End of training

README.md

@@ -1,9 +1,12 @@
 ---
 base_model: Qwen/Qwen2.5-3B-Instruct
+datasets: atomwalk12/linalgzero-distilled
 library_name: transformers
 model_name: Qwen2.5-3B-Instruct-SFT
 tags:
 - generated_from_trainer
+- tool-use
+- linalg-zero
 - sft
 - trl
 licence: license
@@ -11,7 +14,7 @@ licence: license
 
 # Model Card for Qwen2.5-3B-Instruct-SFT
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).
+This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) on the [atomwalk12/linalgzero-distilled](https://huggingface.co/datasets/atomwalk12/linalgzero-distilled) dataset.
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start

all_results.json

@@ -1,4 +1,18 @@
 {
+    "diagnostic/answer_attempted_ratio": 0.0,
+    "diagnostic/avg_turns_taken": 1.0,
+    "diagnostic/format_valid_ratio": 0.1,
+    "diagnostic/stop_reason/extraction_failed": 0.9,
+    "diagnostic/stop_reason/no_action": 0.1,
+    "diagnostic/tool_parse_success_ratio": 0.0,
+    "eval_loss": 0.4209205210208893,
+    "eval_runtime": 4.5445,
+    "eval_samples": 20,
+    "eval_samples_per_second": 4.401,
+    "eval_steps_per_second": 2.2,
+    "reward_final_answer": 0.0,
+    "reward_interaction": 0.0,
+    "reward_response_format": 0.0,
     "total_flos": 5.5251981578153165e+17,
     "train_loss": 0.16806216436064142,
     "train_runtime": 6592.4366,

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "diagnostic/answer_attempted_ratio": 0.0,
+    "diagnostic/avg_turns_taken": 1.0,
+    "diagnostic/format_valid_ratio": 0.1,
+    "diagnostic/stop_reason/extraction_failed": 0.9,
+    "diagnostic/stop_reason/no_action": 0.1,
+    "diagnostic/tool_parse_success_ratio": 0.0,
+    "eval_loss": 0.4209205210208893,
+    "eval_runtime": 4.5445,
+    "eval_samples": 20,
+    "eval_samples_per_second": 4.401,
+    "eval_steps_per_second": 2.2,
+    "reward_final_answer": 0.0,
+    "reward_interaction": 0.0,
+    "reward_response_format": 0.0
+}