Update app.py

app.py

@@ -6,6 +6,176 @@ import torch
 from diffusers import DiffusionPipeline, FluxPipeline, AutoencoderTiny, AutoencoderKL
 from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast
 from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images
+from download import download_all_models, models, download_vaes
+
+# Call the download function at the start of the app
+download_all_models()
+download_vaes()
+
+dtype = torch.bfloat16
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Load VAEs - these can be reused across models
+taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
+good_vae = AutoencoderKL.from_pretrained("black-forest-labs/FLUX.1-dev", subfolder="vae", torch_dtype=dtype).to(device)
+
+def load_model(model_name):
+    model_info = models[model_name]
+    pipeline_class = model_info["pipeline_class"]
+    model_id = model_info["model_id"]
+    config = model_info["config"]
+
+    pipeline = pipeline_class.from_pretrained(model_id, **config, vae=taef1).to(device)
+    
+    # Assign the custom function for live preview if it's a FluxPipeline or DiffusionPipeline
+    if pipeline_class in (FluxPipeline, DiffusionPipeline):
+        pipeline.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipeline)
+
+    return pipeline, model_info.get("description", "No description available.")
+
+# Initialize with default model
+current_model_name = "FLUX.1-dev"
+pipe, model_description = load_model(current_model_name)
+torch.cuda.empty_cache()
+
+MAX_SEED = np.iinfo(np.int32).max
+MAX_IMAGE_SIZE = 2048
+
+@spaces.GPU(duration=75)
+def infer(model_name, prompt, seed=42, randomize_seed=False, width=1024, height=1024, guidance_scale=3.5, num_inference_steps=28, progress=gr.Progress(track_tqdm=True)):
+    global pipe, current_model_name  # Access the global pipe and model name
+
+    if randomize_seed:
+        seed = random.randint(0, MAX_SEED)
+    generator = torch.Generator().manual_seed(seed)
+
+    # Only reload the model if a different one is selected
+    if model_name != current_model_name:
+        pipe, _ = load_model(model_name)
+        current_model_name = model_name
+
+    for img in pipe.flux_pipe_call_that_returns_an_iterable_of_images(
+            prompt=prompt,
+            guidance_scale=guidance_scale,
+            num_inference_steps=num_inference_steps,
+            width=width,
+            height=height,
+            generator=generator,
+            output_type="pil",
+            good_vae=good_vae,
+        ):
+        yield img, seed
+
+examples = [
+    ["FLUX.1-dev", "a tiny astronaut hatching from an egg on the moon"],
+    ["FLUX.1-schnell", "a cat holding a sign that says hello world"],
+    ["Flux.1-lite-8B-alpha", "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 Model Selector
+Select a model to generate images using the FLUX pipeline.
+        """)
+        model_selector = gr.Dropdown(
+            choices=list(models.keys()),
+            value=current_model_name,
+            label="Select Model",
+        )
+        model_description_box = gr.Markdown(model_description)
+        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():
+                guidance_scale = gr.Slider(
+                    label="Guidance Scale",
+                    minimum=1,
+                    maximum=15,
+                    step=0.1,
+                    value=3.5,
+                )
+                num_inference_steps = gr.Slider(
+                    label="Number of inference steps",
+                    minimum=1,
+                    maximum=50,
+                    step=1,
+                    value=28,
+                )
+        
+    def update_description(selected_model):
+        return models[selected_model]["description"]
+
+    model_selector.change(
+        fn=update_description,
+        inputs=[model_selector],
+        outputs=[model_description_box],
+    )
+
+    gr.Examples(
+        examples=examples,
+        fn=infer,
+        inputs=[model_selector, prompt],  # Correct order of inputs
+        outputs=[result, seed],
+        cache_examples=False,
+    )
+
+    gr.on(
+        triggers=[run_button.click, prompt.submit],
+        fn=infer,
+        inputs=[model_selector, prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
+        outputs=[result, seed],
+    )
+
+demo.launch()
+
+#working1 
+''''
+import gradio as gr
+import numpy as np
+import random
+import spaces
+import torch
+from diffusers import DiffusionPipeline, FluxPipeline, AutoencoderTiny, AutoencoderKL
+from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast
+from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images
 
 # Define models and their configurations
 models = {
@@ -183,8 +353,8 @@ Select a model to generate images using the FLUX pipeline.
 
 demo.launch()
 
-
-
+'''
+#orginal
 
 '''
 import gradio as gr