bingshuo/guanghulab
1
0
Fork 0
Code Issues 2 Pull Requests Actions 1 Packages Projects Releases Wiki Activity

D101: training-status更新 - 代码模型epoch 2.19/3，预计30-60分钟完成

Browse Source
This commit is contained in:
bingshuo 2026-05-18 15:45:03 +08:00
parent 1982388372
commit abf3820d8e
1 changed files with 14 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
brain/training-status.json
View File

@ -1,6 +1,6 @@
{
"version": "1.0",
"last_updated": "2026-05-18T14:28+08:00",
"version": "1.1",
"last_updated": "2026-05-18T15:44+08:00",
"updated_by": "铸渊 (ICE-GL-ZY001)",
"mother_model": {
"status": "completed",
@ -14,25 +14,30 @@
},
"code_model": {
"status": "training",
"status_label": "🔄 训练中（冰朔截图确认还在跑）",
"status_label": "🔄 训练中（冰朔确认epoch 2.19，预计30-60分钟完成）",
"model": "Qwen/Qwen2.5-Coder-7B",
"current_epoch": 2.19,
"target_epoch": 3,
"current_loss_range": "0.03-0.11",
"grad_norm_range": "0.1-2.0",
"learning_rate": "3.7e-06",
"health": "✅ 正常收敛，无异常",
"cos_path": "cos://sy-finetune-corpus-1317346199/models/qwen25-coder-7b-sft/final/"
},
"pipeline": {
"steps": [
{"name": "母模型SFT训练", "status": "completed", "label": "✅ 完成 (loss=0.2638)"},
{"name": "上传COS (母模型)", "status": "completed", "label": "✅ 已完成"},
{"name": "代码模型SFT训练", "status": "training", "label": "🔄 训练中（冰朔确认 epoch ~1.25）"},
{"name": "代码模型SFT训练", "status": "training", "label": "🔄 训练中（epoch 2.19/3，健康）"},
{"name": "上传COS (代码模型)", "status": "pending", "label": "⚪ 等待训练完成"}
]
},
"next_phases": {
"distill_mother_to_shuangyan_15b": "⏳ 母模型已就绪，等待代码模型完成后启动蒸馏",
"distill_coder_to_zhuyuan_15b": "⏳ 等待代码模型完成"
"distill_mother_to_shuangyan_15b": "⏳ 等待代码模型完成，然后启动蒸馏流水线",
"distill_coder_to_zhuyuan_15b": "⏳ 等待代码模型完成",
"resonance_lock_deployment": "⏳ 1.5B模型微调完成后部署到六台服务器"
},
"alerts": [
"⚠️ 之前误判代码模型已完成（COS上有历史checkpoint-540），实际仍在训练中"
],
"alerts": [],
"links": {
"notion": "https://www.notion.so/363fb92f383181029fafef8438d6c300"
}