+
"normalized": true,
|
| 170 |
+
"rstrip": false,
|
| 171 |
+
"single_word": false,
|
| 172 |
+
"special": false
|
| 173 |
+
},
|
| 174 |
+
"50273": {
|
| 175 |
+
"content": " ",
|
| 176 |
+
"lstrip": false,
|
| 177 |
+
"normalized": true,
|
| 178 |
+
"rstrip": false,
|
| 179 |
+
"single_word": false,
|
| 180 |
+
"special": false
|
| 181 |
+
},
|
| 182 |
+
"50274": {
|
| 183 |
+
"content": " ",
|
| 184 |
+
"lstrip": false,
|
| 185 |
+
"normalized": true,
|
| 186 |
+
"rstrip": false,
|
| 187 |
+
"single_word": false,
|
| 188 |
+
"special": false
|
| 189 |
+
},
|
| 190 |
+
"50275": {
|
| 191 |
+
"content": " ",
|
| 192 |
+
"lstrip": false,
|
| 193 |
+
"normalized": true,
|
| 194 |
+
"rstrip": false,
|
| 195 |
+
"single_word": false,
|
| 196 |
+
"special": false
|
| 197 |
+
},
|
| 198 |
+
"50276": {
|
| 199 |
+
"content": " ",
|
| 200 |
+
"lstrip": false,
|
| 201 |
+
"normalized": true,
|
| 202 |
+
"rstrip": false,
|
| 203 |
+
"single_word": false,
|
| 204 |
+
"special": false
|
| 205 |
+
},
|
| 206 |
+
"50277": {
|
| 207 |
+
"content": "<|im_start|>",
|
| 208 |
+
"lstrip": false,
|
| 209 |
+
"normalized": false,
|
| 210 |
+
"rstrip": false,
|
| 211 |
+
"single_word": false,
|
| 212 |
+
"special": true
|
| 213 |
+
},
|
| 214 |
+
"50278": {
|
| 215 |
+
"content": "<|im_end|>",
|
| 216 |
+
"lstrip": false,
|
| 217 |
+
"normalized": false,
|
| 218 |
+
"rstrip": false,
|
| 219 |
+
"single_word": false,
|
| 220 |
+
"special": true
|
| 221 |
+
}
|
| 222 |
+
},
|
| 223 |
+
"bos_token": "<|endoftext|>",
|
| 224 |
+
"chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
|
| 225 |
+
"clean_up_tokenization_spaces": true,
|
| 226 |
+
"eos_token": "<|im_end|>",
|
| 227 |
+
"max_length": 1024,
|
| 228 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 229 |
+
"pad_token": "<|endoftext|>",
|
| 230 |
+
"stride": 0,
|
| 231 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
| 232 |
+
"truncation_side": "right",
|
| 233 |
+
"truncation_strategy": "longest_first",
|
| 234 |
+
"unk_token": "<|endoftext|>"
|
| 235 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/trainer_state.json
ADDED
|
@@ -0,0 +1,964 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.0038947028147017244,
|
| 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": 2.9210271110262932e-05,
|
| 13 |
+
"grad_norm": 5.397985935211182,
|
| 14 |
+
"learning_rate": 2.9999999936841446e-05,
|
| 15 |
+
"loss": 2.1156,
|
| 16 |
+
"step": 15
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 5.8420542220525864e-05,
|
| 20 |
+
"grad_norm": 3.999190092086792,
|
| 21 |
+
"learning_rate": 2.999999974736578e-05,
|
| 22 |
+
"loss": 2.5268,
|
| 23 |
+
"step": 30
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 8.76308133307888e-05,
|
| 27 |
+
"grad_norm": 5.580333709716797,
|
| 28 |
+
"learning_rate": 2.999999943157301e-05,
|
| 29 |
+
"loss": 2.3955,
|
| 30 |
+
"step": 45
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.00011684108444105173,
|
| 34 |
+
"grad_norm": 3.6752560138702393,
|
| 35 |
+
"learning_rate": 2.9999998989463132e-05,
|
| 36 |
+
"loss": 2.171,
|
| 37 |
+
"step": 60
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.00014605135555131467,
|
| 41 |
+
"grad_norm": 2.655461072921753,
|
| 42 |
+
"learning_rate": 2.9999998421036153e-05,
|
| 43 |
+
"loss": 2.188,
|
| 44 |
+
"step": 75
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.0001752616266615776,
|
| 48 |
+
"grad_norm": 2.572371006011963,
|
| 49 |
+
"learning_rate": 2.9999997726292083e-05,
|
| 50 |
+
"loss": 2.1977,
|
| 51 |
+
"step": 90
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.00020447189777184053,
|
| 55 |
+
"grad_norm": 8.544811248779297,
|
| 56 |
+
"learning_rate": 2.999999690523092e-05,
|
| 57 |
+
"loss": 2.1871,
|
| 58 |
+
"step": 105
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.00023368216888210345,
|
| 62 |
+
"grad_norm": 3.530529737472534,
|
| 63 |
+
"learning_rate": 2.999999595785267e-05,
|
| 64 |
+
"loss": 2.2238,
|
| 65 |
+
"step": 120
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.00026289243999236636,
|
| 69 |
+
"grad_norm": 3.671182155609131,
|
| 70 |
+
"learning_rate": 2.9999994884157345e-05,
|
| 71 |
+
"loss": 2.1338,
|
| 72 |
+
"step": 135
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.00029210271110262934,
|
| 76 |
+
"grad_norm": 3.9643137454986572,
|
| 77 |
+
"learning_rate": 2.9999993684144956e-05,
|
| 78 |
+
"loss": 2.0788,
|
| 79 |
+
"step": 150
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.00032131298221289227,
|
| 83 |
+
"grad_norm": 2.62164044380188,
|
| 84 |
+
"learning_rate": 2.9999992357815508e-05,
|
| 85 |
+
"loss": 2.117,
|
| 86 |
+
"step": 165
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.0003505232533231552,
|
| 90 |
+
"grad_norm": 4.189517021179199,
|
| 91 |
+
"learning_rate": 2.999999090516902e-05,
|
| 92 |
+
"loss": 2.262,
|
| 93 |
+
"step": 180
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.0003797335244334181,
|
| 97 |
+
"grad_norm": 2.880622625350952,
|
| 98 |
+
"learning_rate": 2.9999989326205494e-05,
|
| 99 |
+
"loss": 2.1102,
|
| 100 |
+
"step": 195
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.00040894379554368105,
|
| 104 |
+
"grad_norm": 5.852839469909668,
|
| 105 |
+
"learning_rate": 2.9999987620924948e-05,
|
| 106 |
+
"loss": 2.0684,
|
| 107 |
+
"step": 210
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.000438154066653944,
|
| 111 |
+
"grad_norm": 3.6339924335479736,
|
| 112 |
+
"learning_rate": 2.9999985789327394e-05,
|
| 113 |
+
"loss": 2.2518,
|
| 114 |
+
"step": 225
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.0004673643377642069,
|
| 118 |
+
"grad_norm": 2.743267297744751,
|
| 119 |
+
"learning_rate": 2.9999983831412858e-05,
|
| 120 |
+
"loss": 2.1236,
|
| 121 |
+
"step": 240
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.0004965746088744698,
|
| 125 |
+
"grad_norm": 3.9490363597869873,
|
| 126 |
+
"learning_rate": 2.9999981747181345e-05,
|
| 127 |
+
"loss": 2.0878,
|
| 128 |
+
"step": 255
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.0005257848799847327,
|
| 132 |
+
"grad_norm": 3.0342047214508057,
|
| 133 |
+
"learning_rate": 2.9999979536632872e-05,
|
| 134 |
+
"loss": 2.1415,
|
| 135 |
+
"step": 270
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.0005549951510949957,
|
| 139 |
+
"grad_norm": 4.045855522155762,
|
| 140 |
+
"learning_rate": 2.9999977199767467e-05,
|
| 141 |
+
"loss": 2.0558,
|
| 142 |
+
"step": 285
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"epoch": 0.0005842054222052587,
|
| 146 |
+
"grad_norm": 4.325068950653076,
|
| 147 |
+
"learning_rate": 2.999997473658514e-05,
|
| 148 |
+
"loss": 2.0617,
|
| 149 |
+
"step": 300
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.0006134156933155216,
|
| 153 |
+
"grad_norm": 4.770871162414551,
|
| 154 |
+
"learning_rate": 2.999997214708592e-05,
|
| 155 |
+
"loss": 2.0922,
|
| 156 |
+
"step": 315
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 0.0006426259644257845,
|
| 160 |
+
"grad_norm": 4.1802754402160645,
|
| 161 |
+
"learning_rate": 2.999996943126982e-05,
|
| 162 |
+
"loss": 2.217,
|
| 163 |
+
"step": 330
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.0006718362355360475,
|
| 167 |
+
"grad_norm": 3.044260263442993,
|
| 168 |
+
"learning_rate": 2.999996658913687e-05,
|
| 169 |
+
"loss": 2.0508,
|
| 170 |
+
"step": 345
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.0007010465066463104,
|
| 174 |
+
"grad_norm": 4.1525187492370605,
|
| 175 |
+
"learning_rate": 2.9999963620687095e-05,
|
| 176 |
+
"loss": 2.1678,
|
| 177 |
+
"step": 360
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.0007302567777565733,
|
| 181 |
+
"grad_norm": 2.5676231384277344,
|
| 182 |
+
"learning_rate": 2.9999960525920515e-05,
|
| 183 |
+
"loss": 2.1693,
|
| 184 |
+
"step": 375
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 0.0007594670488668362,
|
| 188 |
+
"grad_norm": 3.880307674407959,
|
| 189 |
+
"learning_rate": 2.9999957304837156e-05,
|
| 190 |
+
"loss": 1.9008,
|
| 191 |
+
"step": 390
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.0007886773199770992,
|
| 195 |
+
"grad_norm": 3.3433754444122314,
|
| 196 |
+
"learning_rate": 2.999995395743705e-05,
|
| 197 |
+
"loss": 2.0734,
|
| 198 |
+
"step": 405
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.0008178875910873621,
|
| 202 |
+
"grad_norm": 3.451378583908081,
|
| 203 |
+
"learning_rate": 2.999995048372022e-05,
|
| 204 |
+
"loss": 1.9901,
|
| 205 |
+
"step": 420
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.000847097862197625,
|
| 209 |
+
"grad_norm": 3.4411580562591553,
|
| 210 |
+
"learning_rate": 2.9999946883686695e-05,
|
| 211 |
+
"loss": 1.9734,
|
| 212 |
+
"step": 435
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.000876308133307888,
|
| 216 |
+
"grad_norm": 3.514651298522949,
|
| 217 |
+
"learning_rate": 2.999994315733651e-05,
|
| 218 |
+
"loss": 2.1632,
|
| 219 |
+
"step": 450
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.0009055184044181509,
|
| 223 |
+
"grad_norm": 5.380824565887451,
|
| 224 |
+
"learning_rate": 2.999993930466969e-05,
|
| 225 |
+
"loss": 2.1534,
|
| 226 |
+
"step": 465
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.0009347286755284138,
|
| 230 |
+
"grad_norm": 3.768618106842041,
|
| 231 |
+
"learning_rate": 2.999993532568628e-05,
|
| 232 |
+
"loss": 2.2182,
|
| 233 |
+
"step": 480
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.0009639389466386767,
|
| 237 |
+
"grad_norm": 2.9490509033203125,
|
| 238 |
+
"learning_rate": 2.99999312203863e-05,
|
| 239 |
+
"loss": 1.9677,
|
| 240 |
+
"step": 495
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 0.0009931492177489396,
|
| 244 |
+
"grad_norm": 3.4881184101104736,
|
| 245 |
+
"learning_rate": 2.9999926988769788e-05,
|
| 246 |
+
"loss": 2.0786,
|
| 247 |
+
"step": 510
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 0.0010223594888592025,
|
| 251 |
+
"grad_norm": 3.788675546646118,
|
| 252 |
+
"learning_rate": 2.9999922630836784e-05,
|
| 253 |
+
"loss": 1.9696,
|
| 254 |
+
"step": 525
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 0.0010515697599694654,
|
| 258 |
+
"grad_norm": 3.070878744125366,
|
| 259 |
+
"learning_rate": 2.999991814658732e-05,
|
| 260 |
+
"loss": 1.9053,
|
| 261 |
+
"step": 540
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 0.0010807800310797286,
|
| 265 |
+
"grad_norm": 2.456150531768799,
|
| 266 |
+
"learning_rate": 2.9999913536021436e-05,
|
| 267 |
+
"loss": 1.9813,
|
| 268 |
+
"step": 555
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 0.0011099903021899915,
