Model save

README.md

@@ -2,15 +2,12 @@
 license: apache-2.0
 base_model: Minbyul/biomistral-7b-wo-kqa_golden-iter-sft-dpo-step1
 tags:
-- alignment-handbook
-- trl
-- sft
-- generated_from_trainer
 - trl
 - sft
+- alignment-handbook
 - generated_from_trainer
 datasets:
-- HuggingFaceH4/deita-10k-v0-sft
+- generator
 model-index:
 - name: biomistral-7b-wo-kqa_golden-iter-sft-step2
   results: []
@@ -21,9 +18,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # biomistral-7b-wo-kqa_golden-iter-sft-step2
 
-This model is a fine-tuned version of [Minbyul/biomistral-7b-wo-kqa_golden-iter-sft-dpo-step1](https://huggingface.co/Minbyul/biomistral-7b-wo-kqa_golden-iter-sft-dpo-step1) on the HuggingFaceH4/deita-10k-v0-sft dataset.
+This model is a fine-tuned version of [Minbyul/biomistral-7b-wo-kqa_golden-iter-sft-dpo-step1](https://huggingface.co/Minbyul/biomistral-7b-wo-kqa_golden-iter-sft-dpo-step1) on the generator dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.9173
+- Loss: 1.7672
 
 ## Model description
 
@@ -60,9 +57,9 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 0.5168        | 1.0   | 16   | 1.5115          |
-| 0.219         | 2.0   | 32   | 1.7233          |
-| 0.0956        | 3.0   | 48   | 1.9173          |
+| 0.4982        | 0.95  | 13   | 1.4957          |
+| 0.2254        | 1.96  | 27   | 1.6817          |
+| 0.1095        | 2.84  | 39   | 1.7672          |
 
 
 ### Framework versions

all_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 3.0,
+    "epoch": 2.84,
     "eval_loss": 1.9173061847686768,
     "eval_runtime": 29.1672,
     "eval_samples": 4044,
     "eval_samples_per_second": 11.794,
     "eval_steps_per_second": 0.754,
-    "train_loss": 0.2755420195559661,
-    "train_runtime": 906.8406,
+    "train_loss": 0.2916483015586168,
+    "train_runtime": 753.2022,
     "train_samples": 4747,
-    "train_samples_per_second": 3.374,
-    "train_steps_per_second": 0.053
+    "train_samples_per_second": 3.453,
+    "train_steps_per_second": 0.052
 }

config.json

@@ -21,6 +21,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.39.0.dev0",
-  "use_cache": true,
+  "use_cache": false,
   "vocab_size": 32000
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54b1ff6791cec8ef481e26fd70ce3d118a6d8655a166ad8052a55c6f1a496d24
+oid sha256:099b26e30badedeb7f7ee064e387ae0d7db0a9d9dbae28f4bdb2096402b47d09
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c09cc0e05da0ff140b5aea7aa8df73549cbec70e4f6c1e99df33a54224ac6a11
+oid sha256:c0d63bad4dd01e66c8046d83ab95fefd817d914eaffa7c6ea895cd72fb0cd206
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e70d482242b904ffe123885aa6e8fc6322870466a7fd7cf4731bba68a0960abf
+oid sha256:2d1ac10b23a48a2b7f57a4eaae007003211e9bad91bee2d764284714df6ce352
 size 4540516344

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 3.0,
-    "train_loss": 0.2755420195559661,
-    "train_runtime": 906.8406,
+    "epoch": 2.84,
+    "train_loss": 0.2916483015586168,
+    "train_runtime": 753.2022,
     "train_samples": 4747,
-    "train_samples_per_second": 3.374,
-    "train_steps_per_second": 0.053
+    "train_samples_per_second": 3.453,
+    "train_steps_per_second": 0.052
 }

