let's work this out

app.py

@@ -1,115 +1,60 @@
 import gradio as gr
-import numpy as np
-import random
-import spaces
 import torch
-from diffusers import DiffusionPipeline
-
-# Enable TF32 for A100 (this is a form of FP8 computation)
-torch.backends.cuda.matmul.allow_tf32 = True
-torch.backends.cudnn.allow_tf32 = True
+import spaces
+from diffusers import FluxPipeline
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
-dtype = torch.float16  # Use float16 for loading
+dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+
+MODEL_ID = "drbaph/FLUX.1-schnell-dev-merged-fp8-4step"
+MODEL_FILE = "flux1-schnell-dev-merged-fp8-4step.safetensors"
 
-pipe = DiffusionPipeline.from_pretrained(
-    "Kijai/flux-fp8",
-    torch_dtype=dtype,
-    revision="main",
-    filename="flux1-schnell-fp8.safetensors"
-).to(device)
+def load_model():
+    pipe = FluxPipeline.from_single_file(
+        f"https://huggingface.co/{MODEL_ID}/resolve/main/{MODEL_FILE}",
+        torch_dtype=dtype
+    )
+    pipe.to(device)
+    return pipe
 
-MAX_SEED = np.iinfo(np.int32).max
+pipe = load_model()
+
+MAX_SEED = 2**32 - 1
 MAX_IMAGE_SIZE = 2048
 
 @spaces.GPU()
 def infer(prompt, seed=42, randomize_seed=False, width=1024, height=1024, num_inference_steps=4, progress=gr.Progress(track_tqdm=True)):
     if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-    generator = torch.Generator().manual_seed(seed)
+        seed = torch.randint(0, MAX_SEED, (1,)).item()
+    generator = torch.Generator(device=device).manual_seed(seed)
     image = pipe(
         prompt=prompt,
         width=width,
         height=height,
         num_inference_steps=num_inference_steps,
         generator=generator,
-        guidance_scale=0.0
+        guidance_scale=0.0,
+        max_sequence_length=256
     ).images[0]
     return image, seed
 
-examples = [
-    "a tiny astronaut hatching from an egg on the moon",
-    "a cat holding a sign that says hello world",
-    "an anime illustration of a wiener schnitzel",
-]
-
-css="""
-#col-container {
-    margin: 0 auto;
-    max-width: 520px;
-}
-"""
-
-with gr.Blocks(css=css) as demo:
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown(f"""# FLUX.1 [schnell] FP8
-        12B param rectified flow transformer distilled from [FLUX.1 [pro]](https://blackforestlabs.ai/) for 4 step generation
-        [[blog](https://blackforestlabs.ai/announcing-black-forest-labs/)] [[model](https://huggingface.co/Kijai/flux-fp8)]
-        """)
-        with gr.Row():
-            prompt = gr.Text(
-                label="Prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter your prompt",
-                container=False,
-            )
-            run_button = gr.Button("Run", scale=0)
-        result = gr.Image(label="Result", show_label=False)
-        with gr.Accordion("Advanced Settings", open=False):
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
-            )
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-            with gr.Row():
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024,
-                )
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024,
-                )
-            with gr.Row():
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=50,
-                    step=1,
-                    value=4,
-                )
-        gr.Examples(
-            examples=examples,
-            fn=infer,
-            inputs=[prompt],
-            outputs=[result, seed],
-            cache_examples="lazy"
-        )
-        gr.on(
-            triggers=[run_button.click, prompt.submit],
-            fn=infer,
-            inputs=[prompt, seed, randomize_seed, width, height, num_inference_steps],
-            outputs=[result, seed]
-        )
+# Gradio interface
+with gr.Blocks() as demo:
+    gr.Markdown("# FLUX.1 [schnell-dev-merged-fp8-4step]")
+    with gr.Row():
+        prompt = gr.Textbox(label="Prompt")
+        run_button = gr.Button("Generate")
+    with gr.Row():
+        result = gr.Image(label="Generated Image")
+        seed_output = gr.Number(label="Seed Used")
+    with gr.Accordion("Advanced Settings", open=False):
+        seed = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=42)
+        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+        width = gr.Slider(label="Width", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
+        height = gr.Slider(label="Height", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
+        num_inference_steps = gr.Slider(label="Number of inference steps", minimum=1, maximum=10, step=1, value=4)
+    
+    inputs = [prompt, seed, randomize_seed, width, height, num_inference_steps]
+    run_button.click(fn=infer, inputs=inputs, outputs=[result, seed_output])
 
 demo.launch()