|
| 272 |
+
"grad_norm": 4.81670618057251,
|
| 273 |
+
"learning_rate": 2.999990879913917e-05,
|
| 274 |
+
"loss": 1.9317,
|
| 275 |
+
"step": 570
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 0.0011392005733002544,
|
| 279 |
+
"grad_norm": 4.014492988586426,
|
| 280 |
+
"learning_rate": 2.9999903935940567e-05,
|
| 281 |
+
"loss": 1.9709,
|
| 282 |
+
"step": 585
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.0011684108444105174,
|
| 286 |
+
"grad_norm": 4.765859603881836,
|
| 287 |
+
"learning_rate": 2.999989894642566e-05,
|
| 288 |
+
"loss": 1.9252,
|
| 289 |
+
"step": 600
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 0.0011976211155207803,
|
| 293 |
+
"grad_norm": 3.5599727630615234,
|
| 294 |
+
"learning_rate": 2.9999893830594492e-05,
|
| 295 |
+
"loss": 2.0175,
|
| 296 |
+
"step": 615
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 0.0012268313866310432,
|
| 300 |
+
"grad_norm": 4.441170692443848,
|
| 301 |
+
"learning_rate": 2.999988858844711e-05,
|
| 302 |
+
"loss": 2.1166,
|
| 303 |
+
"step": 630
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"epoch": 0.0012560416577413061,
|
| 307 |
+
"grad_norm": 2.691765308380127,
|
| 308 |
+
"learning_rate": 2.9999883219983558e-05,
|
| 309 |
+
"loss": 1.991,
|
| 310 |
+
"step": 645
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.001285251928851569,
|
| 314 |
+
"grad_norm": 4.039743423461914,
|
| 315 |
+
"learning_rate": 2.9999877725203878e-05,
|
| 316 |
+
"loss": 1.9917,
|
| 317 |
+
"step": 660
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 0.001314462199961832,
|
| 321 |
+
"grad_norm": 3.0612881183624268,
|
| 322 |
+
"learning_rate": 2.9999872104108115e-05,
|
| 323 |
+
"loss": 2.0436,
|
| 324 |
+
"step": 675
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.001343672471072095,
|
| 328 |
+
"grad_norm": 3.3579518795013428,
|
| 329 |
+
"learning_rate": 2.9999866356696326e-05,
|
| 330 |
+
"loss": 1.9659,
|
| 331 |
+
"step": 690
|
| 332 |
+
},
|
| 333 |
+
{
|
| 334 |
+
"epoch": 0.0013728827421823579,
|
| 335 |
+
"grad_norm": 4.145946502685547,
|
| 336 |
+
"learning_rate": 2.9999860482968552e-05,
|
| 337 |
+
"loss": 2.0267,
|
| 338 |
+
"step": 705
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 0.0014020930132926208,
|
| 342 |
+
"grad_norm": 5.213028430938721,
|
| 343 |
+
"learning_rate": 2.999985448292484e-05,
|
| 344 |
+
"loss": 2.0953,
|
| 345 |
+
"step": 720
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.0014313032844028837,
|
| 349 |
+
"grad_norm": 2.440793752670288,
|
| 350 |
+
"learning_rate": 2.9999848356565246e-05,
|
| 351 |
+
"loss": 2.0218,
|
| 352 |
+
"step": 735
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.0014605135555131466,
|
| 356 |
+
"grad_norm": 2.6167280673980713,
|
| 357 |
+
"learning_rate": 2.9999842103889813e-05,
|
| 358 |
+
"loss": 2.0638,
|
| 359 |
+
"step": 750
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 0.0014897238266234096,
|
| 363 |
+
"grad_norm": 2.2680811882019043,
|
| 364 |
+
"learning_rate": 2.999983572489861e-05,
|
| 365 |
+
"loss": 1.973,
|
| 366 |
+
"step": 765
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.0015189340977336725,
|
| 370 |
+
"grad_norm": 2.905947208404541,
|
| 371 |
+
"learning_rate": 2.999982921959167e-05,
|
| 372 |
+
"loss": 1.9346,
|
| 373 |
+
"step": 780
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.0015481443688439354,
|
| 377 |
+
"grad_norm": 4.673079967498779,
|
| 378 |
+
"learning_rate": 2.999982258796907e-05,
|
| 379 |
+
"loss": 2.1666,
|
| 380 |
+
"step": 795
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 0.0015773546399541984,
|
| 384 |
+
"grad_norm": 2.5963408946990967,
|
| 385 |
+
"learning_rate": 2.9999815830030846e-05,
|
| 386 |
+
"loss": 2.1063,
|
| 387 |
+
"step": 810
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 0.0016065649110644613,
|
| 391 |
+
"grad_norm": 3.656632661819458,
|
| 392 |
+
"learning_rate": 2.9999808945777066e-05,
|
| 393 |
+
"loss": 2.1328,
|
| 394 |
+
"step": 825
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 0.0016357751821747242,
|
| 398 |
+
"grad_norm": 4.4109368324279785,
|
| 399 |
+
"learning_rate": 2.9999801935207786e-05,
|
| 400 |
+
"loss": 1.9602,
|
| 401 |
+
"step": 840
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.0016649854532849871,
|
| 405 |
+
"grad_norm": 2.10768985748291,
|
| 406 |
+
"learning_rate": 2.9999794798323065e-05,
|
| 407 |
+
"loss": 1.9766,
|
| 408 |
+
"step": 855
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"epoch": 0.00169419572439525,
|
| 412 |
+
"grad_norm": 4.301672458648682,
|
| 413 |
+
"learning_rate": 2.999978753512296e-05,
|
| 414 |
+
"loss": 1.9386,
|
| 415 |
+
"step": 870
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 0.001723405995505513,
|
| 419 |
+
"grad_norm": 8.171302795410156,
|
| 420 |
+
"learning_rate": 2.9999780145607538e-05,
|
| 421 |
+
"loss": 2.1288,
|
| 422 |
+
"step": 885
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 0.001752616266615776,
|
| 426 |
+
"grad_norm": 3.4784069061279297,
|
| 427 |
+
"learning_rate": 2.999977262977685e-05,
|
| 428 |
+
"loss": 2.1346,
|
| 429 |
+
"step": 900
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.0017818265377260389,
|
| 433 |
+
"grad_norm": 2.9945054054260254,
|
| 434 |
+
"learning_rate": 2.9999764987630976e-05,
|
| 435 |
+
"loss": 2.0432,
|
| 436 |
+
"step": 915
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"epoch": 0.0018110368088363018,
|
| 440 |
+
"grad_norm": 3.632039785385132,
|
| 441 |
+
"learning_rate": 2.9999757219169964e-05,
|
| 442 |
+
"loss": 2.0642,
|
| 443 |
+
"step": 930
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.0018402470799465647,
|
| 447 |
+
"grad_norm": 3.9847047328948975,
|
| 448 |
+
"learning_rate": 2.999974932439389e-05,
|
| 449 |
+
"loss": 2.016,
|
| 450 |
+
"step": 945
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.0018694573510568276,
|
| 454 |
+
"grad_norm": 1.89451003074646,
|
| 455 |
+
"learning_rate": 2.9999741303302816e-05,
|
| 456 |
+
"loss": 2.0355,
|
| 457 |
+
"step": 960
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.0018986676221670906,
|
| 461 |
+
"grad_norm": 6.945189952850342,
|
| 462 |
+
"learning_rate": 2.9999733155896814e-05,
|
| 463 |
+
"loss": 1.97,
|
| 464 |
+
"step": 975
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.0019278778932773535,
|
| 468 |
+
"grad_norm": 3.347123384475708,
|
| 469 |
+
"learning_rate": 2.9999724882175947e-05,
|
| 470 |
+
"loss": 2.0421,
|
| 471 |
+
"step": 990
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"epoch": 0.0019570881643876164,
|
| 475 |
+
"grad_norm": 4.331577777862549,
|
| 476 |
+
"learning_rate": 2.9999716482140292e-05,
|
| 477 |
+
"loss": 2.133,
|
| 478 |
+
"step": 1005
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"epoch": 0.001986298435497879,
|
| 482 |
+
"grad_norm": 4.221049785614014,
|
| 483 |
+
"learning_rate": 2.999970795578991e-05,
|
| 484 |
+
"loss": 1.9272,
|
| 485 |
+
"step": 1020
|
| 486 |
+
},
|
| 487 |
+
{
|
| 488 |
+
"epoch": 0.0020155087066081423,
|
| 489 |
+
"grad_norm": 2.820549488067627,
|
| 490 |
+
"learning_rate": 2.9999699303124876e-05,
|
| 491 |
+
"loss": 2.0116,
|
| 492 |
+
"step": 1035
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"epoch": 0.002044718977718405,
|
| 496 |
+
"grad_norm": 2.2391343116760254,
|
| 497 |
+
"learning_rate": 2.9999690524145268e-05,
|
| 498 |
+
"loss": 1.9337,
|
| 499 |
+
"step": 1050
|
| 500 |
+
},
|
| 501 |
+
{
|
| 502 |
+
"epoch": 0.002073929248828668,
|
| 503 |
+
"grad_norm": 2.285027503967285,
|
| 504 |
+
"learning_rate": 2.9999681618851156e-05,
|
| 505 |
+
"loss": 1.9773,
|
| 506 |
+
"step": 1065
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.002103139519938931,
|
| 510 |
+
"grad_norm": 3.266909599304199,
|
| 511 |
+
"learning_rate": 2.9999672587242616e-05,
|
| 512 |
+
"loss": 1.9471,
|
| 513 |
+
"step": 1080
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"epoch": 0.002132349791049194,
|
| 517 |
+
"grad_norm": 5.03222131729126,
|
| 518 |
+
"learning_rate": 2.9999663429319722e-05,
|
| 519 |
+
"loss": 2.1337,
|
| 520 |
+
"step": 1095
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 0.002161560062159457,
|
| 524 |
+
"grad_norm": 2.044246196746826,
|
| 525 |
+
"learning_rate": 2.9999654145082552e-05,
|
| 526 |
+
"loss": 2.06,
|
| 527 |
+
"step": 1110
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 0.00219077033326972,
|
| 531 |
+
"grad_norm": 3.1399905681610107,
|
| 532 |
+
"learning_rate": 2.9999644734531183e-05,
|
| 533 |
+
"loss": 1.9174,
|
| 534 |
+
"step": 1125
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"epoch": 0.002219980604379983,
|
| 538 |
+
"grad_norm": 3.2445454597473145,
|
| 539 |
+
"learning_rate": 2.99996351976657e-05,
|
| 540 |
+
"loss": 1.9728,
|
| 541 |
+
"step": 1140
|
| 542 |
+
},
|
| 543 |
+
{
|
| 544 |
+
"epoch": 0.0022491908754902457,
|
| 545 |
+
"grad_norm": 2.3682138919830322,
|
| 546 |
+
"learning_rate": 2.999962553448618e-05,
|
| 547 |
+
"loss": 1.9779,
|
| 548 |
+
"step": 1155
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 0.002278401146600509,
|
| 552 |
+
"grad_norm": 2.283808708190918,
|
| 553 |
+
"learning_rate": 2.99996157449927e-05,
|
| 554 |
+
"loss": 1.964,
|
| 555 |
+
"step": 1170
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.0023076114177107716,
|
| 559 |
+
"grad_norm": 3.048382043838501,
|
| 560 |
+
"learning_rate": 2.9999605829185353e-05,
|
| 561 |
+
"loss": 2.0908,
|
| 562 |
+
"step": 1185
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"epoch": 0.0023368216888210347,
|
| 566 |
+
"grad_norm": 6.139120101928711,
|
| 567 |
+
"learning_rate": 2.999959578706421e-05,
|
| 568 |
+
"loss": 1.9535,
|
| 569 |
+
"step": 1200
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.0023660319599312974,
|
| 573 |
+
"grad_norm": 2.104069232940674,
|
| 574 |
+
"learning_rate": 2.9999585618629363e-05,
|
| 575 |
+
"loss": 1.8775,
|
| 576 |
+
"step": 1215
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 0.0023952422310415606,
|
| 580 |
+
"grad_norm": 2.531449317932129,
|
| 581 |
+
"learning_rate": 2.9999575323880894e-05,
|
| 582 |
+
"loss": 2.0309,
|
| 583 |
+
"step": 1230
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 0.0024244525021518233,
|
| 587 |
+
"grad_norm": 2.841078996658325,
|
| 588 |
+
"learning_rate": 2.9999564902818894e-05,
|
| 589 |
+
"loss": 2.0546,
|
| 590 |
+
"step": 1245
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 0.0024536627732620864,
|
| 594 |
+
"grad_norm": 2.0627243518829346,
|
| 595 |
+
