Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,12 +1,53 @@
|
|
| 1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
def run_parallel_models(prompt):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
return gr.update(), gr.update(), gr.update()
|
| 5 |
|
|
|
|
|
|
|
| 6 |
with gr.Blocks() as demo:
|
| 7 |
-
gr.Markdown("#Fast Flux Comparison")
|
| 8 |
with gr.Row():
|
| 9 |
-
prompt = gr.Textbox()
|
| 10 |
submit = gr.Button()
|
| 11 |
with gr.Row():
|
| 12 |
schnell = gr.Image(label="FLUX Schnell (4 steps)")
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
from diffusers import DiffusionPipeline
|
| 3 |
+
import spaces
|
| 4 |
+
|
| 5 |
+
dev_model = "black-forest-labs/FLUX.1-dev"
|
| 6 |
+
schnell_model = "black-forest-labs/FLUX.1-schnell"
|
| 7 |
+
|
| 8 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 9 |
+
|
| 10 |
+
pipe_dev = DiffusionPipeline.from_pretrained(dev_model, torch_dtype=torch.bfloat16).to(device)
|
| 11 |
+
pipe_schnell = DiffusionPipeline.from_pretrained(schnell_model, torch_dtype=torch.bfloat16).to(device)
|
| 12 |
+
|
| 13 |
+
@spaces.GPU
|
| 14 |
+
def run_dev_hyper(prompt):
|
| 15 |
+
repo_name = "ByteDance/Hyper-SD"
|
| 16 |
+
ckpt_name = "Hyper-FLUX.1-dev-8steps-lora.safetensors"
|
| 17 |
+
pipe_dev.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
|
| 18 |
+
image = pipe_dev(prompt, num_inference_steps=8, joint_attention_kwargs={"scale": 0.125}).images[0]
|
| 19 |
+
pipe_dev.unload_lora_weights()
|
| 20 |
+
return image
|
| 21 |
+
|
| 22 |
+
@spaces.GPU
|
| 23 |
+
def run_dev_turbo(prompt):
|
| 24 |
+
repo_name = "alimama-creative/FLUX.1-Turbo-Alpha"
|
| 25 |
+
ckpt_name = "diffusion_pytorch_model.safetensors"
|
| 26 |
+
pipe_dev.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
|
| 27 |
+
image = pipe_dev(prompt, num_inference_steps=8).images[0]
|
| 28 |
+
pipe_dev.unload_lora_weights()
|
| 29 |
+
return image
|
| 30 |
+
|
| 31 |
+
@spaces.GPU
|
| 32 |
+
def run_schnell(prompt):
|
| 33 |
+
image = pipe_schnell(prompt).images[0]
|
| 34 |
+
return image
|
| 35 |
|
| 36 |
def run_parallel_models(prompt):
|
| 37 |
+
|
| 38 |
+
with ProcessPoolExecutor(3) as e:
|
| 39 |
+
image_dev_hyper = run_dev_hyper(prompt)
|
| 40 |
+
image_dev_turbo = run_dev_turbo(prompt)
|
| 41 |
+
image_schnell = run_schnell(prompt)
|
| 42 |
+
|
| 43 |
return gr.update(), gr.update(), gr.update()
|
| 44 |
|
| 45 |
+
run_parallel_models.zerogpu = True
|
| 46 |
+
|
| 47 |
with gr.Blocks() as demo:
|
| 48 |
+
gr.Markdown("# Fast Flux Comparison")
|
| 49 |
with gr.Row():
|
| 50 |
+
prompt = gr.Textbox(label="Prompt")
|
| 51 |
submit = gr.Button()
|
| 52 |
with gr.Row():
|
| 53 |
schnell = gr.Image(label="FLUX Schnell (4 steps)")
|