Not use flashattn
Browse files- app.py +0 -1
- requirements.txt +0 -1
app.py
CHANGED
|
@@ -71,7 +71,6 @@ def get_interface(model_name: str):
|
|
| 71 |
backend=outetts.Backend.HF,
|
| 72 |
additional_model_config={
|
| 73 |
"device_map": "auto" if has_cuda else "cpu",
|
| 74 |
-
"attn_implementation": "flash_attention_2" if has_cuda else "eager",
|
| 75 |
"quantization_config": BitsAndBytesConfig(
|
| 76 |
load_in_8bit=True
|
| 77 |
) if has_cuda else None,
|
|
|
|
| 71 |
backend=outetts.Backend.HF,
|
| 72 |
additional_model_config={
|
| 73 |
"device_map": "auto" if has_cuda else "cpu",
|
|
|
|
| 74 |
"quantization_config": BitsAndBytesConfig(
|
| 75 |
load_in_8bit=True
|
| 76 |
) if has_cuda else None,
|
requirements.txt
CHANGED
|
@@ -3,7 +3,6 @@
|
|
| 3 |
llama-cpp-python
|
| 4 |
outetts@git+https://github.com/mybigday/OuteTTS.git@fix
|
| 5 |
accelerate
|
| 6 |
-
flash-attn
|
| 7 |
absl-py==2.3.1
|
| 8 |
aiofiles==24.1.0
|
| 9 |
aiohappyeyeballs==2.6.1
|
|
|
|
| 3 |
llama-cpp-python
|
| 4 |
outetts@git+https://github.com/mybigday/OuteTTS.git@fix
|
| 5 |
accelerate
|
|
|
|
| 6 |
absl-py==2.3.1
|
| 7 |
aiofiles==24.1.0
|
| 8 |
aiohappyeyeballs==2.6.1
|