"learning_rate": 2.9999554355443445e-05,
|
| 596 |
+
"loss": 1.8833,
|
| 597 |
+
"step": 1260
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.002482873044372349,
|
| 601 |
+
"grad_norm": 3.4014151096343994,
|
| 602 |
+
"learning_rate": 2.9999543681754645e-05,
|
| 603 |
+
"loss": 1.8935,
|
| 604 |
+
"step": 1275
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"epoch": 0.0025120833154826123,
|
| 608 |
+
"grad_norm": 3.507380962371826,
|
| 609 |
+
"learning_rate": 2.999953288175257e-05,
|
| 610 |
+
"loss": 1.9512,
|
| 611 |
+
"step": 1290
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 0.002541293586592875,
|
| 615 |
+
"grad_norm": 5.4273362159729,
|
| 616 |
+
"learning_rate": 2.9999521955437325e-05,
|
| 617 |
+
"loss": 1.9794,
|
| 618 |
+
"step": 1305
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 0.002570503857703138,
|
| 622 |
+
"grad_norm": 4.466185092926025,
|
| 623 |
+
"learning_rate": 2.9999510902808993e-05,
|
| 624 |
+
"loss": 1.8266,
|
| 625 |
+
"step": 1320
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"epoch": 0.002599714128813401,
|
| 629 |
+
"grad_norm": 2.4276440143585205,
|
| 630 |
+
"learning_rate": 2.9999499723867672e-05,
|
| 631 |
+
"loss": 2.0149,
|
| 632 |
+
"step": 1335
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 0.002628924399923664,
|
| 636 |
+
"grad_norm": 3.589876413345337,
|
| 637 |
+
"learning_rate": 2.9999488418613454e-05,
|
| 638 |
+
"loss": 1.8865,
|
| 639 |
+
"step": 1350
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 0.0026581346710339267,
|
| 643 |
+
"grad_norm": 3.750687599182129,
|
| 644 |
+
"learning_rate": 2.9999476987046434e-05,
|
| 645 |
+
"loss": 2.13,
|
| 646 |
+
"step": 1365
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.00268734494214419,
|
| 650 |
+
"grad_norm": 3.638850212097168,
|
| 651 |
+
"learning_rate": 2.999946542916671e-05,
|
| 652 |
+
"loss": 1.981,
|
| 653 |
+
"step": 1380
|
| 654 |
+
},
|
| 655 |
+
{
|
| 656 |
+
"epoch": 0.0027165552132544526,
|
| 657 |
+
"grad_norm": 4.610542297363281,
|
| 658 |
+
"learning_rate": 2.9999453744974375e-05,
|
| 659 |
+
"loss": 1.8887,
|
| 660 |
+
"step": 1395
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 0.0027457654843647157,
|
| 664 |
+
"grad_norm": 2.3758933544158936,
|
| 665 |
+
"learning_rate": 2.9999441934469534e-05,
|
| 666 |
+
"loss": 1.7867,
|
| 667 |
+
"step": 1410
|
| 668 |
+
},
|
| 669 |
+
{
|
| 670 |
+
"epoch": 0.0027749757554749784,
|
| 671 |
+
"grad_norm": 1.9892805814743042,
|
| 672 |
+
"learning_rate": 2.999942999765228e-05,
|
| 673 |
+
"loss": 1.8632,
|
| 674 |
+
"step": 1425
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 0.0028041860265852416,
|
| 678 |
+
"grad_norm": 2.5921459197998047,
|
| 679 |
+
"learning_rate": 2.999941793452272e-05,
|
| 680 |
+
"loss": 1.9155,
|
| 681 |
+
"step": 1440
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.0028333962976955043,
|
| 685 |
+
"grad_norm": 6.098895072937012,
|
| 686 |
+
"learning_rate": 2.9999405745080948e-05,
|
| 687 |
+
"loss": 2.0112,
|
| 688 |
+
"step": 1455
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.0028626065688057674,
|
| 692 |
+
"grad_norm": 2.7144930362701416,
|
| 693 |
+
"learning_rate": 2.9999393429327073e-05,
|
| 694 |
+
"loss": 2.0019,
|
| 695 |
+
"step": 1470
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 0.00289181683991603,
|
| 699 |
+
"grad_norm": 2.3964602947235107,
|
| 700 |
+
"learning_rate": 2.9999380987261195e-05,
|
| 701 |
+
"loss": 1.9524,
|
| 702 |
+
"step": 1485
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 0.0029210271110262933,
|
| 706 |
+
"grad_norm": 3.2644741535186768,
|
| 707 |
+
"learning_rate": 2.9999368418883422e-05,
|
| 708 |
+
"loss": 1.881,
|
| 709 |
+
"step": 1500
|
| 710 |
+
},
|
| 711 |
+
{
|
| 712 |
+
"epoch": 0.002950237382136556,
|
| 713 |
+
"grad_norm": 3.0930943489074707,
|
| 714 |
+
"learning_rate": 2.9999355724193854e-05,
|
| 715 |
+
"loss": 2.0001,
|
| 716 |
+
"step": 1515
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 0.002979447653246819,
|
| 720 |
+
"grad_norm": 2.6339657306671143,
|
| 721 |
+
"learning_rate": 2.9999342903192608e-05,
|
| 722 |
+
"loss": 1.9344,
|
| 723 |
+
"step": 1530
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"epoch": 0.003008657924357082,
|
| 727 |
+
"grad_norm": 3.7254459857940674,
|
| 728 |
+
"learning_rate": 2.999932995587978e-05,
|
| 729 |
+
"loss": 2.0673,
|
| 730 |
+
"step": 1545
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 0.003037868195467345,
|
| 734 |
+
"grad_norm": 2.7078473567962646,
|
| 735 |
+
"learning_rate": 2.999931688225549e-05,
|
| 736 |
+
"loss": 2.0223,
|
| 737 |
+
"step": 1560
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.0030670784665776077,
|
| 741 |
+
"grad_norm": 2.7792718410491943,
|
| 742 |
+
"learning_rate": 2.999930368231984e-05,
|
| 743 |
+
"loss": 1.902,
|
| 744 |
+
"step": 1575
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"epoch": 0.003096288737687871,
|
| 748 |
+
"grad_norm": 4.607854843139648,
|
| 749 |
+
"learning_rate": 2.999929035607294e-05,
|
| 750 |
+
"loss": 1.9536,
|
| 751 |
+
"step": 1590
|
| 752 |
+
},
|
| 753 |
+
{
|
| 754 |
+
"epoch": 0.0031254990087981336,
|
| 755 |
+
"grad_norm": 2.4656715393066406,
|
| 756 |
+
"learning_rate": 2.999927690351491e-05,
|
| 757 |
+
"loss": 2.1792,
|
| 758 |
+
"step": 1605
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 0.0031547092799083967,
|
| 762 |
+
"grad_norm": 2.8231770992279053,
|
| 763 |
+
"learning_rate": 2.9999263324645863e-05,
|
| 764 |
+
"loss": 1.8908,
|
| 765 |
+
"step": 1620
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 0.0031839195510186594,
|
| 769 |
+
"grad_norm": 3.5741395950317383,
|
| 770 |
+
"learning_rate": 2.9999249619465904e-05,
|
| 771 |
+
"loss": 1.9505,
|
| 772 |
+
"step": 1635
|
| 773 |
+
},
|
| 774 |
+
{
|
| 775 |
+
"epoch": 0.0032131298221289226,
|
| 776 |
+
"grad_norm": 2.0066606998443604,
|
| 777 |
+
"learning_rate": 2.999923578797516e-05,
|
| 778 |
+
"loss": 2.0346,
|
| 779 |
+
"step": 1650
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 0.0032423400932391853,
|
| 783 |
+
"grad_norm": 2.5830793380737305,
|
| 784 |
+
"learning_rate": 2.999922183017374e-05,
|
| 785 |
+
"loss": 1.9546,
|
| 786 |
+
"step": 1665
|
| 787 |
+
},
|
| 788 |
+
{
|
| 789 |
+
"epoch": 0.0032715503643494484,
|
| 790 |
+
"grad_norm": 2.8811633586883545,
|
| 791 |
+
"learning_rate": 2.999920774606176e-05,
|
| 792 |
+
"loss": 1.8517,
|
| 793 |
+
"step": 1680
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.003300760635459711,
|
| 797 |
+
"grad_norm": 3.0571601390838623,
|
| 798 |
+
"learning_rate": 2.999919353563935e-05,
|
| 799 |
+
"loss": 1.8746,
|
| 800 |
+
"step": 1695
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 0.0033299709065699743,
|
| 804 |
+
"grad_norm": 2.927371025085449,
|
| 805 |
+
"learning_rate": 2.9999179198906614e-05,
|
| 806 |
+
"loss": 2.0249,
|
| 807 |
+
"step": 1710
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 0.003359181177680237,
|
| 811 |
+
"grad_norm": 2.425579071044922,
|
| 812 |
+
"learning_rate": 2.9999164735863685e-05,
|
| 813 |
+
"loss": 1.9149,
|
| 814 |
+
"step": 1725
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"epoch": 0.0033883914487905,
|
| 818 |
+
"grad_norm": 3.0711071491241455,
|
| 819 |
+
"learning_rate": 2.9999150146510678e-05,
|
| 820 |
+
"loss": 1.8335,
|
| 821 |
+
"step": 1740
|
| 822 |
+
},
|
| 823 |
+
{
|
| 824 |
+
"epoch": 0.003417601719900763,
|
| 825 |
+
"grad_norm": 4.509474277496338,
|
| 826 |
+
"learning_rate": 2.9999135430847718e-05,
|
| 827 |
+
"loss": 2.0002,
|
| 828 |
+
"step": 1755
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"epoch": 0.003446811991011026,
|
| 832 |
+
"grad_norm": 2.784403085708618,
|
| 833 |
+
"learning_rate": 2.9999120588874927e-05,
|
| 834 |
+
"loss": 1.985,
|
| 835 |
+
"step": 1770
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"epoch": 0.0034760222621212887,
|
| 839 |
+
"grad_norm": 4.556781768798828,
|
| 840 |
+
"learning_rate": 2.9999105620592434e-05,
|
| 841 |
+
"loss": 1.8832,
|
| 842 |
+
"step": 1785
|
| 843 |
+
},
|
| 844 |
+
{
|
| 845 |
+
"epoch": 0.003505232533231552,
|
| 846 |
+
"grad_norm": 3.1926143169403076,
|
| 847 |
+
"learning_rate": 2.9999090526000364e-05,
|
| 848 |
+
"loss": 2.0136,
|
| 849 |
+
"step": 1800
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"epoch": 0.0035344428043418146,
|
| 853 |
+
"grad_norm": 3.627634286880493,
|
| 854 |
+
"learning_rate": 2.9999075305098846e-05,
|
| 855 |
+
"loss": 2.0413,
|
| 856 |
+
"step": 1815
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"epoch": 0.0035636530754520777,
|
| 860 |
+
"grad_norm": 4.250141620635986,
|
| 861 |
+
"learning_rate": 2.9999059957887998e-05,
|
| 862 |
+
"loss": 1.9076,
|
| 863 |
+
"step": 1830
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 0.0035928633465623404,
|
| 867 |
+
"grad_norm": 4.263555526733398,
|
| 868 |
+
"learning_rate": 2.999904448436796e-05,
|
| 869 |
+
"loss": 1.9782,
|
| 870 |
+
"step": 1845
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"epoch": 0.0036220736176726036,
|
| 874 |
+
"grad_norm": 2.5871047973632812,
|
| 875 |
+
"learning_rate": 2.999902888453886e-05,
|
| 876 |
+
"loss": 1.9703,
|
| 877 |
+
"step": 1860
|
| 878 |
+
},
|
| 879 |
+
{
|
| 880 |
+
"epoch": 0.0036512838887828663,
|
| 881 |
+
"grad_norm": 2.2850048542022705,
|
| 882 |
+
"learning_rate": 2.9999013158400827e-05,
|
| 883 |
+
"loss": 1.9571,
|
| 884 |
+
"step": 1875
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.0036804941598931294,
|
| 888 |
+
"grad_norm": 2.419032096862793,
|
| 889 |
+
"learning_rate": 2.9998997305953993e-05,
|
| 890 |
+
"loss": 2.1583,
|
| 891 |
+
"step": 1890
|
| 892 |
+
},
|
| 893 |
+
{
|
| 894 |
+
"epoch": 0.003709704431003392,
|
| 895 |
+
"grad_norm": 3.190084934234619,
|
| 896 |
+
"learning_rate": 2.9998981327198497e-05,
|
| 897 |
+
"loss": 1.9732,
|
| 898 |
+
"step": 1905
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 0.0037389147021136553,
|
| 902 |
+
"grad_norm": 3.31540846824646,
|
| 903 |
+
"learning_rate": 2.9998965222134468e-05,
|
| 904 |
+
"loss": 1.8646,
|
| 905 |
+
"step": 1920
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.003768124973223918,
|
| 909 |
+
"grad_norm": 2.423227310180664,
|
| 910 |
+
"learning_rate": 2.9998948990762044e-05,
|
| 911 |
+
"loss": 1.9282,
|
| 912 |
+
"step": 1935
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"epoch": 0.003797335244334181,
|
| 916 |
+
"grad_norm": 3.4635472297668457,
|
| 917 |
+
"learning_rate": 2.999893263308136e-05,
|
| 918 |
+
"loss": 1.9859,
|
| 919 |
