Upload eval_deepscaler/deepscaler_eval_ckpt50/summary.json with huggingface_hub
Browse files
eval_deepscaler/deepscaler_eval_ckpt50/summary.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "Qwen/Qwen3-1.7B",
|
| 3 |
+
"checkpoint": "/workspace/outputs/hint_follow_lr5e5_250steps/checkpoint-50",
|
| 4 |
+
"num_problems": 100,
|
| 5 |
+
"num_generations_per_problem": 1,
|
| 6 |
+
"total_completions": 100,
|
| 7 |
+
"overall_accuracy": 0.89,
|
| 8 |
+
"mean_tokens": 2789.85,
|
| 9 |
+
"median_tokens": 2013,
|
| 10 |
+
"min_tokens": 590,
|
| 11 |
+
"max_tokens": 8192,
|
| 12 |
+
"std_tokens": 2061.3368253393232,
|
| 13 |
+
"frac_has_think": 0.93,
|
| 14 |
+
"frac_has_boxed": 0.93,
|
| 15 |
+
"difficulty_range": [
|
| 16 |
+
99.0,
|
| 17 |
+
100.0
|
| 18 |
+
],
|
| 19 |
+
"generation_params": {
|
| 20 |
+
"temperature": 0.6,
|
| 21 |
+
"top_p": 0.95,
|
| 22 |
+
"max_new_tokens": 8192
|
| 23 |
+
},
|
| 24 |
+
"backend": "vllm",
|
| 25 |
+
"timestamp": "2026-02-23T18:37:31.823210"
|
| 26 |
+
}
|