权重改为fp32, 其他部分保持bf16

Browse Source

目前参数--encoder_mode有三种选择:
1. fp32: 全部使用fp32, 适合显存充足的情况
2. autocast: 使用torch.cuda.amp.autocast自动混合精度, 稍微快一些, psnr下降较多
3. bf16_full: 全部使用bf16, 精度较高

...

This commit is contained in:

olivame

2026-01-18 20:24:37 +08:00

parent 44379f3e31

commit e1b029201e

4 changed files with 75 additions and 16 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

BIN

unitree_g1_pack_camera/case1/output/tensorboard/events.out.tfevents.1768734530.node-0.417718.0 Normal file

Binary file not shown.