+
"step": 1950
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"epoch": 0.003826545515444444,
|
| 923 |
+
"grad_norm": 2.474353790283203,
|
| 924 |
+
"learning_rate": 2.9998916149092556e-05,
|
| 925 |
+
"loss": 2.0055,
|
| 926 |
+
"step": 1965
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 0.003855755786554707,
|
| 930 |
+
"grad_norm": 3.264435052871704,
|
| 931 |
+
"learning_rate": 2.9998899538795766e-05,
|
| 932 |
+
"loss": 1.9755,
|
| 933 |
+
"step": 1980
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"epoch": 0.0038849660576649697,
|
| 937 |
+
"grad_norm": 2.028916358947754,
|
| 938 |
+
"learning_rate": 2.9998882802191137e-05,
|
| 939 |
+
"loss": 1.9914,
|
| 940 |
+
"step": 1995
|
| 941 |
+
}
|
| 942 |
+
],
|
| 943 |
+
"logging_steps": 15,
|
| 944 |
+
"max_steps": 513518,
|
| 945 |
+
"num_input_tokens_seen": 0,
|
| 946 |
+
"num_train_epochs": 1,
|
| 947 |
+
"save_steps": 500,
|
| 948 |
+
"stateful_callbacks": {
|
| 949 |
+
"TrainerControl": {
|
| 950 |
+
"args": {
|
| 951 |
+
"should_epoch_stop": false,
|
| 952 |
+
"should_evaluate": false,
|
| 953 |
+
"should_log": false,
|
| 954 |
+
"should_save": true,
|
| 955 |
+
"should_training_stop": false
|
| 956 |
+
},
|
| 957 |
+
"attributes": {}
|
| 958 |
+
}
|
| 959 |
+
},
|
| 960 |
+
"total_flos": 3130656085370880.0,
|
| 961 |
+
"train_batch_size": 4,
|
| 962 |
+
"trial_name": null,
|
| 963 |
+
"trial_params": null
|
| 964 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2000/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0287fbd96bea9652b04a1bb7ed4b6b1e8de92cccbc4b448c2b5d263a08664d48
|
| 3 |
+
size 5304
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "jonathanjordan21/mos-mamba-6x130m-trainer-sft",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MoSMambaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "jonathanjordan21/mos-mamba-6x130m-hf--configuration_mos_mamba.MoSMambaConfig",
|
| 8 |
+
"AutoModelForCausalLM": "jonathanjordan21/mos-mamba-6x130m-hf--modeling_mos_mamba.MoSMambaForCausalLM"
|
| 9 |
+
},
|
| 10 |
+
"bos_token_id": 0,
|
| 11 |
+
"conv_kernel": 4,
|
| 12 |
+
"d_inner": 1536,
|
| 13 |
+
"d_model": 768,
|
| 14 |
+
"eos_token_id": 50278,
|
| 15 |
+
"expand": 2,
|
| 16 |
+
"fused_add_norm": true,
|
| 17 |
+
"hidden_act": "silu",
|
| 18 |
+
"hidden_size": 768,
|
| 19 |
+
"initializer_range": 0.1,
|
| 20 |
+
"intermediate_size": 1536,
|
| 21 |
+
"layer_norm_epsilon": 1e-05,
|
| 22 |
+
"model_type": "MoSMamba",
|
| 23 |
+
"n_layer": 24,
|
| 24 |
+
"num_hidden_layers": 24,
|
| 25 |
+
"num_selectivities": 6,
|
| 26 |
+
"num_selectivities_per_tok": 2,
|
| 27 |
+
"output_router_logits": true,
|
| 28 |
+
"pad_token_id": 0,
|
| 29 |
+
"pad_vocab_size_multiple": 8,
|
| 30 |
+
"rescale_prenorm_residual": false,
|
| 31 |
+
"residual_in_fp32": true,
|
| 32 |
+
"rms_norm": true,
|
| 33 |
+
"ssm_cfg": {},
|
| 34 |
+
"state_size": 16,
|
| 35 |
+
"time_step_floor": 0.0001,
|
| 36 |
+
"time_step_init_scheme": "random",
|
| 37 |
+
"time_step_max": 0.1,
|
| 38 |
+
"time_step_min": 0.001,
|
| 39 |
+
"time_step_rank": 48,
|
| 40 |
+
"time_step_scale": 1.0,
|
| 41 |
+
"torch_dtype": "float32",
|
| 42 |
+
"transformers_version": "4.42.4",
|
| 43 |
+
"use_bias": false,
|
| 44 |
+
"use_cache": true,
|
| 45 |
+
"use_conv_bias": true,
|
| 46 |
+
"vocab_size": 50280
|
| 47 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 50278,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.42.4"
|
| 7 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fb904cd60075b474e59a133ceb6ff998ab4a84f6ee47142bc4525c7b58c7cbfd
|
| 3 |
+
size 576008736
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:805d52d3e5ab04d565b3b482ed9b94de9cc50129e061a151e7dd76c6ac80e1bf
|
| 3 |
+
size 1152256984
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cda3aa960c48aa29a55f9fbb6530c088667fac02b2c205ec065408bdbd5c21bc
|
| 3 |
+
size 14244
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:419a24c9a8e40edb958fec8adda5f5d49a9532e5482286127bbc1aa188bfd070
|
| 3 |
+
size 1064
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/special_tokens_map.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<|endoftext|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "<|im_end|>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": false,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": {
|
| 17 |
+
"content": "<|endoftext|>",
|
| 18 |
+
"lstrip": false,
|
| 19 |
+
"normalized": false,
|
| 20 |
+
"rstrip": false,
|
| 21 |
+
"single_word": false
|
| 22 |
+
},
|
| 23 |
+
"unk_token": {
|
| 24 |
+
"content": "<|endoftext|>",
|
| 25 |
+
"lstrip": false,
|
| 26 |
+
"normalized": false,
|
| 27 |
+
"rstrip": false,
|
| 28 |
+
"single_word": false
|
| 29 |
+
}
|
| 30 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/tokenizer_config.json
ADDED
|
@@ -0,0 +1,235 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"add_prefix_space": false,
|
| 5 |
+
"added_tokens_decoder": {
|
| 6 |
+
"0": {
|
| 7 |
+
"content": "<|endoftext|>",
|
| 8 |
+
"lstrip": false,
|
| 9 |
+
"normalized": false,
|
| 10 |
+
"rstrip": false,
|
| 11 |
+
"single_word": false,
|
| 12 |
+
"special": true
|
| 13 |
+
},
|
| 14 |
+
"1": {
|
| 15 |
+
"content": "<|padding|>",
|
| 16 |
+
"lstrip": false,
|
| 17 |
+
"normalized": false,
|
| 18 |
+
"rstrip": false,
|
| 19 |
+
"single_word": false,
|
| 20 |
+
"special": true
|
| 21 |
+
},
|
| 22 |
+
"50254": {
|
| 23 |
+
"content": " ",
|
| 24 |
+
"lstrip": false,
|
| 25 |
+
"normalized": true,
|
| 26 |
+
"rstrip": false,
|
| 27 |
+
"single_word": false,
|
| 28 |
+
"special": false
|
| 29 |
+
},
|
| 30 |
+
"50255": {
|
| 31 |
+
"content": " ",
|
| 32 |
+
"lstrip": false,
|
| 33 |
+
"normalized": true,
|
| 34 |
+
"rstrip": false,
|
| 35 |
+
"single_word": false,
|
| 36 |
+
"special": false
|
| 37 |
+
},
|
| 38 |
+
"50256": {
|
| 39 |
+
"content": " ",
|
| 40 |
+
"lstrip": false,
|
| 41 |
+
"normalized": true,
|
| 42 |
+
"rstrip": false,
|
| 43 |
+
"single_word": false,
|
| 44 |
+
"special": false
|
| 45 |
+
},
|
| 46 |
+
"50257": {
|
| 47 |
+
"content": " ",
|
| 48 |
+
"lstrip": false,
|
| 49 |
+
"normalized": true,
|
| 50 |
+
"rstrip": false,
|
| 51 |
+
"single_word": false,
|
| 52 |
+
"special": false
|
| 53 |
+
},
|
| 54 |
+
"50258": {
|
| 55 |
+
"content": " ",
|
| 56 |
+
"lstrip": false,
|
| 57 |
+
"normalized": true,
|
| 58 |
+
"rstrip": false,
|
| 59 |
+
"single_word": false,
|
| 60 |
+
"special": false
|
| 61 |
+
},
|
| 62 |
+
"50259": {
|
| 63 |
+
"content": " ",
|
| 64 |
+
"lstrip": false,
|
| 65 |
+
"normalized": true,
|
| 66 |
+
"rstrip": false,
|
| 67 |
+
"single_word": false,
|
| 68 |
+
"special": false
|
| 69 |
+
},
|
| 70 |
+
"50260": {
|
| 71 |
+
"content": " ",
|
| 72 |
+
"lstrip": false,
|
| 73 |
+
"normalized": true,
|
| 74 |
+
"rstrip": false,
|
| 75 |
+
"single_word": false,
|
| 76 |
+
"special": false
|
| 77 |
+
},
|
| 78 |
+
"50261": {
|
| 79 |
+
"content": " ",
|
| 80 |
+
"lstrip": false,
|
| 81 |
+
"normalized": true,
|
| 82 |
+
"rstrip": false,
|
| 83 |
+
"single_word": false,
|
| 84 |
+
"special": false
|
| 85 |
+
},
|
| 86 |
+
"50262": {
|
| 87 |
+
"content": " ",
|
| 88 |
+
"lstrip": false,
|
| 89 |
+
"normalized": true,
|
| 90 |
+
"rstrip": false,
|
| 91 |
+
"single_word": false,
|
| 92 |
+
"special": false
|
| 93 |
+
},
|
| 94 |
+
"50263": {
|
| 95 |
+
"content": " ",
|
| 96 |
+
"lstrip": false,
|
| 97 |
+
"normalized": true,
|
| 98 |
+
"rstrip": false,
|
| 99 |
+
"single_word": false,
|
| 100 |
+
"special": false
|
| 101 |
+
},
|
| 102 |
+
"50264": {
|
| 103 |
+
"content": " ",
|
| 104 |
+
"lstrip": false,
|
| 105 |
+
"normalized": true,
|
| 106 |
+
"rstrip": false,
|
| 107 |
+
"single_word": false,
|
| 108 |
+
"special": false
|
| 109 |
+
},
|
| 110 |
+
"50265": {
|
| 111 |
+
"content": " ",
|
| 112 |
+
"lstrip": false,
|
| 113 |
+
"normalized": true,
|
| 114 |
+
"rstrip": false,
|
| 115 |
+
"single_word": false,
|
| 116 |
+
"special": false
|
| 117 |
+
},
|
| 118 |
+
"50266": {
|
| 119 |
+
"content": " ",
|
| 120 |
+
"lstrip": false,
|
| 121 |
+
"normalized": true,
|
| 122 |
+
"rstrip": false,
|
| 123 |
+
"single_word": false,
|
| 124 |
+
"special": false
|
| 125 |
+
},
|
| 126 |
+
"50267": {
|
| 127 |
+
"content": " ",
|
| 128 |
+
"lstrip": false,
|
| 129 |
+
"normalized": true,
|
| 130 |
+
"rstrip": false,
|
| 131 |
+
"single_word": false,
|
| 132 |
+
"special": false
|
| 133 |
+
},
|
| 134 |
+
"50268": {
|
| 135 |
+
"content": " ",
|
| 136 |
+
"lstrip": false,
|
| 137 |
+
"normalized": true,
|
| 138 |
+
"rstrip": false,
|
| 139 |
+
"single_word": false,
|
| 140 |
+
"special": false
|
| 141 |
+
},
|
| 142 |
+
"50269": {
|
| 143 |
+
"content": " ",
|
| 144 |
+
"lstrip": false,
|
| 145 |
+
"normalized": true,
|
| 146 |
+
"rstrip": false,
|
| 147 |
+
"single_word": false,
|
| 148 |
+
"special": false
|
| 149 |
+
},
|
| 150 |
+
"50270": {
|
| 151 |
+
"content": " ",
|
| 152 |
+
"lstrip": false,
|
| 153 |
+
"normalized": true,
|
| 154 |
+
"rstrip": false,
|
| 155 |
+
"single_word": false,
|
| 156 |
+
"special": false
|
| 157 |
+
},
|
| 158 |
+
"50271": {
|
| 159 |
+
"content": " ",
|
| 160 |
+
"lstrip": false,
|
| 161 |
+
"normalized": true,
|
| 162 |
+
"rstrip": false,
|
| 163 |
+
"single_word": false,
|
| 164 |
+
"special": false
|
| 165 |
+
},
|
| 166 |
+
"50272": {
|
| 167 |
+
"content": " ",
|
| 168 |
+
"lstrip": false,
|
| 169 |
+
"normalized": true,
|
| 170 |
+
"rstrip": false,
|
| 171 |
+
"single_word": false,
|
| 172 |
+
"special": false
|
| 173 |
+
},
|
| 174 |
+
"50273": {
|
| 175 |
+
"content": " ",
|
| 176 |
+
"lstrip": false,
|
| 177 |
+
"normalized": true,
|
| 178 |
+
"rstrip": false,
|
| 179 |
+
"single_word": false,
|
| 180 |
+
"special": false
|
| 181 |
+
},
|
| 182 |
+
"50274": {
|
| 183 |
+
"content": " ",
|
| 184 |
+
"lstrip": false,
|
| 185 |
+
"normalized": true,
|
| 186 |
+
"rstrip": false,
|
| 187 |
+
"single_word": false,
|
| 188 |
+
"special": false
|
| 189 |
+
},
|
| 190 |
+
"50275": {
|
| 191 |
+
"content": " ",
|
| 192 |
+
"lstrip": false,
|
| 193 |
+
"normalized": true,
|
| 194 |
+
"rstrip": false,
|
| 195 |
+
"single_word": false,
|
| 196 |
+
"special": false
|
| 197 |
+
},
|
| 198 |
+
"50276": {
|
| 199 |
+
"content": " ",
|
| 200 |
+
"lstrip": false,
|
| 201 |
+
"normalized": true,
|
| 202 |
+
"rstrip": false,
|
| 203 |
+
"single_word": false,
|
| 204 |
+
"special": false
|
| 205 |
+
},
|
| 206 |
+
"50277": {
|
| 207 |
+
"content": "<|im_start|>",
|
| 208 |
+
"lstrip": false,
|
| 209 |
+
"normalized": false,
|
| 210 |
+
"rstrip": false,
|
| 211 |
+
"single_word": false,
|
| 212 |
+
