Model save

README.md

@@ -1,11 +1,9 @@
 ---
 base_model: Qwen/Qwen2.5-1.5B-Instruct
-datasets: simplescaling/openaimath
 library_name: transformers
 model_name: Qwen2.5-1.5B-hl-true-v6
 tags:
 - generated_from_trainer
-- open-r1
 - trl
 - grpo
 licence: license
@@ -13,7 +11,7 @@ licence: license
 
 # Model Card for Qwen2.5-1.5B-hl-true-v6
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct) on the [simplescaling/openaimath](https://huggingface.co/datasets/simplescaling/openaimath) dataset.
+This model is a fine-tuned version of [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -29,7 +27,7 @@ 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/muennighoff/halos/runs/n0pjcped) 
+[<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/muennighoff/halos/runs/5v4dg2rh) 
 
 
 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": -0.036251500838914125,
-    "train_runtime": 129592.8422,
+    "train_loss": -2.503701759680296e-05,
+    "train_runtime": 134.9948,
     "train_samples": 12000,
-    "train_samples_per_second": 11.062,
-    "train_steps_per_second": 0.012
+    "train_samples_per_second": 10619.672,
+    "train_steps_per_second": 11.852
 }

generation_config.json

@@ -1,10 +1,7 @@
 {
   "bos_token_id": 151643,
   "do_sample": true,
-  "eos_token_id": [
-    151645,
-    151643
-  ],
+  "eos_token_id": 151645,
   "pad_token_id": 151643,
   "repetition_penalty": 1.1,
   "temperature": 0.7,

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": -0.036251500838914125,
-    "train_runtime": 129592.8422,
+    "train_loss": -2.503701759680296e-05,
+    "train_runtime": 134.9948,
     "train_samples": 12000,
-    "train_samples_per_second": 11.062,
-    "train_steps_per_second": 0.012
+    "train_samples_per_second": 10619.672,
+    "train_steps_per_second": 11.852
 }

trainer_state.json

@@ -2,9 +2,9 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 14.951603498542275,
+  "epoch": 3.7395742198891804,
   "eval_steps": 500,
-  "global_step": 1600,
+  "global_step": 1601,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -38410,19 +38410,43 @@
       "step": 1600
     },
     {
-      "epoch": 14.951603498542275,
-      "step": 1600,
+      "clip_ratio/high_max": 0.04218487173784524,
+      "clip_ratio/high_mean": 0.02072980930097401,
+      "clip_ratio/low_mean": 0.0005389717216530698,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.021268780867103487,
+      "completions/clipped_ratio": 0.1037946428571429,
+      "completions/max_length": 4096.0,
+      "completions/max_terminated_length": 2385.0,
+      "completions/mean_length": 866.5313110351562,
+      "completions/mean_terminated_length": 492.5080871582031,
+      "completions/min_length": 106.0,
+      "completions/min_terminated_length": 106.0,
+      "epoch": 3.7395742198891804,
+      "grad_norm": 0.23319926857948303,
+      "learning_rate": 1e-06,
+      "loss": -0.0401,
+      "num_tokens": 871906108.0,
+      "reward": 0.7433035969734192,
+      "reward_std": 0.14162540435791016,
+      "rewards/verify_math_reward/mean": 0.7433035969734192,
+      "rewards/verify_math_reward/std": 0.43705445528030396,
+      "step": 1601
+    },
+    {
+      "epoch": 3.7395742198891804,
+      "step": 1601,
       "total_flos": 0.0,
-      "train_loss": -0.036251500838914125,
-      "train_runtime": 129592.8422,
-      "train_samples_per_second": 11.062,
-      "train_steps_per_second": 0.012
+      "train_loss": -2.503701759680296e-05,
+      "train_runtime": 134.9948,
+      "train_samples_per_second": 10619.672,
+      "train_steps_per_second": 11.852
     }
   ],
   "logging_steps": 1,
   "max_steps": 1600,
-  "num_input_tokens_seen": 871420576,
-  "num_train_epochs": 15,
+  "num_input_tokens_seen": 871906108,
+  "num_train_epochs": 4,
   "save_steps": 160,
   "stateful_callbacks": {
     "TrainerControl": {