update readme
Browse files- README.md +2 -0
- README_CN.md +2 -0
README.md
CHANGED
@@ -234,6 +234,8 @@ Generating images with 1K resolution will result in artifacts.
|
|
234 |
Additionally, we recommend using the full generation pipeline for better quality (i.e. enabling prompt enhancement and refinment).
|
235 |
|
236 |
```python
|
|
|
|
|
237 |
import torch
|
238 |
from hyimage.diffusion.pipelines.hunyuanimage_pipeline import HunyuanImagePipeline
|
239 |
|
|
|
234 |
Additionally, we recommend using the full generation pipeline for better quality (i.e. enabling prompt enhancement and refinment).
|
235 |
|
236 |
```python
|
237 |
+
import os
|
238 |
+
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
|
239 |
import torch
|
240 |
from hyimage.diffusion.pipelines.hunyuanimage_pipeline import HunyuanImagePipeline
|
241 |
|
README_CN.md
CHANGED
@@ -210,6 +210,8 @@ HunyuanImage-2.1 仅支持 2K 分辨率图像生成(如 1:1 时为 2048x2048
|
|
210 |
|
211 |
|
212 |
```python
|
|
|
|
|
213 |
import torch
|
214 |
from hyimage.diffusion.pipelines.hunyuanimage_pipeline import HunyuanImagePipeline
|
215 |
|
|
|
210 |
|
211 |
|
212 |
```python
|
213 |
+
import os
|
214 |
+
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
|
215 |
import torch
|
216 |
from hyimage.diffusion.pipelines.hunyuanimage_pipeline import HunyuanImagePipeline
|
217 |
|