Upload qwen-qwen2-5-3b-instruct checkpoint metadata
Browse files
checkpoints/qwen-qwen2-5-3b-instruct/sft_checkpoint.json
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"status": "ok",
|
| 3 |
+
"backend": "trl_transformers",
|
| 4 |
+
"examples_used": 2000,
|
| 5 |
+
"model_id": "Qwen/Qwen2.5-3B-Instruct",
|
| 6 |
+
"unsloth_available": false,
|
| 7 |
+
"train_runtime": 715.2908,
|
| 8 |
+
"train_loss": 0.15688225453009363,
|
| 9 |
+
"train_metrics": {
|
| 10 |
+
"train_runtime": 715.2908,
|
| 11 |
+
"train_samples_per_second": 5.592,
|
| 12 |
+
"train_steps_per_second": 2.796,
|
| 13 |
+
"total_flos": 2.949554402500608e+16,
|
| 14 |
+
"train_loss": 0.15688225453009363
|
| 15 |
+
},
|
| 16 |
+
"history_path": "/app/checkpoints/sweeps/qwen-qwen2-5-3b-instruct/sft_history.json",
|
| 17 |
+
"artifact_path": "/app/checkpoints/sweeps/qwen-qwen2-5-3b-instruct/sft_adapter"
|
| 18 |
+
}
|