Upload exp_phase7_moe_joint_T1T2T3_r64_20260502_203210/manifest.json with huggingface_hub
Browse files
exp_phase7_moe_joint_T1T2T3_r64_20260502_203210/manifest.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"trainer": "phase7/train_moe_joint.py",
|
| 3 |
+
"llm_name": "Qwen/Qwen3.5-0.8B",
|
| 4 |
+
"lora_r": 64,
|
| 5 |
+
"task_counts": {
|
| 6 |
+
"enhancer_generation": 13098,
|
| 7 |
+
"pair_prediction": 11532,
|
| 8 |
+
"enhancer_editing": 6268
|
| 9 |
+
},
|
| 10 |
+
"n_train_rows": 30898,
|
| 11 |
+
"trainable_params": 12230176,
|
| 12 |
+
"num_steps": 6000,
|
| 13 |
+
"peak_lr": 2e-05,
|
| 14 |
+
"reasoning_loss_weight": 1.0,
|
| 15 |
+
"regression_loss_weight": 1.0,
|
| 16 |
+
"combined_jsonl": "/workspace/dnathinker/data/multiturn_dataset/combined_t1t2t3_for_moe.jsonl"
|
| 17 |
+
}
|