Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -8,6 +8,14 @@ import spaces
|
|
| 8 |
|
| 9 |
HF_TOKEN = os.getenv("HF_TOKEN")
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
model_path = snapshot_download(
|
| 12 |
repo_id="stabilityai/stable-diffusion-3-medium",
|
| 13 |
revision="refs/pr/26",
|
|
@@ -17,24 +25,19 @@ model_path = snapshot_download(
|
|
| 17 |
token=HF_TOKEN,
|
| 18 |
)
|
| 19 |
|
| 20 |
-
if torch.cuda.is_available():
|
| 21 |
-
device = "cuda"
|
| 22 |
-
print("Using GPU")
|
| 23 |
-
else:
|
| 24 |
-
device = "cpu"
|
| 25 |
-
print("Using CPU")
|
| 26 |
|
| 27 |
# Initialize the pipeline and download the model
|
| 28 |
pipe = StableDiffusion3Pipeline.from_pretrained(model_path, torch_dtype=torch.float16)
|
| 29 |
pipe.to(device)
|
| 30 |
|
|
|
|
| 31 |
tokenizer = T5Tokenizer.from_pretrained("roborovski/superprompt-v1")
|
| 32 |
model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", device_map="auto", torch_dtype="auto")
|
| 33 |
model.to(device)
|
| 34 |
|
| 35 |
# Define the image generation function
|
| 36 |
@spaces.GPU(duration=60)
|
| 37 |
-
def generate_image(prompt, negative_prompt, num_inference_steps, height, width, guidance_scale, seed, num_images_per_prompt):
|
| 38 |
if seed == 0:
|
| 39 |
seed = random.randint(1, 2**32-1)
|
| 40 |
|
|
|
|
| 8 |
|
| 9 |
HF_TOKEN = os.getenv("HF_TOKEN")
|
| 10 |
|
| 11 |
+
if torch.cuda.is_available():
|
| 12 |
+
device = "cuda"
|
| 13 |
+
print("Using GPU")
|
| 14 |
+
else:
|
| 15 |
+
device = "cpu"
|
| 16 |
+
print("Using CPU")
|
| 17 |
+
|
| 18 |
+
# download sd3 medium weights
|
| 19 |
model_path = snapshot_download(
|
| 20 |
repo_id="stabilityai/stable-diffusion-3-medium",
|
| 21 |
revision="refs/pr/26",
|
|
|
|
| 25 |
token=HF_TOKEN,
|
| 26 |
)
|
| 27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
# Initialize the pipeline and download the model
|
| 30 |
pipe = StableDiffusion3Pipeline.from_pretrained(model_path, torch_dtype=torch.float16)
|
| 31 |
pipe.to(device)
|
| 32 |
|
| 33 |
+
# superprompt-v1
|
| 34 |
tokenizer = T5Tokenizer.from_pretrained("roborovski/superprompt-v1")
|
| 35 |
model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", device_map="auto", torch_dtype="auto")
|
| 36 |
model.to(device)
|
| 37 |
|
| 38 |
# Define the image generation function
|
| 39 |
@spaces.GPU(duration=60)
|
| 40 |
+
def generate_image(prompt, enhance_prompt, negative_prompt, num_inference_steps, height, width, guidance_scale, seed, num_images_per_prompt):
|
| 41 |
if seed == 0:
|
| 42 |
seed = random.randint(1, 2**32-1)
|
| 43 |
|