Model save

README.md

@@ -26,7 +26,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/aaron-weihuang-the-university-of-hong-kong/huggingface/runs/bc8byckb) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/aaron-weihuang-the-university-of-hong-kong/huggingface/runs/kz4f7c4b) 
 
 
 This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 5.110737800786022,
-    "train_runtime": 360798.095,
+    "train_loss": 1.286716509427883,
+    "train_runtime": 229250.8975,
     "train_samples": 72441,
-    "train_samples_per_second": 0.201,
-    "train_steps_per_second": 0.004
+    "train_samples_per_second": 0.316,
+    "train_steps_per_second": 0.003
 }

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/mnt/sdd/hw/LLM/Qwen2.5-3B-Instruct",
+  "_name_or_path": "/mnt/sdd/hw/LLM/RL_R1/Qwen2.5-3B-Open-R1-Distill",
   "architectures": [
     "Qwen2ForCausalLM"
   ],

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c7ae80f504ae3d3fb30365724bef62927f9b11f7d80b9cf445f2f2536992a582
+oid sha256:c8042c984b33d36c39d9b0062e02b1a6d51b81a45d804cf42edb8f7b41885e0b
 size 4957560304

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d0f36b69207d23b20c4a6561fef2b45f09a3536f21810fc06967327f352246b
+oid sha256:60b3d6a0a7c0dcf0c59c2d1ba5eb9230884e032fb50c29dab30ce73f6dabc6ec
 size 1836696752

special_tokens_map.json

@@ -22,7 +22,7 @@
     "single_word": false
   },
   "pad_token": {
-    "content": "<|endoftext|>",
+    "content": "<|im_end|>",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5eee858c5123a4279c3e1f7b81247343f356ac767940b2692a928ad929543214
-size 11422063
+oid sha256:63a2951d5edfa5cc0a2346ef872f8c77a2920274cfc3b503b04e3799104dee80
+size 11422060

tokenizer_config.json

@@ -201,7 +201,7 @@
   "errors": "replace",
   "extra_special_tokens": {},
   "model_max_length": 131072,
-  "pad_token": "<|endoftext|>",
+  "pad_token": "<|im_end|>",
   "padding_side": "left",
   "split_special_tokens": false,
   "tokenizer_class": "Qwen2Tokenizer",

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 5.110737800786022,
-    "train_runtime": 360798.095,
+    "train_loss": 1.286716509427883,
+    "train_runtime": 229250.8975,
     "train_samples": 72441,
-    "train_samples_per_second": 0.201,
-    "train_steps_per_second": 0.004
+    "train_samples_per_second": 0.316,
+    "train_steps_per_second": 0.003
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0c295910081897b75568d50ca5708994c1a9ff69694cb44627d289e2d19a7430
+oid sha256:6fd60a654e558783f388f58c0cb2053266bcfc3ded9151cd548c76d7bbe00563
 size 7544