End of training

README.md

@@ -1,17 +1,19 @@
 ---
 base_model: sohyunan/gemma-2-2b-it_controller
+datasets: maze_5x5
 library_name: transformers
-model_name: gemma-2-2b-it_controller-grpo
+model_name: sohyunan/gemma-2-2b-it_controller-grpo
 tags:
 - generated_from_trainer
+- controller-grpo
 - trl
 - grpo
 licence: license
 ---
 
-# Model Card for gemma-2-2b-it_controller-grpo
+# Model Card for sohyunan/gemma-2-2b-it_controller-grpo
 
-This model is a fine-tuned version of [sohyunan/gemma-2-2b-it_controller](https://huggingface.co/sohyunan/gemma-2-2b-it_controller).
+This model is a fine-tuned version of [sohyunan/gemma-2-2b-it_controller](https://huggingface.co/sohyunan/gemma-2-2b-it_controller) on the [maze_5x5](https://huggingface.co/datasets/maze_5x5) dataset.
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start

all_results.json

@@ -1,4 +1,9 @@
 {
+    "eval_loss": 0.026073778048157692,
+    "eval_runtime": 1435.6116,
+    "eval_samples": 400,
+    "eval_samples_per_second": 0.279,
+    "eval_steps_per_second": 0.017,
     "total_flos": 0.0,
     "train_loss": 3628.8164483806245,
     "train_runtime": 61845.1528,

config.json

@@ -31,6 +31,6 @@
   "sliding_window": 4096,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.49.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 256001
 }

eval_results.json

@@ -1,7 +1,7 @@
 {
-    "eval_loss": -1.0927518445669193e-08,
-    "eval_runtime": 263.0804,
-    "eval_samples": 30,
-    "eval_samples_per_second": 0.114,
-    "eval_steps_per_second": 0.011
+    "eval_loss": 0.026073778048157692,
+    "eval_runtime": 1435.6116,
+    "eval_samples": 400,
+    "eval_samples_per_second": 0.279,
+    "eval_steps_per_second": 0.017
 }