Spaces:
Running
on
Zero
Running
on
Zero
Update ootd/inference_ootd.py
Browse files- ootd/inference_ootd.py +7 -7
ootd/inference_ootd.py
CHANGED
|
@@ -33,7 +33,7 @@ MODEL_PATH = "./checkpoints/ootd"
|
|
| 33 |
class OOTDiffusion:
|
| 34 |
|
| 35 |
def __init__(self, gpu_id):
|
| 36 |
-
self.gpu_id = 'cuda:' + str(gpu_id)
|
| 37 |
|
| 38 |
vae = AutoencoderKL.from_pretrained(
|
| 39 |
VAE_PATH,
|
|
@@ -64,12 +64,12 @@ class OOTDiffusion:
|
|
| 64 |
use_safetensors=True,
|
| 65 |
safety_checker=None,
|
| 66 |
requires_safety_checker=False,
|
| 67 |
-
)
|
| 68 |
|
| 69 |
self.pipe.scheduler = UniPCMultistepScheduler.from_config(self.pipe.scheduler.config)
|
| 70 |
|
| 71 |
self.auto_processor = AutoProcessor.from_pretrained(VIT_PATH)
|
| 72 |
-
self.image_encoder = CLIPVisionModelWithProjection.from_pretrained(VIT_PATH)
|
| 73 |
|
| 74 |
self.tokenizer = CLIPTokenizer.from_pretrained(
|
| 75 |
MODEL_PATH,
|
|
@@ -78,7 +78,7 @@ class OOTDiffusion:
|
|
| 78 |
self.text_encoder = CLIPTextModel.from_pretrained(
|
| 79 |
MODEL_PATH,
|
| 80 |
subfolder="text_encoder",
|
| 81 |
-
)
|
| 82 |
|
| 83 |
|
| 84 |
def tokenize_captions(self, captions, max_length):
|
|
@@ -107,14 +107,14 @@ class OOTDiffusion:
|
|
| 107 |
generator = torch.manual_seed(seed)
|
| 108 |
|
| 109 |
with torch.no_grad():
|
| 110 |
-
prompt_image = self.auto_processor(images=image_garm, return_tensors="pt").to(
|
| 111 |
prompt_image = self.image_encoder(prompt_image.data['pixel_values']).image_embeds
|
| 112 |
prompt_image = prompt_image.unsqueeze(1)
|
| 113 |
if model_type == 'hd':
|
| 114 |
-
prompt_embeds = self.text_encoder(self.tokenize_captions([""], 2).to(
|
| 115 |
prompt_embeds[:, 1:] = prompt_image[:]
|
| 116 |
elif model_type == 'dc':
|
| 117 |
-
prompt_embeds = self.text_encoder(self.tokenize_captions([category], 3).to(
|
| 118 |
prompt_embeds = torch.cat([prompt_embeds, prompt_image], dim=1)
|
| 119 |
else:
|
| 120 |
raise ValueError("model_type must be \'hd\' or \'dc\'!")
|
|
|
|
| 33 |
class OOTDiffusion:
|
| 34 |
|
| 35 |
def __init__(self, gpu_id):
|
| 36 |
+
# self.gpu_id = 'cuda:' + str(gpu_id)
|
| 37 |
|
| 38 |
vae = AutoencoderKL.from_pretrained(
|
| 39 |
VAE_PATH,
|
|
|
|
| 64 |
use_safetensors=True,
|
| 65 |
safety_checker=None,
|
| 66 |
requires_safety_checker=False,
|
| 67 |
+
)#.to(self.gpu_id)
|
| 68 |
|
| 69 |
self.pipe.scheduler = UniPCMultistepScheduler.from_config(self.pipe.scheduler.config)
|
| 70 |
|
| 71 |
self.auto_processor = AutoProcessor.from_pretrained(VIT_PATH)
|
| 72 |
+
self.image_encoder = CLIPVisionModelWithProjection.from_pretrained(VIT_PATH)#.to(self.gpu_id)
|
| 73 |
|
| 74 |
self.tokenizer = CLIPTokenizer.from_pretrained(
|
| 75 |
MODEL_PATH,
|
|
|
|
| 78 |
self.text_encoder = CLIPTextModel.from_pretrained(
|
| 79 |
MODEL_PATH,
|
| 80 |
subfolder="text_encoder",
|
| 81 |
+
)#.to(self.gpu_id)
|
| 82 |
|
| 83 |
|
| 84 |
def tokenize_captions(self, captions, max_length):
|
|
|
|
| 107 |
generator = torch.manual_seed(seed)
|
| 108 |
|
| 109 |
with torch.no_grad():
|
| 110 |
+
prompt_image = self.auto_processor(images=image_garm, return_tensors="pt").to('cuda')
|
| 111 |
prompt_image = self.image_encoder(prompt_image.data['pixel_values']).image_embeds
|
| 112 |
prompt_image = prompt_image.unsqueeze(1)
|
| 113 |
if model_type == 'hd':
|
| 114 |
+
prompt_embeds = self.text_encoder(self.tokenize_captions([""], 2).to('cuda'))[0]
|
| 115 |
prompt_embeds[:, 1:] = prompt_image[:]
|
| 116 |
elif model_type == 'dc':
|
| 117 |
+
prompt_embeds = self.text_encoder(self.tokenize_captions([category], 3).to('cuda'))[0]
|
| 118 |
prompt_embeds = torch.cat([prompt_embeds, prompt_image], dim=1)
|
| 119 |
else:
|
| 120 |
raise ValueError("model_type must be \'hd\' or \'dc\'!")
|