Training in progress, step 500

README.md

@@ -20,15 +20,15 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.9218
-- Rewards/chosen: -0.0573
-- Rewards/rejected: -0.1441
-- Rewards/accuracies: 0.5573
-- Rewards/margins: 0.0868
-- Logps/rejected: -37.7225
-- Logps/chosen: -34.1164
-- Logits/rejected: -2.1841
-- Logits/chosen: -2.1889
+- Loss: 0.6957
+- Rewards/chosen: -0.0294
+- Rewards/rejected: -0.0431
+- Rewards/accuracies: 0.5158
+- Rewards/margins: 0.0138
+- Logps/rejected: -37.5783
+- Logps/chosen: -34.0765
+- Logits/rejected: -2.2231
+- Logits/chosen: -2.2280
 
 ## Model description
 
@@ -57,33 +57,21 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 4
+- num_epochs: 1
 
 ### Training results
 
-| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
-|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
-| 0.8961        | 0.26  | 100  | -2.2315       | -2.2266         | -34.0252     | -37.5409       | 0.9767          | 0.5365             | 0.0065         | 0.0235          | -0.0170          |
-| 0.7699        | 0.52  | 200  | -2.2282       | -2.2234         | -34.0547     | -37.5737       | 0.9742          | 0.5303             | -0.0141        | 0.0259          | -0.0400          |
-| 0.6723        | 0.78  | 300  | -2.2234       | -2.2186         | -34.0868     | -37.6047       | 0.9761          | 0.5299             | -0.0366        | 0.0250          | -0.0616          |
-| 0.4211        | 1.04  | 400  | 0.9373        | -0.0429         | -0.1064      | 0.5482         | 0.0635          | -37.6686           | -34.0958       | -2.1988         | -2.2037          |
-| 0.3278        | 1.3   | 500  | 0.9598        | -0.1367         | -0.1786      | 0.5511         | 0.0418          | -37.7717           | -34.2299       | -2.2015         | -2.2064          |
-| 0.3797        | 1.56  | 600  | 0.9195        | -0.0682         | -0.1551      | 0.5951         | 0.0868          | -37.7381           | -34.1320       | -2.1871         | -2.1920          |
-| 0.3051        | 1.82  | 700  | 0.9317        | -0.0587         | -0.1357      | 0.5685         | 0.0769          | -37.7104           | -34.1184       | -2.1861         | -2.1909          |
-| 0.1161        | 2.08  | 800  | 0.9294        | -0.0807         | -0.1629      | 0.5573         | 0.0822          | -37.7493           | -34.1499       | -2.1868         | -2.1916          |
-| 0.1219        | 2.34  | 900  | 0.9271        | -0.0710         | -0.1562      | 0.5627         | 0.0852          | -37.7398           | -34.1360       | -2.1850         | -2.1898          |
-| 0.1667        | 2.6   | 1000 | 0.9150        | -0.0668         | -0.1633      | 0.5864         | 0.0965          | -37.7499           | -34.1300       | -2.1819         | -2.1867          |
-| 0.1242        | 2.86  | 1100 | 0.9304        | -0.0665         | -0.1474      | 0.5718         | 0.0809          | -37.7272           | -34.1295       | -2.1837         | -2.1885          |
-| 0.1476        | 3.12  | 1200 | 0.9214        | -0.0512         | -0.1408      | 0.5831         | 0.0895          | -37.7177           | -34.1077       | -2.1839         | -2.1887          |
-| 0.0718        | 3.38  | 1300 | 0.9157        | -0.0524         | -0.1471      | 0.5660         | 0.0947          | -37.7268           | -34.1094       | -2.1847         | -2.1895          |
-| 0.045         | 3.64  | 1400 | 0.9232        | -0.0581         | -0.1471      | 0.5889         | 0.0890          | -37.7267           | -34.1176       | -2.1848         | -2.1896          |
-| 0.1201        | 3.9   | 1500 | 0.9239        | -0.0601         | -0.1468      | 0.5864         | 0.0867          | -37.7263           | -34.1204       | -2.1840         | -2.1888          |
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 0.6577        | 0.26  | 100  | 0.6949          | -0.0038        | -0.0133          | 0.5316             | 0.0095          | -37.5357       | -34.0400     | -2.2289         | -2.2338       |
+| 0.6156        | 0.52  | 200  | 0.6943          | -0.0132        | -0.0288          | 0.5191             | 0.0156          | -37.5578       | -34.0535     | -2.2268         | -2.2317       |
+| 0.5468        | 0.78  | 300  | 0.6903          | -0.0237        | -0.0486          | 0.5191             | 0.0249          | -37.5860       | -34.0684     | -2.2229         | -2.2277       |
 
