Spaces:
Paused
Paused
Kunpeng Song
commited on
Commit
•
338f71e
1
Parent(s):
8a4a948
float16
Browse files- app.py +1 -1
- model_lib/moMA_generator.py +2 -2
- model_lib/modules.py +2 -2
app.py
CHANGED
@@ -28,7 +28,7 @@ def inference(rgb, mask, subject, prompt):
|
|
28 |
seed_everything(0)
|
29 |
args = parse_args()
|
30 |
#load MoMA from HuggingFace. Auto download
|
31 |
-
model = MoMA_main_modal(args).to(args.device, dtype=torch.
|
32 |
|
33 |
|
34 |
################ change texture ##################
|
|
|
28 |
seed_everything(0)
|
29 |
args = parse_args()
|
30 |
#load MoMA from HuggingFace. Auto download
|
31 |
+
model = MoMA_main_modal(args).to(args.device, dtype=torch.float16)
|
32 |
|
33 |
|
34 |
################ change texture ##################
|
model_lib/moMA_generator.py
CHANGED
@@ -93,7 +93,7 @@ class MoMA_generator:
|
|
93 |
print('Loading StableDiffusion: Realistic_Vision...')
|
94 |
self.pipe = StableDiffusionPipeline.from_pretrained(
|
95 |
"SG161222/Realistic_Vision_V4.0_noVAE",
|
96 |
-
torch_dtype=torch.
|
97 |
scheduler=noise_scheduler,
|
98 |
vae=vae,
|
99 |
feature_extractor=None,
|
@@ -112,7 +112,7 @@ class MoMA_generator:
|
|
112 |
cross_attention_dim=768,
|
113 |
clip_embeddings_dim=1024,
|
114 |
clip_extra_context_tokens=4,
|
115 |
-
).to(self.device, dtype=torch.
|
116 |
return image_proj_model
|
117 |
|
118 |
def set_ip_adapter(self):
|
|
|
93 |
print('Loading StableDiffusion: Realistic_Vision...')
|
94 |
self.pipe = StableDiffusionPipeline.from_pretrained(
|
95 |
"SG161222/Realistic_Vision_V4.0_noVAE",
|
96 |
+
torch_dtype=torch.float16,
|
97 |
scheduler=noise_scheduler,
|
98 |
vae=vae,
|
99 |
feature_extractor=None,
|
|
|
112 |
cross_attention_dim=768,
|
113 |
clip_embeddings_dim=1024,
|
114 |
clip_extra_context_tokens=4,
|
115 |
+
).to(self.device, dtype=torch.float16)
|
116 |
return image_proj_model
|
117 |
|
118 |
def set_ip_adapter(self):
|
model_lib/modules.py
CHANGED
@@ -87,7 +87,7 @@ class MoMA_main_modal(nn.Module):
|
|
87 |
|
88 |
add_function(self.model_llava)
|
89 |
|
90 |
-
self.mapping = LlamaMLP_mapping(4096,1024).to(self.device, dtype=torch.
|
91 |
self.load_saved_components()
|
92 |
self.freeze_modules()
|
93 |
|
@@ -140,7 +140,7 @@ class MoMA_main_modal(nn.Module):
|
|
140 |
batch = Dataset_evaluate_MoMA(rgb_path, prompt, subject, mask_path,self)
|
141 |
self.moMA_generator.set_selfAttn_strength(strength)
|
142 |
|
143 |
-
with torch.cuda.amp.autocast(enabled=True, dtype=torch.
|
144 |
with torch.no_grad():
|
145 |
### key steps
|
146 |
llava_emb = self.forward_MLLM(batch).clone().detach()
|
|
|
87 |
|
88 |
add_function(self.model_llava)
|
89 |
|
90 |
+
self.mapping = LlamaMLP_mapping(4096,1024).to(self.device, dtype=torch.float16)
|
91 |
self.load_saved_components()
|
92 |
self.freeze_modules()
|
93 |
|
|
|
140 |
batch = Dataset_evaluate_MoMA(rgb_path, prompt, subject, mask_path,self)
|
141 |
self.moMA_generator.set_selfAttn_strength(strength)
|
142 |
|
143 |
+
with torch.cuda.amp.autocast(enabled=True, dtype=torch.float16, cache_enabled=True):
|
144 |
with torch.no_grad():
|
145 |
### key steps
|
146 |
llava_emb = self.forward_MLLM(batch).clone().detach()
|