"special": true
|
| 213 |
+
},
|
| 214 |
+
"50278": {
|
| 215 |
+
"content": "<|im_end|>",
|
| 216 |
+
"lstrip": false,
|
| 217 |
+
"normalized": false,
|
| 218 |
+
"rstrip": false,
|
| 219 |
+
"single_word": false,
|
| 220 |
+
"special": true
|
| 221 |
+
}
|
| 222 |
+
},
|
| 223 |
+
"bos_token": "<|endoftext|>",
|
| 224 |
+
"chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
|
| 225 |
+
"clean_up_tokenization_spaces": true,
|
| 226 |
+
"eos_token": "<|im_end|>",
|
| 227 |
+
"max_length": 1024,
|
| 228 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 229 |
+
"pad_token": "<|endoftext|>",
|
| 230 |
+
"stride": 0,
|
| 231 |
+
"tokenizer_class": "GPTNeoXTokenizer",
|
| 232 |
+
"truncation_side": "right",
|
| 233 |
+
"truncation_strategy": "longest_first",
|
| 234 |
+
"unk_token": "<|endoftext|>"
|
| 235 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/trainer_state.json
ADDED
|
@@ -0,0 +1,1195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.004868378518377155,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2500,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 2.9210271110262932e-05,
|
| 13 |
+
"grad_norm": 5.397985935211182,
|
| 14 |
+
"learning_rate": 2.9999999936841446e-05,
|
| 15 |
+
"loss": 2.1156,
|
| 16 |
+
"step": 15
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 5.8420542220525864e-05,
|
| 20 |
+
"grad_norm": 3.999190092086792,
|
| 21 |
+
"learning_rate": 2.999999974736578e-05,
|
| 22 |
+
"loss": 2.5268,
|
| 23 |
+
"step": 30
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 8.76308133307888e-05,
|
| 27 |
+
"grad_norm": 5.580333709716797,
|
| 28 |
+
"learning_rate": 2.999999943157301e-05,
|
| 29 |
+
"loss": 2.3955,
|
| 30 |
+
"step": 45
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.00011684108444105173,
|
| 34 |
+
"grad_norm": 3.6752560138702393,
|
| 35 |
+
"learning_rate": 2.9999998989463132e-05,
|
| 36 |
+
"loss": 2.171,
|
| 37 |
+
"step": 60
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.00014605135555131467,
|
| 41 |
+
"grad_norm": 2.655461072921753,
|
| 42 |
+
"learning_rate": 2.9999998421036153e-05,
|
| 43 |
+
"loss": 2.188,
|
| 44 |
+
"step": 75
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.0001752616266615776,
|
| 48 |
+
"grad_norm": 2.572371006011963,
|
| 49 |
+
"learning_rate": 2.9999997726292083e-05,
|
| 50 |
+
"loss": 2.1977,
|
| 51 |
+
"step": 90
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.00020447189777184053,
|
| 55 |
+
"grad_norm": 8.544811248779297,
|
| 56 |
+
"learning_rate": 2.999999690523092e-05,
|
| 57 |
+
"loss": 2.1871,
|
| 58 |
+
"step": 105
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.00023368216888210345,
|
| 62 |
+
"grad_norm": 3.530529737472534,
|
| 63 |
+
"learning_rate": 2.999999595785267e-05,
|
| 64 |
+
"loss": 2.2238,
|
| 65 |
+
"step": 120
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.00026289243999236636,
|
| 69 |
+
"grad_norm": 3.671182155609131,
|
| 70 |
+
"learning_rate": 2.9999994884157345e-05,
|
| 71 |
+
"loss": 2.1338,
|
| 72 |
+
"step": 135
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.00029210271110262934,
|
| 76 |
+
"grad_norm": 3.9643137454986572,
|
| 77 |
+
"learning_rate": 2.9999993684144956e-05,
|
| 78 |
+
"loss": 2.0788,
|
| 79 |
+
"step": 150
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.00032131298221289227,
|
| 83 |
+
"grad_norm": 2.62164044380188,
|
| 84 |
+
"learning_rate": 2.9999992357815508e-05,
|
| 85 |
+
"loss": 2.117,
|
| 86 |
+
"step": 165
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.0003505232533231552,
|
| 90 |
+
"grad_norm": 4.189517021179199,
|
| 91 |
+
"learning_rate": 2.999999090516902e-05,
|
| 92 |
+
"loss": 2.262,
|
| 93 |
+
"step": 180
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.0003797335244334181,
|
| 97 |
+
"grad_norm": 2.880622625350952,
|
| 98 |
+
"learning_rate": 2.9999989326205494e-05,
|
| 99 |
+
"loss": 2.1102,
|
| 100 |
+
"step": 195
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.00040894379554368105,
|
| 104 |
+
"grad_norm": 5.852839469909668,
|
| 105 |
+
"learning_rate": 2.9999987620924948e-05,
|
| 106 |
+
"loss": 2.0684,
|
| 107 |
+
"step": 210
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.000438154066653944,
|
| 111 |
+
"grad_norm": 3.6339924335479736,
|
| 112 |
+
"learning_rate": 2.9999985789327394e-05,
|
| 113 |
+
"loss": 2.2518,
|
| 114 |
+
"step": 225
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.0004673643377642069,
|
| 118 |
+
"grad_norm": 2.743267297744751,
|
| 119 |
+
"learning_rate": 2.9999983831412858e-05,
|
| 120 |
+
"loss": 2.1236,
|
| 121 |
+
"step": 240
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.0004965746088744698,
|
| 125 |
+
"grad_norm": 3.9490363597869873,
|
| 126 |
+
"learning_rate": 2.9999981747181345e-05,
|
| 127 |
+
"loss": 2.0878,
|
| 128 |
+
"step": 255
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.0005257848799847327,
|
| 132 |
+
"grad_norm": 3.0342047214508057,
|
| 133 |
+
"learning_rate": 2.9999979536632872e-05,
|
| 134 |
+
"loss": 2.1415,
|
| 135 |
+
"step": 270
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.0005549951510949957,
|
| 139 |
+
"grad_norm": 4.045855522155762,
|
| 140 |
+
"learning_rate": 2.9999977199767467e-05,
|
| 141 |
+
"loss": 2.0558,
|
| 142 |
+
"step": 285
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"epoch": 0.0005842054222052587,
|
| 146 |
+
"grad_norm": 4.325068950653076,
|
| 147 |
+
"learning_rate": 2.999997473658514e-05,
|
| 148 |
+
"loss": 2.0617,
|
| 149 |
+
"step": 300
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.0006134156933155216,
|
| 153 |
+
"grad_norm": 4.770871162414551,
|
| 154 |
+
"learning_rate": 2.999997214708592e-05,
|
| 155 |
+
"loss": 2.0922,
|
| 156 |
+
"step": 315
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 0.0006426259644257845,
|
| 160 |
+
"grad_norm": 4.1802754402160645,
|
| 161 |
+
"learning_rate": 2.999996943126982e-05,
|
| 162 |
+
"loss": 2.217,
|
| 163 |
+
"step": 330
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.0006718362355360475,
|
| 167 |
+
"grad_norm": 3.044260263442993,
|
| 168 |
+
"learning_rate": 2.999996658913687e-05,
|
| 169 |
+
"loss": 2.0508,
|
| 170 |
+
"step": 345
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.0007010465066463104,
|
| 174 |
+
"grad_norm": 4.1525187492370605,
|
| 175 |
+
"learning_rate": 2.9999963620687095e-05,
|
| 176 |
+
"loss": 2.1678,
|
| 177 |
+
"step": 360
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.0007302567777565733,
|
| 181 |
+
"grad_norm": 2.5676231384277344,
|
| 182 |
+
"learning_rate": 2.9999960525920515e-05,
|
| 183 |
+
"loss": 2.1693,
|
| 184 |
+
"step": 375
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 0.0007594670488668362,
|
| 188 |
+
"grad_norm": 3.880307674407959,
|
| 189 |
+
"learning_rate": 2.9999957304837156e-05,
|
| 190 |
+
"loss": 1.9008,
|
| 191 |
+
"step": 390
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.0007886773199770992,
|
| 195 |
+
"grad_norm": 3.3433754444122314,
|
| 196 |
+
"learning_rate": 2.999995395743705e-05,
|
| 197 |
+
"loss": 2.0734,
|
| 198 |
+
"step": 405
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.0008178875910873621,
|
| 202 |
+
"grad_norm": 3.451378583908081,
|
| 203 |
+
"learning_rate": 2.999995048372022e-05,
|
| 204 |
+
"loss": 1.9901,
|
| 205 |
+
"step": 420
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.000847097862197625,
|
| 209 |
+
"grad_norm": 3.4411580562591553,
|
| 210 |
+
"learning_rate": 2.9999946883686695e-05,
|
| 211 |
+
"loss": 1.9734,
|
| 212 |
+
"step": 435
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.000876308133307888,
|
| 216 |
+
"grad_norm": 3.514651298522949,
|
| 217 |
+
"learning_rate": 2.999994315733651e-05,
|
| 218 |
+
"loss": 2.1632,
|
| 219 |
+
"step": 450
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.0009055184044181509,
|
| 223 |
+
"grad_norm": 5.380824565887451,
|
| 224 |
+
"learning_rate": 2.999993930466969e-05,
|
| 225 |
+
"loss": 2.1534,
|
| 226 |
+
"step": 465
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.0009347286755284138,
|
| 230 |
+
"grad_norm": 3.768618106842041,
|
| 231 |
+
"learning_rate": 2.999993532568628e-05,
|
| 232 |
+
"loss": 2.2182,
|
| 233 |
+
"step": 480
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.0009639389466386767,
|
| 237 |
+
"grad_norm": 2.9490509033203125,
|
| 238 |
+
"learning_rate": 2.99999312203863e-05,
|
| 239 |
+
"loss": 1.9677,
|
| 240 |
+
"step": 495
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 0.0009931492177489396,
|
| 244 |
+
"grad_norm": 3.4881184101104736,
|
| 245 |
+
"learning_rate": 2.9999926988769788e-05,
|
| 246 |
+
"loss": 2.0786,
|
| 247 |
+
"step": 510
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 0.0010223594888592025,
|
| 251 |
+
"grad_norm": 3.788675546646118,
|
| 252 |
+
"learning_rate": 2.9999922630836784e-05,
|
| 253 |
+
"loss": 1.9696,
|
| 254 |
+
"step": 525
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 0.0010515697599694654,
|
| 258 |
+
"grad_norm": 3.070878744125366,
|
| 259 |
+
"learning_rate": 2.999991814658732e-05,
|
| 260 |
+
"loss": 1.9053,
|
| 261 |
+
"step": 540
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 0.0010807800310797286,
|
| 265 |
+
"grad_norm": 2.456150531768799,
|
| 266 |
+
"learning_rate": 2.9999913536021436e-05,
|
| 267 |
+
"loss": 1.9813,
|
| 268 |
+
"step": 555
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 0.0011099903021899915,
|
| 272 |
+
"grad_norm": 4.81670618057251,
|
| 273 |
+
"learning_rate": 2.999990879913917e-05,
|
| 274 |
+
"loss": 1.9317,
|
| 275 |
+
"step": 570
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 0.0011392005733002544,
|
| 279 |
+
"grad_norm": 4.014492988586426,
|
| 280 |
+
"learning_rate": 2.9999903935940567e-05,
|
| 281 |
+
"loss": 1.9709,
|
| 282 |
+
"step": 585
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.0011684108444105174,
|
| 286 |
+
"grad_norm": 4.765859603881836,
|
| 287 |
+
"learning_rate": 2.999989894642566e-05,
|
| 288 |
+
"loss": 1.9252,
|
| 289 |
+
"step": 600
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 0.0011976211155207803,
|
| 293 |
+
"grad_norm": 3.5599727630615234,
|
| 294 |
+
"learning_rate": 2.9999893830594492e-05,
|
| 295 |
+
"loss": 2.0175,
|
| 296 |
+
"step": 615
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 0.0012268313866310432,
|
| 300 |
+
"grad_norm": 4.441170692443848,
|
| 301 |
+
"learning_rate": 2.999988858844711e-05,
|
| 302 |
+
"loss": 2.1166,
|
| 303 |
+
"step": 630
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"epoch": 0.0012560416577413061,
|
