Upload training_info.json with huggingface_hub
Browse files- training_info.json +32 -0
training_info.json
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name": "MedGemma3N-chinese-qa-v1",
|
| 3 |
+
"version": "v1.0 (第一阶段基础版本)",
|
| 4 |
+
"base_model": "unsloth/gemma-3n-E4B-it",
|
| 5 |
+
"training_date": "2025-07-27T04:22:09.329107",
|
| 6 |
+
"training_data": {
|
| 7 |
+
"total_samples": 13153,
|
| 8 |
+
"data_sources": {
|
| 9 |
+
"huatuo_encyclopedia_qa": 7960,
|
| 10 |
+
"medical_o1_reasoning": 2924,
|
| 11 |
+
"huatuo_knowledge_graph": 2269
|
| 12 |
+
}
|
| 13 |
+
},
|
| 14 |
+
"training_config": {
|
| 15 |
+
"lora_rank": 16,
|
| 16 |
+
"lora_alpha": 32,
|
| 17 |
+
"learning_rate": 0.0002,
|
| 18 |
+
"max_steps": 8000,
|
| 19 |
+
"batch_size": 4,
|
| 20 |
+
"optimizer": "adamw_8bit"
|
| 21 |
+
},
|
| 22 |
+
"performance": {
|
| 23 |
+
"data_expansion": "4.7x",
|
| 24 |
+
"training_time": "~11 hours",
|
| 25 |
+
"gpu_used": "NVIDIA A100-SXM4-40GB"
|
| 26 |
+
},
|
| 27 |
+
"next_phase": {
|
| 28 |
+
"planned_version": "v2.0",
|
| 29 |
+
"planned_data": "85,000条专业临床数据",
|
| 30 |
+
"enhancement_focus": "临床推理、诊断思维、医患对话"
|
| 31 |
+
}
|
| 32 |
+
}
|