Abe
commited on
Commit
·
d142a83
1
Parent(s):
1803c22
updated docs
Browse files- README.md +6 -1
- requirements.txt +3 -1
README.md
CHANGED
|
@@ -6,10 +6,15 @@ colorTo: yellow
|
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 5.29.0
|
| 8 |
app_file: app_hf.py
|
| 9 |
-
pinned:
|
| 10 |
license: other
|
| 11 |
---
|
| 12 |
|
| 13 |
Flex.2-Preview experiment by [JustLab.ai](https://justlab.ai)
|
| 14 |
|
| 15 |
Credits to [Ostris](https://huggingface.co/ostris) for this amazing all-in-one model
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 5.29.0
|
| 8 |
app_file: app_hf.py
|
| 9 |
+
pinned: true
|
| 10 |
license: other
|
| 11 |
---
|
| 12 |
|
| 13 |
Flex.2-Preview experiment by [JustLab.ai](https://justlab.ai)
|
| 14 |
|
| 15 |
Credits to [Ostris](https://huggingface.co/ostris) for this amazing all-in-one model
|
| 16 |
+
|
| 17 |
+
## Running locally.
|
| 18 |
+
use `app.py` which uses the `int8` version and `torchao` as recommended by Model creator.
|
| 19 |
+
For Windows you need to install `triton-windows` so uncomment in requirements.txt
|
| 20 |
+
I have tested on 24GB VRAM 3090 and 4090 and it works well. You can try with 16GBVRAM
|
requirements.txt
CHANGED
|
@@ -11,4 +11,6 @@ xformers>=0.0.20
|
|
| 11 |
huggingface-hub>=0.19.0
|
| 12 |
torchao>=0.1.0
|
| 13 |
spaces
|
| 14 |
-
sentencepiece
|
|
|
|
|
|
|
|
|
| 11 |
huggingface-hub>=0.19.0
|
| 12 |
torchao>=0.1.0
|
| 13 |
spaces
|
| 14 |
+
sentencepiece
|
| 15 |
+
### Windows
|
| 16 |
+
#triton-windows
|