Model save

README.md

@@ -2,7 +2,6 @@
 license: gemma
 base_model: scottsuk0306/easylm-sft-gemma-2-2b
 tags:
-- easylm
 - trl
 - reward-trainer
 - generated_from_trainer
@@ -16,13 +15,12 @@ model-index:
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/lklab_kaist/grm/runs/kp392ohf)
 # easylm-rm-gemma-2-2b
 
 This model is a fine-tuned version of [scottsuk0306/easylm-sft-gemma-2-2b](https://huggingface.co/scottsuk0306/easylm-sft-gemma-2-2b) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.7170
-- Accuracy: 0.6042
+- Loss: 6.0469
+- Accuracy: 0.5833
 
 ## Model description
 
@@ -57,14 +55,14 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss | Accuracy |
 |:-------------:|:-----:|:----:|:---------------:|:--------:|
-| No log        | 1.0   | 25   | 0.6833          | 0.5417   |
-| No log        | 2.0   | 50   | 1.2481          | 0.5625   |
-| No log        | 3.0   | 75   | 1.7170          | 0.6042   |
+| 0.6795        | 1.0   | 2399 | 0.7247          | 0.5938   |
+| 0.4438        | 2.0   | 4798 | 2.9242          | 0.5938   |
+| 0.0414        | 3.0   | 7197 | 6.0469          | 0.5833   |
 
 
 ### Framework versions
 
-- Transformers 4.42.4
+- Transformers 4.43.3
 - Pytorch 2.3.0+cu121
 - Datasets 2.20.0
 - Tokenizers 0.19.1

all_results.json

@@ -1,15 +1,9 @@
 {
     "epoch": 3.0,
-    "eval_accuracy": 0.6041666666666666,
-    "eval_loss": 1.7170292139053345,
-    "eval_runtime": 1.6843,
-    "eval_samples": 96,
-    "eval_samples_per_second": 56.996,
-    "eval_steps_per_second": 14.249,
     "total_flos": 0.0,
-    "train_loss": 0.5482939656575521,
-    "train_runtime": 190.4657,
-    "train_samples": 100,
-    "train_samples_per_second": 1.575,
-    "train_steps_per_second": 0.394
+    "train_loss": 0.38497807171802156,
+    "train_runtime": 4417.8859,
+    "train_samples": 9595,
+    "train_samples_per_second": 6.516,
+    "train_steps_per_second": 1.629
 }

config.json

@@ -33,7 +33,7 @@
   "rope_theta": 10000.0,
   "sliding_window": 4096,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.42.4",
+  "transformers_version": "4.43.3",
   "use_cache": true,
   "vocab_size": 256000
 }

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b00ac57221c6afcf0c9d6b84dd92b855c0fde0c5bef251a4bf68be9462cf5f8a
+oid sha256:ebdc412ae8e6303311b7295afc2426709718573671e7dc3fbbb856569273e82b
 size 4988025760

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7dc4b3fadae1742a61cbcb9c8f61144918b5b86476fb7e673979b4670ee6502
+oid sha256:ab88370ef0a7abfcf32f7b00c3788ae7e30f00749c2146b0fafdb348c75eecc2
 size 240696424

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 3.0,
     "total_flos": 0.0,
-    "train_loss": 0.5482939656575521,
-    "train_runtime": 190.4657,
-    "train_samples": 100,
-    "train_samples_per_second": 1.575,
-    "train_steps_per_second": 0.394
+    "train_loss": 0.38497807171802156,
+    "train_runtime": 4417.8859,
+    "train_samples": 9595,
+    "train_samples_per_second": 6.516,
+    "train_steps_per_second": 1.629
 }