| 307 |
+
"grad_norm": 2.691765308380127,
|
| 308 |
+
"learning_rate": 2.9999883219983558e-05,
|
| 309 |
+
"loss": 1.991,
|
| 310 |
+
"step": 645
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.001285251928851569,
|
| 314 |
+
"grad_norm": 4.039743423461914,
|
| 315 |
+
"learning_rate": 2.9999877725203878e-05,
|
| 316 |
+
"loss": 1.9917,
|
| 317 |
+
"step": 660
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 0.001314462199961832,
|
| 321 |
+
"grad_norm": 3.0612881183624268,
|
| 322 |
+
"learning_rate": 2.9999872104108115e-05,
|
| 323 |
+
"loss": 2.0436,
|
| 324 |
+
"step": 675
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.001343672471072095,
|
| 328 |
+
"grad_norm": 3.3579518795013428,
|
| 329 |
+
"learning_rate": 2.9999866356696326e-05,
|
| 330 |
+
"loss": 1.9659,
|
| 331 |
+
"step": 690
|
| 332 |
+
},
|
| 333 |
+
{
|
| 334 |
+
"epoch": 0.0013728827421823579,
|
| 335 |
+
"grad_norm": 4.145946502685547,
|
| 336 |
+
"learning_rate": 2.9999860482968552e-05,
|
| 337 |
+
"loss": 2.0267,
|
| 338 |
+
"step": 705
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 0.0014020930132926208,
|
| 342 |
+
"grad_norm": 5.213028430938721,
|
| 343 |
+
"learning_rate": 2.999985448292484e-05,
|
| 344 |
+
"loss": 2.0953,
|
| 345 |
+
"step": 720
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.0014313032844028837,
|
| 349 |
+
"grad_norm": 2.440793752670288,
|
| 350 |
+
"learning_rate": 2.9999848356565246e-05,
|
| 351 |
+
"loss": 2.0218,
|
| 352 |
+
"step": 735
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.0014605135555131466,
|
| 356 |
+
"grad_norm": 2.6167280673980713,
|
| 357 |
+
"learning_rate": 2.9999842103889813e-05,
|
| 358 |
+
"loss": 2.0638,
|
| 359 |
+
"step": 750
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 0.0014897238266234096,
|
| 363 |
+
"grad_norm": 2.2680811882019043,
|
| 364 |
+
"learning_rate": 2.999983572489861e-05,
|
| 365 |
+
"loss": 1.973,
|
| 366 |
+
"step": 765
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.0015189340977336725,
|
| 370 |
+
"grad_norm": 2.905947208404541,
|
| 371 |
+
"learning_rate": 2.999982921959167e-05,
|
| 372 |
+
"loss": 1.9346,
|
| 373 |
+
"step": 780
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.0015481443688439354,
|
| 377 |
+
"grad_norm": 4.673079967498779,
|
| 378 |
+
"learning_rate": 2.999982258796907e-05,
|
| 379 |
+
"loss": 2.1666,
|
| 380 |
+
"step": 795
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 0.0015773546399541984,
|
| 384 |
+
"grad_norm": 2.5963408946990967,
|
| 385 |
+
"learning_rate": 2.9999815830030846e-05,
|
| 386 |
+
"loss": 2.1063,
|
| 387 |
+
"step": 810
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 0.0016065649110644613,
|
| 391 |
+
"grad_norm": 3.656632661819458,
|
| 392 |
+
"learning_rate": 2.9999808945777066e-05,
|
| 393 |
+
"loss": 2.1328,
|
| 394 |
+
"step": 825
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 0.0016357751821747242,
|
| 398 |
+
"grad_norm": 4.4109368324279785,
|
| 399 |
+
"learning_rate": 2.9999801935207786e-05,
|
| 400 |
+
"loss": 1.9602,
|
| 401 |
+
"step": 840
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.0016649854532849871,
|
| 405 |
+
"grad_norm": 2.10768985748291,
|
| 406 |
+
"learning_rate": 2.9999794798323065e-05,
|
| 407 |
+
"loss": 1.9766,
|
| 408 |
+
"step": 855
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"epoch": 0.00169419572439525,
|
| 412 |
+
"grad_norm": 4.301672458648682,
|
| 413 |
+
"learning_rate": 2.999978753512296e-05,
|
| 414 |
+
"loss": 1.9386,
|
| 415 |
+
"step": 870
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 0.001723405995505513,
|
| 419 |
+
"grad_norm": 8.171302795410156,
|
| 420 |
+
"learning_rate": 2.9999780145607538e-05,
|
| 421 |
+
"loss": 2.1288,
|
| 422 |
+
"step": 885
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 0.001752616266615776,
|
| 426 |
+
"grad_norm": 3.4784069061279297,
|
| 427 |
+
"learning_rate": 2.999977262977685e-05,
|
| 428 |
+
"loss": 2.1346,
|
| 429 |
+
"step": 900
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.0017818265377260389,
|
| 433 |
+
"grad_norm": 2.9945054054260254,
|
| 434 |
+
"learning_rate": 2.9999764987630976e-05,
|
| 435 |
+
"loss": 2.0432,
|
| 436 |
+
"step": 915
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"epoch": 0.0018110368088363018,
|
| 440 |
+
"grad_norm": 3.632039785385132,
|
| 441 |
+
"learning_rate": 2.9999757219169964e-05,
|
| 442 |
+
"loss": 2.0642,
|
| 443 |
+
"step": 930
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.0018402470799465647,
|
| 447 |
+
"grad_norm": 3.9847047328948975,
|
| 448 |
+
"learning_rate": 2.999974932439389e-05,
|
| 449 |
+
"loss": 2.016,
|
| 450 |
+
"step": 945
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"epoch": 0.0018694573510568276,
|
| 454 |
+
"grad_norm": 1.89451003074646,
|
| 455 |
+
"learning_rate": 2.9999741303302816e-05,
|
| 456 |
+
"loss": 2.0355,
|
| 457 |
+
"step": 960
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.0018986676221670906,
|
| 461 |
+
"grad_norm": 6.945189952850342,
|
| 462 |
+
"learning_rate": 2.9999733155896814e-05,
|
| 463 |
+
"loss": 1.97,
|
| 464 |
+
"step": 975
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.0019278778932773535,
|
| 468 |
+
"grad_norm": 3.347123384475708,
|
| 469 |
+
"learning_rate": 2.9999724882175947e-05,
|
| 470 |
+
"loss": 2.0421,
|
| 471 |
+
"step": 990
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"epoch": 0.0019570881643876164,
|
| 475 |
+
"grad_norm": 4.331577777862549,
|
| 476 |
+
"learning_rate": 2.9999716482140292e-05,
|
| 477 |
+
"loss": 2.133,
|
| 478 |
+
"step": 1005
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"epoch": 0.001986298435497879,
|
| 482 |
+
"grad_norm": 4.221049785614014,
|
| 483 |
+
"learning_rate": 2.999970795578991e-05,
|
| 484 |
+
"loss": 1.9272,
|
| 485 |
+
"step": 1020
|
| 486 |
+
},
|
| 487 |
+
{
|
| 488 |
+
"epoch": 0.0020155087066081423,
|
| 489 |
+
"grad_norm": 2.820549488067627,
|
| 490 |
+
"learning_rate": 2.9999699303124876e-05,
|
| 491 |
+
"loss": 2.0116,
|
| 492 |
+
"step": 1035
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"epoch": 0.002044718977718405,
|
| 496 |
+
"grad_norm": 2.2391343116760254,
|
| 497 |
+
"learning_rate": 2.9999690524145268e-05,
|
| 498 |
+
"loss": 1.9337,
|
| 499 |
+
"step": 1050
|
| 500 |
+
},
|
| 501 |
+
{
|
| 502 |
+
"epoch": 0.002073929248828668,
|
| 503 |
+
"grad_norm": 2.285027503967285,
|
| 504 |
+
"learning_rate": 2.9999681618851156e-05,
|
| 505 |
+
"loss": 1.9773,
|
| 506 |
+
"step": 1065
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.002103139519938931,
|
| 510 |
+
"grad_norm": 3.266909599304199,
|
| 511 |
+
"learning_rate": 2.9999672587242616e-05,
|
| 512 |
+
"loss": 1.9471,
|
| 513 |
+
"step": 1080
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"epoch": 0.002132349791049194,
|
| 517 |
+
"grad_norm": 5.03222131729126,
|
| 518 |
+
"learning_rate": 2.9999663429319722e-05,
|
| 519 |
+
"loss": 2.1337,
|
| 520 |
+
"step": 1095
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 0.002161560062159457,
|
| 524 |
+
"grad_norm": 2.044246196746826,
|
| 525 |
+
"learning_rate": 2.9999654145082552e-05,
|
| 526 |
+
"loss": 2.06,
|
| 527 |
+
"step": 1110
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 0.00219077033326972,
|
| 531 |
+
"grad_norm": 3.1399905681610107,
|
| 532 |
+
"learning_rate": 2.9999644734531183e-05,
|
| 533 |
+
"loss": 1.9174,
|
| 534 |
+
"step": 1125
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"epoch": 0.002219980604379983,
|
| 538 |
+
"grad_norm": 3.2445454597473145,
|
| 539 |
+
"learning_rate": 2.99996351976657e-05,
|
| 540 |
+
"loss": 1.9728,
|
| 541 |
+
"step": 1140
|
| 542 |
+
},
|
| 543 |
+
{
|
| 544 |
+
"epoch": 0.0022491908754902457,
|
| 545 |
+
"grad_norm": 2.3682138919830322,
|
| 546 |
+
"learning_rate": 2.999962553448618e-05,
|
| 547 |
+
"loss": 1.9779,
|
| 548 |
+
"step": 1155
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 0.002278401146600509,
|
| 552 |
+
"grad_norm": 2.283808708190918,
|
| 553 |
+
"learning_rate": 2.99996157449927e-05,
|
| 554 |
+
"loss": 1.964,
|
| 555 |
+
"step": 1170
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.0023076114177107716,
|
| 559 |
+
"grad_norm": 3.048382043838501,
|
| 560 |
+
"learning_rate": 2.9999605829185353e-05,
|
| 561 |
+
"loss": 2.0908,
|
| 562 |
+
"step": 1185
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"epoch": 0.0023368216888210347,
|
| 566 |
+
"grad_norm": 6.139120101928711,
|
| 567 |
+
"learning_rate": 2.999959578706421e-05,
|
| 568 |
+
"loss": 1.9535,
|
| 569 |
+
"step": 1200
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.0023660319599312974,
|
| 573 |
+
"grad_norm": 2.104069232940674,
|
| 574 |
+
"learning_rate": 2.9999585618629363e-05,
|
| 575 |
+
"loss": 1.8775,
|
| 576 |
+
"step": 1215
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 0.0023952422310415606,
|
| 580 |
+
"grad_norm": 2.531449317932129,
|
| 581 |
+
"learning_rate": 2.9999575323880894e-05,
|
| 582 |
+
"loss": 2.0309,
|
| 583 |
+
"step": 1230
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 0.0024244525021518233,
|
| 587 |
+
"grad_norm": 2.841078996658325,
|
| 588 |
+
"learning_rate": 2.9999564902818894e-05,
|
| 589 |
+
"loss": 2.0546,
|
| 590 |
+
"step": 1245
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 0.0024536627732620864,
|
| 594 |
+
"grad_norm": 2.0627243518829346,
|
| 595 |
+
"learning_rate": 2.9999554355443445e-05,
|
| 596 |
+
"loss": 1.8833,
|
| 597 |
+
"step": 1260
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.002482873044372349,
|
| 601 |
+
"grad_norm": 3.4014151096343994,
|
| 602 |
+
"learning_rate": 2.9999543681754645e-05,
|
| 603 |
+
"loss": 1.8935,
|
| 604 |
+
"step": 1275
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"epoch": 0.0025120833154826123,
|
| 608 |
+
"grad_norm": 3.507380962371826,
|
| 609 |
+
"learning_rate": 2.999953288175257e-05,
|
| 610 |
+
"loss": 1.9512,
|
| 611 |
+
"step": 1290
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 0.002541293586592875,
|
| 615 |
+
"grad_norm": 5.4273362159729,
|
| 616 |
+
"learning_rate": 2.9999521955437325e-05,
|
| 617 |
+
"loss": 1.9794,
|
| 618 |
+
"step": 1305
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 0.002570503857703138,
|
| 622 |
+
"grad_norm": 4.466185092926025,
|
| 623 |
+
"learning_rate": 2.9999510902808993e-05,
|
| 624 |
+
"loss": 1.8266,
|
| 625 |
+
"step": 1320
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"epoch": 0.002599714128813401,
|
| 629 |
+
"grad_norm": 2.4276440143585205,
|
| 630 |
+
"learning_rate": 2.9999499723867672e-05,