trainer_state.json

@@ -1,123 +1,109 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 3.0,
+  "epoch": 2.8363636363636364,
   "eval_steps": 500,
-  "global_step": 48,
+  "global_step": 39,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.06,
-      "grad_norm": 4.589228707438539,
-      "learning_rate": 4.000000000000001e-06,
-      "loss": 0.4768,
+      "epoch": 0.07,
+      "grad_norm": 4.349675350321153,
+      "learning_rate": 5e-06,
+      "loss": 0.4569,
       "step": 1
     },
     {
-      "epoch": 0.31,
-      "grad_norm": 4.436834276153056,
-      "learning_rate": 2e-05,
-      "loss": 0.4474,
+      "epoch": 0.36,
+      "grad_norm": 4.161390499285026,
+      "learning_rate": 1.9959742939952393e-05,
+      "loss": 0.4656,
       "step": 5
     },
     {
-      "epoch": 0.62,
-      "grad_norm": 3.3329122323772036,
-      "learning_rate": 1.9340161087325483e-05,
-      "loss": 0.5144,
+      "epoch": 0.73,
+      "grad_norm": 2.4085887995111475,
+      "learning_rate": 1.8584487936018663e-05,
+      "loss": 0.4982,
       "step": 10
     },
     {
-      "epoch": 0.94,
-      "grad_norm": 10.026674486358512,
-      "learning_rate": 1.744772182743782e-05,
-      "loss": 0.5168,
-      "step": 15
+      "epoch": 0.95,
+      "eval_loss": 1.4956705570220947,
+      "eval_runtime": 29.3808,
+      "eval_samples_per_second": 11.708,
+      "eval_steps_per_second": 0.749,
+      "step": 13
     },
     {
-      "epoch": 1.0,
-      "eval_loss": 1.5115139484405518,
-      "eval_runtime": 29.4255,
-      "eval_samples_per_second": 11.691,
-      "eval_steps_per_second": 0.748,
-      "step": 16
+      "epoch": 1.09,
+      "grad_norm": 2.520152396295157,
+      "learning_rate": 1.5508969814521026e-05,
+      "loss": 0.4421,
+      "step": 15
     },
     {
-      "epoch": 1.25,
-      "grad_norm": 2.2850083262335183,
-      "learning_rate": 1.4572423233046386e-05,
-      "loss": 0.3017,
+      "epoch": 1.45,
+      "grad_norm": 2.2813305230772256,
+      "learning_rate": 1.1342332658176556e-05,
+      "loss": 0.2341,
       "step": 20
     },
     {
-      "epoch": 1.56,
-      "grad_norm": 2.097044805034422,
-      "learning_rate": 1.1093712083778748e-05,
-      "loss": 0.235,
+      "epoch": 1.82,
+      "grad_norm": 3.7724059311334783,
+      "learning_rate": 6.909830056250527e-06,
+      "loss": 0.2254,
       "step": 25
     },
     {
-      "epoch": 1.88,
-      "grad_norm": 5.851950306103885,
-      "learning_rate": 7.470666176083193e-06,
-      "loss": 0.219,
-      "step": 30
+      "epoch": 1.96,
+      "eval_loss": 1.6817222833633423,
+      "eval_runtime": 29.2131,
+      "eval_samples_per_second": 11.776,
+      "eval_steps_per_second": 0.753,
+      "step": 27
     },
     {
-      "epoch": 2.0,
-      "eval_loss": 1.7232524156570435,
-      "eval_runtime": 29.1725,
-      "eval_samples_per_second": 11.792,
-      "eval_steps_per_second": 0.754,
-      "step": 32
+      "epoch": 2.18,
+      "grad_norm": 7.992205963709856,
+      "learning_rate": 3.089373510131354e-06,
+      "loss": 0.2145,
+      "step": 30
     },
     {
-      "epoch": 2.19,
-      "grad_norm": 1.585838384340871,
-      "learning_rate": 4.181410844420473e-06,
-      "loss": 0.1536,
+      "epoch": 2.55,
+      "grad_norm": 3.619511209880637,
+      "learning_rate": 6.37651293602628e-07,
+      "loss": 0.1095,
       "step": 35
     },
     {
-      "epoch": 2.5,
-      "grad_norm": 1.229574939486859,
-      "learning_rate": 1.660021821101222e-06,
-      "loss": 0.1024,
-      "step": 40
-    },
-    {
-      "epoch": 2.81,
-      "grad_norm": 1.0616061576941005,
-      "learning_rate": 2.392412244407294e-07,
-      "loss": 0.0956,
-      "step": 45
-    },
-    {
-      "epoch": 3.0,
-      "eval_loss": 1.9173061847686768,
-      "eval_runtime": 29.2012,
-      "eval_samples_per_second": 11.78,
-      "eval_steps_per_second": 0.753,
-      "step": 48
+      "epoch": 2.84,
+      "eval_loss": 1.767195224761963,
+      "eval_runtime": 29.1785,
+      "eval_samples_per_second": 11.789,
+      "eval_steps_per_second": 0.754,
+      "step": 39
     },
     {
-      "epoch": 3.0,
-      "step": 48,
-      "total_flos": 9997878558720.0,
-      "train_loss": 0.2755420195559661,
-      "train_runtime": 906.8406,
-      "train_samples_per_second": 3.374,
-      "train_steps_per_second": 0.053
+      "epoch": 2.84,
+      "step": 39,
+      "total_flos": 8113461657600.0,
+      "train_loss": 0.2916483015586168,
+      "train_runtime": 753.2022,
+      "train_samples_per_second": 3.453,
+      "train_steps_per_second": 0.052
     }
   ],
   "logging_steps": 5,
-  "max_steps": 48,
+  "max_steps": 39,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 3,
   "save_steps": 500,
-  "total_flos": 9997878558720.0,
+  "total_flos": 8113461657600.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d569194e0749bd93479d279be77149b330f099e0e858ac87e41547912ed5ba11
+oid sha256:f2ebc5d2a83d9f6df24e05e71dd49ede03c3dbec15879a7ca5386f4ec03d7609
 size 6200