trainer_state.json

@@ -3,50 +3,148 @@
   "best_model_checkpoint": null,
   "epoch": 3.0,
   "eval_steps": 500,
-  "global_step": 75,
+  "global_step": 7197,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
+    {
+      "epoch": 0.20842017507294705,
+      "grad_norm": 7.509331703186035,
+      "learning_rate": 2.9644144799482153e-06,
+      "loss": 0.7098,
+      "step": 500
+    },
+    {
+      "epoch": 0.4168403501458941,
+      "grad_norm": 5.73619270324707,
+      "learning_rate": 2.859346358776001e-06,
+      "loss": 0.6948,
+      "step": 1000
+    },
+    {
+      "epoch": 0.6252605252188412,
+      "grad_norm": 7.396816253662109,
+      "learning_rate": 2.689780841460394e-06,
+      "loss": 0.6855,
+      "step": 1500
+    },
+    {
+      "epoch": 0.8336807002917882,
+      "grad_norm": 9.113743782043457,
+      "learning_rate": 2.4637633641567626e-06,
+      "loss": 0.6795,
+      "step": 2000
+    },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.5416666666666666,
-      "eval_loss": 0.6833394169807434,
-      "eval_runtime": 1.6468,
-      "eval_samples_per_second": 58.294,
-      "eval_steps_per_second": 14.574,
-      "step": 25
+      "eval_accuracy": 0.59375,
+      "eval_loss": 0.7246907353401184,
+      "eval_runtime": 1.6903,
+      "eval_samples_per_second": 56.793,
+      "eval_steps_per_second": 14.198,
+      "step": 2399
+    },
+    {
+      "epoch": 1.0421008753647354,
+      "grad_norm": 33.38862609863281,
+      "learning_rate": 2.1920178594926297e-06,
+      "loss": 0.6189,
+      "step": 2500
+    },
+    {
+      "epoch": 1.2505210504376825,
+      "grad_norm": 3.111920118331909,
+      "learning_rate": 1.8874379342749385e-06,
+      "loss": 0.4549,
+      "step": 3000
+    },
+    {
+      "epoch": 1.4589412255106295,
+      "grad_norm": 0.5629640221595764,
+      "learning_rate": 1.5644751018852966e-06,
+      "loss": 0.4627,
+      "step": 3500
+    },
+    {
+      "epoch": 1.6673614005835766,
+      "grad_norm": 0.07236230373382568,
+      "learning_rate": 1.2384530961210145e-06,
+      "loss": 0.4569,
+      "step": 4000
+    },
+    {
+      "epoch": 1.8757815756565237,
+      "grad_norm": 0.02289917692542076,
+      "learning_rate": 9.24840800480023e-07,
+      "loss": 0.4438,
+      "step": 4500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.5625,
-      "eval_loss": 1.2481266260147095,
-      "eval_runtime": 1.672,
-      "eval_samples_per_second": 57.415,
-      "eval_steps_per_second": 14.354,
-      "step": 50
+      "eval_accuracy": 0.59375,
+      "eval_loss": 2.9241831302642822,
+      "eval_runtime": 1.6517,
+      "eval_samples_per_second": 58.123,
+      "eval_steps_per_second": 14.531,
+      "step": 4798
+    },
+    {
+      "epoch": 2.084201750729471,
+      "grad_norm": 4.127718925476074,
+      "learning_rate": 6.385182904756809e-07,
+      "loss": 0.2538,
+      "step": 5000
+    },
+    {
+      "epoch": 2.292621925802418,
+      "grad_norm": 1.510969994955591e-11,
+      "learning_rate": 3.93070813336037e-07,
+      "loss": 0.0133,
+      "step": 5500
+    },
+    {
+      "epoch": 2.501042100875365,
+      "grad_norm": 3.9616837588241083e-10,
+      "learning_rate": 2.001442038869754e-07,
+      "loss": 0.0213,
+      "step": 6000
+    },
+    {
+      "epoch": 2.7094622759483116,
+      "grad_norm": 1.3078284837320098e-06,
+      "learning_rate": 6.889232043392568e-08,
+      "loss": 0.0046,
+      "step": 6500
+    },
+    {
+      "epoch": 2.917882451021259,
+      "grad_norm": 3.3587113534849777e-07,
+      "learning_rate": 5.5427183508257595e-09,
+      "loss": 0.0414,
+      "step": 7000
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.6041666666666666,
-      "eval_loss": 1.7170292139053345,
-      "eval_runtime": 1.7763,
-      "eval_samples_per_second": 54.045,
-      "eval_steps_per_second": 13.511,
-      "step": 75
+      "eval_accuracy": 0.5833333333333334,
+      "eval_loss": 6.046913146972656,
+      "eval_runtime": 1.6474,
+      "eval_samples_per_second": 58.275,
+      "eval_steps_per_second": 14.569,
+      "step": 7197
     },
     {
       "epoch": 3.0,
-      "step": 75,
+      "step": 7197,
       "total_flos": 0.0,
-      "train_loss": 0.5482939656575521,
-      "train_runtime": 190.4657,
-      "train_samples_per_second": 1.575,
-      "train_steps_per_second": 0.394
+      "train_loss": 0.38497807171802156,
+      "train_runtime": 4417.8859,
+      "train_samples_per_second": 6.516,
+      "train_steps_per_second": 1.629
     }
   ],
   "logging_steps": 500,
-  "max_steps": 75,
+  "max_steps": 7197,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 3,
   "save_steps": 500,
@@ -56,8 +154,8 @@
         "should_epoch_stop": false,
         "should_evaluate": false,
         "should_log": false,
-        "should_save": true,
-        "should_training_stop": true
+        "should_save": false,
+        "should_training_stop": false
       },
       "attributes": {}
     }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17509e8d66be67771ace016820380128ce14fad5edf7c3939a29014c02226a57
-size 6200
+oid sha256:b3a81758449ae164f1bdd71e177b28bb7eae16b05cdabb93ec9211a34cf70b14
+size 6264