|
| 631 |
+
"loss": 2.0149,
|
| 632 |
+
"step": 1335
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 0.002628924399923664,
|
| 636 |
+
"grad_norm": 3.589876413345337,
|
| 637 |
+
"learning_rate": 2.9999488418613454e-05,
|
| 638 |
+
"loss": 1.8865,
|
| 639 |
+
"step": 1350
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 0.0026581346710339267,
|
| 643 |
+
"grad_norm": 3.750687599182129,
|
| 644 |
+
"learning_rate": 2.9999476987046434e-05,
|
| 645 |
+
"loss": 2.13,
|
| 646 |
+
"step": 1365
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.00268734494214419,
|
| 650 |
+
"grad_norm": 3.638850212097168,
|
| 651 |
+
"learning_rate": 2.999946542916671e-05,
|
| 652 |
+
"loss": 1.981,
|
| 653 |
+
"step": 1380
|
| 654 |
+
},
|
| 655 |
+
{
|
| 656 |
+
"epoch": 0.0027165552132544526,
|
| 657 |
+
"grad_norm": 4.610542297363281,
|
| 658 |
+
"learning_rate": 2.9999453744974375e-05,
|
| 659 |
+
"loss": 1.8887,
|
| 660 |
+
"step": 1395
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 0.0027457654843647157,
|
| 664 |
+
"grad_norm": 2.3758933544158936,
|
| 665 |
+
"learning_rate": 2.9999441934469534e-05,
|
| 666 |
+
"loss": 1.7867,
|
| 667 |
+
"step": 1410
|
| 668 |
+
},
|
| 669 |
+
{
|
| 670 |
+
"epoch": 0.0027749757554749784,
|
| 671 |
+
"grad_norm": 1.9892805814743042,
|
| 672 |
+
"learning_rate": 2.999942999765228e-05,
|
| 673 |
+
"loss": 1.8632,
|
| 674 |
+
"step": 1425
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 0.0028041860265852416,
|
| 678 |
+
"grad_norm": 2.5921459197998047,
|
| 679 |
+
"learning_rate": 2.999941793452272e-05,
|
| 680 |
+
"loss": 1.9155,
|
| 681 |
+
"step": 1440
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.0028333962976955043,
|
| 685 |
+
"grad_norm": 6.098895072937012,
|
| 686 |
+
"learning_rate": 2.9999405745080948e-05,
|
| 687 |
+
"loss": 2.0112,
|
| 688 |
+
"step": 1455
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.0028626065688057674,
|
| 692 |
+
"grad_norm": 2.7144930362701416,
|
| 693 |
+
"learning_rate": 2.9999393429327073e-05,
|
| 694 |
+
"loss": 2.0019,
|
| 695 |
+
"step": 1470
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 0.00289181683991603,
|
| 699 |
+
"grad_norm": 2.3964602947235107,
|
| 700 |
+
"learning_rate": 2.9999380987261195e-05,
|
| 701 |
+
"loss": 1.9524,
|
| 702 |
+
"step": 1485
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 0.0029210271110262933,
|
| 706 |
+
"grad_norm": 3.2644741535186768,
|
| 707 |
+
"learning_rate": 2.9999368418883422e-05,
|
| 708 |
+
"loss": 1.881,
|
| 709 |
+
"step": 1500
|
| 710 |
+
},
|
| 711 |
+
{
|
| 712 |
+
"epoch": 0.002950237382136556,
|
| 713 |
+
"grad_norm": 3.0930943489074707,
|
| 714 |
+
"learning_rate": 2.9999355724193854e-05,
|
| 715 |
+
"loss": 2.0001,
|
| 716 |
+
"step": 1515
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 0.002979447653246819,
|
| 720 |
+
"grad_norm": 2.6339657306671143,
|
| 721 |
+
"learning_rate": 2.9999342903192608e-05,
|
| 722 |
+
"loss": 1.9344,
|
| 723 |
+
"step": 1530
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"epoch": 0.003008657924357082,
|
| 727 |
+
"grad_norm": 3.7254459857940674,
|
| 728 |
+
"learning_rate": 2.999932995587978e-05,
|
| 729 |
+
"loss": 2.0673,
|
| 730 |
+
"step": 1545
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 0.003037868195467345,
|
| 734 |
+
"grad_norm": 2.7078473567962646,
|
| 735 |
+
"learning_rate": 2.999931688225549e-05,
|
| 736 |
+
"loss": 2.0223,
|
| 737 |
+
"step": 1560
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.0030670784665776077,
|
| 741 |
+
"grad_norm": 2.7792718410491943,
|
| 742 |
+
"learning_rate": 2.999930368231984e-05,
|
| 743 |
+
"loss": 1.902,
|
| 744 |
+
"step": 1575
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"epoch": 0.003096288737687871,
|
| 748 |
+
"grad_norm": 4.607854843139648,
|
| 749 |
+
"learning_rate": 2.999929035607294e-05,
|
| 750 |
+
"loss": 1.9536,
|
| 751 |
+
"step": 1590
|
| 752 |
+
},
|
| 753 |
+
{
|
| 754 |
+
"epoch": 0.0031254990087981336,
|
| 755 |
+
"grad_norm": 2.4656715393066406,
|
| 756 |
+
"learning_rate": 2.999927690351491e-05,
|
| 757 |
+
"loss": 2.1792,
|
| 758 |
+
"step": 1605
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 0.0031547092799083967,
|
| 762 |
+
"grad_norm": 2.8231770992279053,
|
| 763 |
+
"learning_rate": 2.9999263324645863e-05,
|
| 764 |
+
"loss": 1.8908,
|
| 765 |
+
"step": 1620
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 0.0031839195510186594,
|
| 769 |
+
"grad_norm": 3.5741395950317383,
|
| 770 |
+
"learning_rate": 2.9999249619465904e-05,
|
| 771 |
+
"loss": 1.9505,
|
| 772 |
+
"step": 1635
|
| 773 |
+
},
|
| 774 |
+
{
|
| 775 |
+
"epoch": 0.0032131298221289226,
|
| 776 |
+
"grad_norm": 2.0066606998443604,
|
| 777 |
+
"learning_rate": 2.999923578797516e-05,
|
| 778 |
+
"loss": 2.0346,
|
| 779 |
+
"step": 1650
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 0.0032423400932391853,
|
| 783 |
+
"grad_norm": 2.5830793380737305,
|
| 784 |
+
"learning_rate": 2.999922183017374e-05,
|
| 785 |
+
"loss": 1.9546,
|
| 786 |
+
"step": 1665
|
| 787 |
+
},
|
| 788 |
+
{
|
| 789 |
+
"epoch": 0.0032715503643494484,
|
| 790 |
+
"grad_norm": 2.8811633586883545,
|
| 791 |
+
"learning_rate": 2.999920774606176e-05,
|
| 792 |
+
"loss": 1.8517,
|
| 793 |
+
"step": 1680
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.003300760635459711,
|
| 797 |
+
"grad_norm": 3.0571601390838623,
|
| 798 |
+
"learning_rate": 2.999919353563935e-05,
|
| 799 |
+
"loss": 1.8746,
|
| 800 |
+
"step": 1695
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 0.0033299709065699743,
|
| 804 |
+
"grad_norm": 2.927371025085449,
|
| 805 |
+
"learning_rate": 2.9999179198906614e-05,
|
| 806 |
+
"loss": 2.0249,
|
| 807 |
+
"step": 1710
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 0.003359181177680237,
|
| 811 |
+
"grad_norm": 2.425579071044922,
|
| 812 |
+
"learning_rate": 2.9999164735863685e-05,
|
| 813 |
+
"loss": 1.9149,
|
| 814 |
+
"step": 1725
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"epoch": 0.0033883914487905,
|
| 818 |
+
"grad_norm": 3.0711071491241455,
|
| 819 |
+
"learning_rate": 2.9999150146510678e-05,
|
| 820 |
+
"loss": 1.8335,
|
| 821 |
+
"step": 1740
|
| 822 |
+
},
|
| 823 |
+
{
|
| 824 |
+
"epoch": 0.003417601719900763,
|
| 825 |
+
"grad_norm": 4.509474277496338,
|
| 826 |
+
"learning_rate": 2.9999135430847718e-05,
|
| 827 |
+
"loss": 2.0002,
|
| 828 |
+
"step": 1755
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"epoch": 0.003446811991011026,
|
| 832 |
+
"grad_norm": 2.784403085708618,
|
| 833 |
+
"learning_rate": 2.9999120588874927e-05,
|
| 834 |
+
"loss": 1.985,
|
| 835 |
+
"step": 1770
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"epoch": 0.0034760222621212887,
|
| 839 |
+
"grad_norm": 4.556781768798828,
|
| 840 |
+
"learning_rate": 2.9999105620592434e-05,
|
| 841 |
+
"loss": 1.8832,
|
| 842 |
+
"step": 1785
|
| 843 |
+
},
|
| 844 |
+
{
|
| 845 |
+
"epoch": 0.003505232533231552,
|
| 846 |
+
"grad_norm": 3.1926143169403076,
|
| 847 |
+
"learning_rate": 2.9999090526000364e-05,
|
| 848 |
+
"loss": 2.0136,
|
| 849 |
+
"step": 1800
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"epoch": 0.0035344428043418146,
|
| 853 |
+
"grad_norm": 3.627634286880493,
|
| 854 |
+
"learning_rate": 2.9999075305098846e-05,
|
| 855 |
+
"loss": 2.0413,
|
| 856 |
+
"step": 1815
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"epoch": 0.0035636530754520777,
|
| 860 |
+
"grad_norm": 4.250141620635986,
|
| 861 |
+
"learning_rate": 2.9999059957887998e-05,
|
| 862 |
+
"loss": 1.9076,
|
| 863 |
+
"step": 1830
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 0.0035928633465623404,
|
| 867 |
+
"grad_norm": 4.263555526733398,
|
| 868 |
+
"learning_rate": 2.999904448436796e-05,
|
| 869 |
+
"loss": 1.9782,
|
| 870 |
+
"step": 1845
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"epoch": 0.0036220736176726036,
|
| 874 |
+
"grad_norm": 2.5871047973632812,
|
| 875 |
+
"learning_rate": 2.999902888453886e-05,
|
| 876 |
+
"loss": 1.9703,
|
| 877 |
+
"step": 1860
|
| 878 |
+
},
|
| 879 |
+
{
|
| 880 |
+
"epoch": 0.0036512838887828663,
|
| 881 |
+
"grad_norm": 2.2850048542022705,
|
| 882 |
+
"learning_rate": 2.9999013158400827e-05,
|
| 883 |
+
"loss": 1.9571,
|
| 884 |
+
"step": 1875
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.0036804941598931294,
|
| 888 |
+
"grad_norm": 2.419032096862793,
|
| 889 |
+
"learning_rate": 2.9998997305953993e-05,
|
| 890 |
+
"loss": 2.1583,
|
| 891 |
+
"step": 1890
|
| 892 |
+
},
|
| 893 |
+
{
|
| 894 |
+
"epoch": 0.003709704431003392,
|
| 895 |
+
"grad_norm": 3.190084934234619,
|
| 896 |
+
"learning_rate": 2.9998981327198497e-05,
|
| 897 |
+
"loss": 1.9732,
|
| 898 |
+
"step": 1905
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 0.0037389147021136553,
|
| 902 |
+
"grad_norm": 3.31540846824646,
|
| 903 |
+
"learning_rate": 2.9998965222134468e-05,
|
| 904 |
+
"loss": 1.8646,
|
| 905 |
+
"step": 1920
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.003768124973223918,
|
| 909 |
+
"grad_norm": 2.423227310180664,
|
| 910 |
+
"learning_rate": 2.9998948990762044e-05,
|
| 911 |
+
"loss": 1.9282,
|
| 912 |
+
"step": 1935
|
| 913 |
+
},
|
| 914 |
+
{
|
| 915 |
+
"epoch": 0.003797335244334181,
|
| 916 |
+
"grad_norm": 3.4635472297668457,
|
| 917 |
+
"learning_rate": 2.999893263308136e-05,
|
| 918 |
+
"loss": 1.9859,
|
| 919 |
+
"step": 1950
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"epoch": 0.003826545515444444,
|
| 923 |
+
"grad_norm": 2.474353790283203,
|
| 924 |
+
"learning_rate": 2.9998916149092556e-05,
|
| 925 |
+
"loss": 2.0055,
|
| 926 |
+
"step": 1965
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 0.003855755786554707,
|
| 930 |
+
"grad_norm": 3.264435052871704,
|
| 931 |
+
"learning_rate": 2.9998899538795766e-05,
|
| 932 |
+
"loss": 1.9755,
|
| 933 |
+
"step": 1980
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"epoch": 0.0038849660576649697,
|
| 937 |
+
"grad_norm": 2.028916358947754,
|
| 938 |
+
"learning_rate": 2.9998882802191137e-05,
|
| 939 |
+
"loss": 1.9914,
|
| 940 |
+
"step": 1995
|
| 941 |
+
},
|
| 942 |
+
{
|
| 943 |
+
"epoch": 0.003914176328775233,
|
| 944 |
+
"grad_norm": 1.9189203977584839,
|
| 945 |
+
"learning_rate": 2.9998865939278805e-05,
|
| 946 |
+
"loss": 1.8109,
|
| 947 |
+
"step": 2010
|
| 948 |
+
},
|
| 949 |
+
{
|
| 950 |
+
"epoch": 0.0039433865998854956,
|
| 951 |
+
"grad_norm": 3.143857955932617,
|
| 952 |
+
"learning_rate": 2.9998848950058913e-05,
|
| 953 |
+
"loss": 1.8233,
|
| 954 |
+
"step": 2025
|
| 955 |
+
},
|
