Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -30,16 +30,16 @@ import diffusers
|
|
30 |
|
31 |
# init
|
32 |
dtype = torch.bfloat16
|
33 |
-
device =
|
|
|
34 |
print(device)
|
35 |
base_model = "black-forest-labs/FLUX.1-dev"
|
36 |
|
37 |
# load pipe
|
38 |
-
taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
|
39 |
-
good_vae = AutoencoderKL.from_pretrained(base_model, subfolder="vae", torch_dtype=dtype).to(device)
|
40 |
|
41 |
-
txt2img_pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=dtype)
|
42 |
txt2img_pipe.__class__.load_lora_into_transformer = classmethod(load_lora_into_transformer)
|
|
|
43 |
|
44 |
|
45 |
MAX_SEED = 2**32 - 1
|
@@ -108,6 +108,7 @@ def generate_random_4_digit_string():
|
|
108 |
return ''.join(random.choices(string.digits, k=4))
|
109 |
|
110 |
@spaces.GPU(duration=120)
|
|
|
111 |
def run_lora(prompt, image_url, lora_strings_json, image_strength, cfg_scale, steps, randomize_seed, seed, width, height, upload_to_r2, account_id, access_key, secret_key, bucket, progress=gr.Progress(track_tqdm=True)):
|
112 |
print("run_lora", prompt, lora_strings_json, cfg_scale, steps, width, height)
|
113 |
gr.Info("Starting process")
|
|
|
30 |
|
31 |
# init
|
32 |
dtype = torch.bfloat16
|
33 |
+
device = "cuda"
|
34 |
+
|
35 |
print(device)
|
36 |
base_model = "black-forest-labs/FLUX.1-dev"
|
37 |
|
38 |
# load pipe
|
|
|
|
|
39 |
|
40 |
+
txt2img_pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=dtype)
|
41 |
txt2img_pipe.__class__.load_lora_into_transformer = classmethod(load_lora_into_transformer)
|
42 |
+
txt2img_pipe = txt2img_pipe.to(device)
|
43 |
|
44 |
|
45 |
MAX_SEED = 2**32 - 1
|
|
|
108 |
return ''.join(random.choices(string.digits, k=4))
|
109 |
|
110 |
@spaces.GPU(duration=120)
|
111 |
+
|
112 |
def run_lora(prompt, image_url, lora_strings_json, image_strength, cfg_scale, steps, randomize_seed, seed, width, height, upload_to_r2, account_id, access_key, secret_key, bucket, progress=gr.Progress(track_tqdm=True)):
|
113 |
print("run_lora", prompt, lora_strings_json, cfg_scale, steps, width, height)
|
114 |
gr.Info("Starting process")
|