Spaces:
Running
on
Zero
Running
on
Zero
Add variant argument to support loading fp16 model
Browse files- app.py +1 -0
- pipeline_objectclear.py +2 -0
app.py
CHANGED
@@ -283,6 +283,7 @@ device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
|
283 |
pipe = ObjectClearPipeline.from_pretrained_with_custom_modules(
|
284 |
"jixin0101/ObjectClear",
|
285 |
torch_dtype=torch.float16,
|
|
|
286 |
save_cross_attn=True,
|
287 |
cache_dir="/home/jovyan/shared/jixinzhao/models",
|
288 |
)
|
|
|
283 |
pipe = ObjectClearPipeline.from_pretrained_with_custom_modules(
|
284 |
"jixin0101/ObjectClear",
|
285 |
torch_dtype=torch.float16,
|
286 |
+
variant='fp16',
|
287 |
save_cross_attn=True,
|
288 |
cache_dir="/home/jovyan/shared/jixinzhao/models",
|
289 |
)
|
pipeline_objectclear.py
CHANGED
@@ -469,6 +469,7 @@ class ObjectClearPipeline(
|
|
469 |
pretrained_model_name_or_path,
|
470 |
torch_dtype=torch.float32,
|
471 |
cache_dir=None,
|
|
|
472 |
**kwargs,
|
473 |
):
|
474 |
from safetensors.torch import load_file
|
@@ -497,6 +498,7 @@ class ObjectClearPipeline(
|
|
497 |
image_prompt_encoder=image_prompt_encoder,
|
498 |
postfuse_module=postfuse_module,
|
499 |
cache_dir=cache_dir,
|
|
|
500 |
**kwargs,
|
501 |
)
|
502 |
|
|
|
469 |
pretrained_model_name_or_path,
|
470 |
torch_dtype=torch.float32,
|
471 |
cache_dir=None,
|
472 |
+
variant=None,
|
473 |
**kwargs,
|
474 |
):
|
475 |
from safetensors.torch import load_file
|
|
|
498 |
image_prompt_encoder=image_prompt_encoder,
|
499 |
postfuse_module=postfuse_module,
|
500 |
cache_dir=cache_dir,
|
501 |
+
variant=variant,
|
502 |
**kwargs,
|
503 |
)
|
504 |
|