JiaruiLi/DeepHealth
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Log and save model parameter counts

Browse Source
This commit is contained in:
Jiarui Li
2026-07-22 14:44:31 +08:00
parent db0947ce9d
commit 978c88a4ed
5 changed files with 50 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
TrajMixer_设计方案.md
View File

@@ -319,7 +319,7 @@ output_init_std: 0.001
group_wise_layer_norm: false
```
训练时必须将 `model_architecture: traj_mixer_v1` 写入 `train_config.json`。本分支的评估和导出入口只接受带有该标识、且 checkpoint 中包含 TrajMixer 参数张量的模型；其他分支生成的模型应直接拒绝加载。
训练时必须将 `model_architecture: traj_mixer_v1`、`model_parameter_count` 和 `trainable_parameter_count` 写入 `train_config.json`，并在训练日志中显式打印总参数量与可训练参数量。本分支的评估和导出入口只接受带有该架构标识、且 checkpoint 中包含 TrajMixer 参数张量的模型；其他分支生成的模型应直接拒绝加载。
必须满足：