| 956 |
+
{
|
| 957 |
+
"epoch": 0.003972596870995758,
|
| 958 |
+
"grad_norm": 2.662853479385376,
|
| 959 |
+
"learning_rate": 2.9998831834531608e-05,
|
| 960 |
+
"loss": 2.0015,
|
| 961 |
+
"step": 2040
|
| 962 |
+
},
|
| 963 |
+
{
|
| 964 |
+
"epoch": 0.004001807142106022,
|
| 965 |
+
"grad_norm": 2.0305709838867188,
|
| 966 |
+
"learning_rate": 2.9998814592697027e-05,
|
| 967 |
+
"loss": 1.8817,
|
| 968 |
+
"step": 2055
|
| 969 |
+
},
|
| 970 |
+
{
|
| 971 |
+
"epoch": 0.0040310174132162846,
|
| 972 |
+
"grad_norm": 4.386982440948486,
|
| 973 |
+
"learning_rate": 2.999879722455532e-05,
|
| 974 |
+
"loss": 1.8511,
|
| 975 |
+
"step": 2070
|
| 976 |
+
},
|
| 977 |
+
{
|
| 978 |
+
"epoch": 0.004060227684326547,
|
| 979 |
+
"grad_norm": 3.3638317584991455,
|
| 980 |
+
"learning_rate": 2.9998779730106633e-05,
|
| 981 |
+
"loss": 1.9738,
|
| 982 |
+
"step": 2085
|
| 983 |
+
},
|
| 984 |
+
{
|
| 985 |
+
"epoch": 0.00408943795543681,
|
| 986 |
+
"grad_norm": 3.6872494220733643,
|
| 987 |
+
"learning_rate": 2.9998762109351107e-05,
|
| 988 |
+
"loss": 1.8496,
|
| 989 |
+
"step": 2100
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.0041186482265470736,
|
| 993 |
+
"grad_norm": 3.2654178142547607,
|
| 994 |
+
"learning_rate": 2.9998744362288902e-05,
|
| 995 |
+
"loss": 1.9003,
|
| 996 |
+
"step": 2115
|
| 997 |
+
},
|
| 998 |
+
{
|
| 999 |
+
"epoch": 0.004147858497657336,
|
| 1000 |
+
"grad_norm": 3.4839179515838623,
|
| 1001 |
+
"learning_rate": 2.9998726488920162e-05,
|
| 1002 |
+
"loss": 2.016,
|
| 1003 |
+
"step": 2130
|
| 1004 |
+
},
|
| 1005 |
+
{
|
| 1006 |
+
"epoch": 0.004177068768767599,
|
| 1007 |
+
"grad_norm": 2.537320613861084,
|
| 1008 |
+
"learning_rate": 2.9998708489245034e-05,
|
| 1009 |
+
"loss": 2.0073,
|
| 1010 |
+
"step": 2145
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"epoch": 0.004206279039877862,
|
| 1014 |
+
"grad_norm": 3.2071378231048584,
|
| 1015 |
+
"learning_rate": 2.999869036326367e-05,
|
| 1016 |
+
"loss": 1.9835,
|
| 1017 |
+
"step": 2160
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"epoch": 0.004235489310988125,
|
| 1021 |
+
"grad_norm": 2.7338645458221436,
|
| 1022 |
+
"learning_rate": 2.999867211097623e-05,
|
| 1023 |
+
"loss": 1.8995,
|
| 1024 |
+
"step": 2175
|
| 1025 |
+
},
|
| 1026 |
+
{
|
| 1027 |
+
"epoch": 0.004264699582098388,
|
| 1028 |
+
"grad_norm": 4.866488933563232,
|
| 1029 |
+
"learning_rate": 2.999865373238286e-05,
|
| 1030 |
+
"loss": 1.9656,
|
| 1031 |
+
"step": 2190
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"epoch": 0.004293909853208651,
|
| 1035 |
+
"grad_norm": 2.67726731300354,
|
| 1036 |
+
"learning_rate": 2.9998635227483715e-05,
|
| 1037 |
+
"loss": 2.0401,
|
| 1038 |
+
"step": 2205
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"epoch": 0.004323120124318914,
|
| 1042 |
+
"grad_norm": 3.1271047592163086,
|
| 1043 |
+
"learning_rate": 2.9998616596278955e-05,
|
| 1044 |
+
"loss": 2.0406,
|
| 1045 |
+
"step": 2220
|
| 1046 |
+
},
|
| 1047 |
+
{
|
| 1048 |
+
"epoch": 0.004352330395429177,
|
| 1049 |
+
"grad_norm": 5.673630237579346,
|
| 1050 |
+
"learning_rate": 2.9998597838768738e-05,
|
| 1051 |
+
"loss": 1.9733,
|
| 1052 |
+
"step": 2235
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 0.00438154066653944,
|
| 1056 |
+
"grad_norm": 2.696476697921753,
|
| 1057 |
+
"learning_rate": 2.9998578954953216e-05,
|
| 1058 |
+
"loss": 2.16,
|
| 1059 |
+
"step": 2250
|
| 1060 |
+
},
|
| 1061 |
+
{
|
| 1062 |
+
"epoch": 0.004410750937649702,
|
| 1063 |
+
"grad_norm": 5.409951686859131,
|
| 1064 |
+
"learning_rate": 2.9998559944832553e-05,
|
| 1065 |
+
"loss": 1.9401,
|
| 1066 |
+
"step": 2265
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"epoch": 0.004439961208759966,
|
| 1070 |
+
"grad_norm": 2.183530330657959,
|
| 1071 |
+
"learning_rate": 2.9998540808406903e-05,
|
| 1072 |
+
"loss": 1.9405,
|
| 1073 |
+
"step": 2280
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"epoch": 0.004469171479870229,
|
| 1077 |
+
"grad_norm": 3.3076109886169434,
|
| 1078 |
+
"learning_rate": 2.9998521545676438e-05,
|
| 1079 |
+
"loss": 1.9617,
|
| 1080 |
+
"step": 2295
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 0.004498381750980491,
|
| 1084 |
+
"grad_norm": 2.793837547302246,
|
| 1085 |
+
"learning_rate": 2.999850215664131e-05,
|
| 1086 |
+
"loss": 1.9463,
|
| 1087 |
+
"step": 2310
|
| 1088 |
+
},
|
| 1089 |
+
{
|
| 1090 |
+
"epoch": 0.004527592022090754,
|
| 1091 |
+
"grad_norm": 2.6732029914855957,
|
| 1092 |
+
"learning_rate": 2.9998482641301687e-05,
|
| 1093 |
+
"loss": 1.9114,
|
| 1094 |
+
"step": 2325
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 0.004556802293201018,
|
| 1098 |
+
"grad_norm": 2.1354994773864746,
|
| 1099 |
+
"learning_rate": 2.999846299965773e-05,
|
| 1100 |
+
"loss": 2.0321,
|
| 1101 |
+
"step": 2340
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"epoch": 0.00458601256431128,
|
| 1105 |
+
"grad_norm": 3.889298439025879,
|
| 1106 |
+
"learning_rate": 2.9998443231709608e-05,
|
| 1107 |
+
"loss": 2.0722,
|
| 1108 |
+
"step": 2355
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"epoch": 0.004615222835421543,
|
| 1112 |
+
"grad_norm": 3.6255850791931152,
|
| 1113 |
+
"learning_rate": 2.9998423337457486e-05,
|
| 1114 |
+
"loss": 2.0355,
|
| 1115 |
+
"step": 2370
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 0.004644433106531806,
|
| 1119 |
+
"grad_norm": 2.4627747535705566,
|
| 1120 |
+
"learning_rate": 2.9998403316901533e-05,
|
| 1121 |
+
"loss": 2.0076,
|
| 1122 |
+
"step": 2385
|
| 1123 |
+
},
|
| 1124 |
+
{
|
| 1125 |
+
"epoch": 0.004673643377642069,
|
| 1126 |
+
"grad_norm": 3.3934834003448486,
|
| 1127 |
+
"learning_rate": 2.9998383170041916e-05,
|
| 1128 |
+
"loss": 1.9877,
|
| 1129 |
+
"step": 2400
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.004702853648752332,
|
| 1133 |
+
"grad_norm": 2.2475860118865967,
|
| 1134 |
+
"learning_rate": 2.99983628968788e-05,
|
| 1135 |
+
"loss": 1.8463,
|
| 1136 |
+
"step": 2415
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 0.004732063919862595,
|
| 1140 |
+
"grad_norm": 2.8804898262023926,
|
| 1141 |
+
"learning_rate": 2.9998342497412365e-05,
|
| 1142 |
+
"loss": 2.0943,
|
| 1143 |
+
"step": 2430
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 0.0047612741909728576,
|
| 1147 |
+
"grad_norm": 2.2566545009613037,
|
| 1148 |
+
"learning_rate": 2.999832197164278e-05,
|
| 1149 |
+
"loss": 1.8834,
|
| 1150 |
+
"step": 2445
|
| 1151 |
+
},
|
| 1152 |
+
{
|
| 1153 |
+
"epoch": 0.004790484462083121,
|
| 1154 |
+
"grad_norm": 2.948420524597168,
|
| 1155 |
+
"learning_rate": 2.9998301319570216e-05,
|
| 1156 |
+
"loss": 2.002,
|
| 1157 |
+
"step": 2460
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 0.004819694733193384,
|
| 1161 |
+
"grad_norm": 3.6838879585266113,
|
| 1162 |
+
"learning_rate": 2.999828054119484e-05,
|
| 1163 |
+
"loss": 1.9071,
|
| 1164 |
+
"step": 2475
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"epoch": 0.0048489050043036466,
|
| 1168 |
+
"grad_norm": 2.2608137130737305,
|
| 1169 |
+
"learning_rate": 2.9998259636516845e-05,
|
| 1170 |
+
"loss": 1.9241,
|
| 1171 |
+
"step": 2490
|
| 1172 |
+
}
|
| 1173 |
+
],
|
| 1174 |
+
"logging_steps": 15,
|
| 1175 |
+
"max_steps": 513518,
|
| 1176 |
+
"num_input_tokens_seen": 0,
|
| 1177 |
+
"num_train_epochs": 1,
|
| 1178 |
+
"save_steps": 500,
|
| 1179 |
+
"stateful_callbacks": {
|
| 1180 |
+
"TrainerControl": {
|
| 1181 |
+
"args": {
|
| 1182 |
+
"should_epoch_stop": false,
|
| 1183 |
+
"should_evaluate": false,
|
| 1184 |
+
"should_log": false,
|
| 1185 |
+
"should_save": true,
|
| 1186 |
+
"should_training_stop": false
|
| 1187 |
+
},
|
| 1188 |
+
"attributes": {}
|
| 1189 |
+
}
|
| 1190 |
+
},
|
| 1191 |
+
"total_flos": 3936391872509952.0,
|
| 1192 |
+
"train_batch_size": 4,
|
| 1193 |
+
"trial_name": null,
|
| 1194 |
+
"trial_params": null
|
| 1195 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-2500/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0287fbd96bea9652b04a1bb7ed4b6b1e8de92cccbc4b448c2b5d263a08664d48
|
| 3 |
+
size 5304
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "jonathanjordan21/mos-mamba-6x130m-trainer-sft",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MoSMambaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "jonathanjordan21/mos-mamba-6x130m-hf--configuration_mos_mamba.MoSMambaConfig",
|
| 8 |
+
"AutoModelForCausalLM": "jonathanjordan21/mos-mamba-6x130m-hf--modeling_mos_mamba.MoSMambaForCausalLM"
|
| 9 |
+
},
|
| 10 |
+
"bos_token_id": 0,
|
| 11 |
+
"conv_kernel": 4,
|
| 12 |
+
"d_inner": 1536,
|
| 13 |
+
"d_model": 768,
|
| 14 |
+
"eos_token_id": 50278,
|
| 15 |
+
"expand": 2,
|
| 16 |
+
"fused_add_norm": true,
|
| 17 |
+
"hidden_act": "silu",
|
| 18 |
+
"hidden_size": 768,
|
| 19 |
+
"initializer_range": 0.1,
|
| 20 |
+
"intermediate_size": 1536,
|
| 21 |
+
"layer_norm_epsilon": 1e-05,
|
| 22 |
+
"model_type": "MoSMamba",
|
| 23 |
+
"n_layer": 24,
|
| 24 |
+
"num_hidden_layers": 24,
|
| 25 |
+
"num_selectivities": 6,
|
| 26 |
+
"num_selectivities_per_tok": 2,
|
| 27 |
+
"output_router_logits": true,
|
| 28 |
+
"pad_token_id": 0,
|
| 29 |
+
"pad_vocab_size_multiple": 8,
|
| 30 |
+
"rescale_prenorm_residual": false,
|
| 31 |
+
"residual_in_fp32": true,
|
| 32 |
+
"rms_norm": true,
|
| 33 |
+
"ssm_cfg": {},
|
| 34 |
+
"state_size": 16,
|
| 35 |
+
"time_step_floor": 0.0001,
|
| 36 |
+
"time_step_init_scheme": "random",
|
| 37 |
+
"time_step_max": 0.1,
|
| 38 |
+
"time_step_min": 0.001,
|
| 39 |
+
"time_step_rank": 48,
|
| 40 |
+
"time_step_scale": 1.0,
|
| 41 |
+
"torch_dtype": "float32",
|
| 42 |
+
"transformers_version": "4.42.4",
|
| 43 |
+
"use_bias": false,
|
| 44 |
+
"use_cache": true,
|
| 45 |
+
"use_conv_bias": true,
|
| 46 |
+
"vocab_size": 50280
|
| 47 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 50278,
|
| 5 |
+
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.42.4"
|
| 7 |
+
}
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d18241b8cce7e29615f02b3dda5863324d16cfc1da1abf6f85ab568e49122de
|
| 3 |
+
size 576008736
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca82d37270885d8f17d8e6730f027d7c6dfa6786984ae4a210d00ae176b82896
|
| 3 |
+
size 1152256984
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eac24a6faa5ec8b2faf37093a9d4ae29821df535acefbdb21ed8e7097d0484d9
|
| 3 |
+
size 14244
|
data/mos-mamba-6x130m-trainer-sft/checkpoint-3000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14ab0e69b0aef02a14ac79155cbb70b46767177399038a372265d1b4d3e251d0
|
| 3 |
+
size 1064
|