Skip to content

Commit e0fc7d4

Browse files
committed
chore: tidy wan2.2 docs and quant exports
Made-with: Cursor
1 parent 5a51ded commit e0fc7d4

3 files changed

Lines changed: 1 addition & 5 deletions

File tree

.gitignore

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -24,5 +24,4 @@ save*
2424
*.ipynb*
2525
model/
2626
output_*
27-
HiFloat4/
2827
datasets/

docs/wan2.2_quantization_guide.md

Lines changed: 0 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -43,8 +43,6 @@ quant:
4343
clip_sym: True
4444
```
4545
46-
首次使用需编译仓库内 hif4 GPU 扩展:`HiFloat4/hif4_gpu/build.sh`。
47-
4846
## 运行步骤
4947
5048
### 1. 环境
@@ -129,7 +127,6 @@ torchrun \
129127

130128
- **OOM**:减小 `sample_steps``num_frames`、分辨率;`bs: 1`;可开 `use_cpu_to_save_cuda_mem_for_catcher`
131129
- **无法 `import wan`**:安装官方仓库或配置 `wan2_repo_path`
132-
- **hif4 扩展编译失败**:核对 CUDA / PyTorch 与 `HiFloat4/hif4_gpu/build.sh` 日志。
133130
- **画质下降**:增加/多样化校准 prompt;在支持范围内微调 `special` 与校准规模。
134131

135132
## 参考

llmc/compression/quantization/__init__.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@
1010
from .ntweak import NormTweaking
1111
from .omniq import OmniQuant
1212
from .osplus import OsPlus
13-
from .quant import FloatQuantizer, HiFloat4Quantizer, IntegerQuantizer
13+
from .quant import FloatQuantizer, IntegerQuantizer
1414
from .quarot import Quarot
1515
from .quik import QUIK
1616
from .rtn import RTN

0 commit comments

Comments
 (0)