 
 ### Framework versions
 
-- PEFT 0.10.0
-- Transformers 4.39.0.dev0
+- PEFT 0.8.2
+- Transformers 4.37.2
 - Pytorch 2.1.2+cu121
-- Datasets 2.14.6
+- Datasets 2.17.0
 - Tokenizers 0.15.1

adapter_config.json

@@ -20,13 +20,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "down_proj",
-    "up_proj",
-    "o_proj",
-    "q_proj",
     "k_proj",
     "v_proj",
-    "gate_proj"
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj",
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b15c8f6f5b190ba807daf940437159053d92dcf1cfeeaa44f49aa0b7a6be087
+oid sha256:cd693db8a5c7cc0e67d6474c6be8058969b5f2a7be9ecce3d2779608f038c05d
 size 176183216

all_results.json

@@ -1,21 +1,21 @@
 {
-    "epoch": 4.0,
-    "eval_logits/chosen": -2.1889240741729736,
-    "eval_logits/rejected": -2.184124231338501,
-    "eval_logps/chosen": -34.11641311645508,
-    "eval_logps/rejected": -37.722450256347656,
-    "eval_loss": 0.9218093156814575,
-    "eval_rewards/accuracies": 0.5573089718818665,
-    "eval_rewards/chosen": -0.05729999020695686,
-    "eval_rewards/margins": 0.08677948266267776,
-    "eval_rewards/rejected": -0.14407947659492493,
-    "eval_runtime": 145.1213,
+    "epoch": 1.0,
+    "eval_logits/chosen": -2.227961301803589,
+    "eval_logits/rejected": -2.2231268882751465,
+    "eval_logps/chosen": -34.07651138305664,
+    "eval_logps/rejected": -37.578250885009766,
+    "eval_loss": 0.6956681609153748,
+    "eval_rewards/accuracies": 0.5157807469367981,
+    "eval_rewards/chosen": -0.029370848089456558,
+    "eval_rewards/margins": 0.013771643862128258,
+    "eval_rewards/rejected": -0.043142493814229965,
+    "eval_runtime": 145.5641,
     "eval_samples": 343,
-    "eval_samples_per_second": 2.364,
-    "eval_steps_per_second": 0.296,
-    "train_loss": 0.17721207693412708,
-    "train_runtime": 10735.2543,
+    "eval_samples_per_second": 2.356,
+    "eval_steps_per_second": 0.295,
+    "train_loss": 0.6291831790626823,
+    "train_runtime": 3251.1508,
     "train_samples": 3079,
-    "train_samples_per_second": 1.147,
-    "train_steps_per_second": 0.143
+    "train_samples_per_second": 0.947,
+    "train_steps_per_second": 0.118
 }

config.json

@@ -18,8 +18,6 @@
   "n_layer": 28,
   "n_positions": 2048,
   "quantization_config": {
-    "_load_in_4bit": true,
-    "_load_in_8bit": false,
     "bnb_4bit_compute_dtype": "float16",
     "bnb_4bit_quant_type": "nf4",
     "bnb_4bit_use_double_quant": false,
@@ -47,7 +45,7 @@
   },
   "tie_word_embeddings": false,
   "tokenizer_class": "GPT2Tokenizer",
-  "transformers_version": "4.39.0.dev0",
+  "transformers_version": "4.37.2",
   "use_cache": true,
   "vocab_size": 50400
 }

eval_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 4.0,
-    "eval_logits/chosen": -2.1889240741729736,
-    "eval_logits/rejected": -2.184124231338501,
-    "eval_logps/chosen": -34.11641311645508,
-    "eval_logps/rejected": -37.722450256347656,
-    "eval_loss": 0.9218093156814575,
-    "eval_rewards/accuracies": 0.5573089718818665,
-    "eval_rewards/chosen": -0.05729999020695686,
-    "eval_rewards/margins": 0.08677948266267776,
-    "eval_rewards/rejected": -0.14407947659492493,
-    "eval_runtime": 145.1213,
+    "epoch": 1.0,
+    "eval_logits/chosen": -2.227961301803589,
+    "eval_logits/rejected": -2.2231268882751465,
+    "eval_logps/chosen": -34.07651138305664,
+    "eval_logps/rejected": -37.578250885009766,
+    "eval_loss": 0.6956681609153748,
+    "eval_rewards/accuracies": 0.5157807469367981,
+    "eval_rewards/chosen": -0.029370848089456558,
+    "eval_rewards/margins": 0.013771643862128258,
+    "eval_rewards/rejected": -0.043142493814229965,
+    "eval_runtime": 145.5641,
     "eval_samples": 343,
-    "eval_samples_per_second": 2.364,
-    "eval_steps_per_second": 0.296
+    "eval_samples_per_second": 2.356,
+    "eval_steps_per_second": 0.295
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 4.0,
-    "train_loss": 0.17721207693412708,
-    "train_runtime": 10735.2543,
+    "epoch": 1.0,
+    "train_loss": 0.6291831790626823,
+    "train_runtime": 3251.1508,
     "train_samples": 3079,
-    "train_samples_per_second": 1.147,
-    "train_steps_per_second": 0.143
+    "train_samples_per_second": 0.947,
+    "train_steps_per_second": 0.118
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b715c426f0a7cb99b86fb1b4c90c55353be43921743323373dda6cb1c70085d8
+oid sha256:6a7873256b464917bacb6f47f59f51881713a7cee62cb71ff848461a2de6635e
 size 5176