zero gpu usage activate

app.py

@@ -2,133 +2,76 @@ import gradio as gr
 import numpy as np
 import random
 
-import spaces #[uncomment to use ZeroGPU]
 from diffusers import DiffusionPipeline
 import torch
+import os
 
 from peft import PeftModel, LoraConfig
-import os
 
-def get_lora_sd_pipeline(
-    ckpt_dir='./lora', 
-    base_model_name_or_path=None, 
-    dtype=torch.float16, 
-    adapter_name="default"
-    ):
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
+def get_lora_sd_pipeline(
+    ckpt_dir="./lora",
+    base_model_name_or_path="sd-legacy/stable-diffusion-v1-5",
+    dtype=torch.float16,
+    adapter_name="default",
+):
     unet_sub_dir = os.path.join(ckpt_dir, "unet")
     text_encoder_sub_dir = os.path.join(ckpt_dir, "text_encoder")
-    
-    if os.path.exists(text_encoder_sub_dir) and base_model_name_or_path is None:
-        config = LoraConfig.from_pretrained(text_encoder_sub_dir)
-        base_model_name_or_path = config.base_model_name_or_path
-    
-    if base_model_name_or_path is None:
-        raise ValueError("Please specify the base model name or path")
-    
-    pipe = DiffusionPipeline.from_pretrained(base_model_name_or_path, torch_dtype=dtype)
-    before_params = pipe.unet.parameters()
-    pipe.unet = PeftModel.from_pretrained(pipe.unet, unet_sub_dir, adapter_name=adapter_name)
-    pipe.unet.set_adapter(adapter_name)
-    after_params = pipe.unet.parameters()
-    print("Parameters changed:", any(torch.any(b != a) for b, a in zip(before_params, after_params)))
-    
-    if os.path.exists(text_encoder_sub_dir):
-        pipe.text_encoder = PeftModel.from_pretrained(pipe.text_encoder, text_encoder_sub_dir, adapter_name=adapter_name)
-    
-    if dtype in (torch.float16, torch.bfloat16):
-        pipe.unet.half()
-        pipe.text_encoder.half()
-    
-    return pipe
-
-def process_prompt(prompt, tokenizer, text_encoder, max_length=77):
-    tokens = tokenizer(prompt, truncation=False, return_tensors="pt")["input_ids"]
-    chunks = [tokens[:, i:i + max_length] for i in range(0, tokens.shape[1], max_length)]
-    
-    with torch.no_grad():
-        embeds = [text_encoder(chunk.to(text_encoder.device))[0] for chunk in chunks]
-    
-    return torch.cat(embeds, dim=1)
-
-def align_embeddings(prompt_embeds, negative_prompt_embeds):
-    max_length = max(prompt_embeds.shape[1], negative_prompt_embeds.shape[1])
-    return torch.nn.functional.pad(prompt_embeds, (0, 0, 0, max_length - prompt_embeds.shape[1])), \
-           torch.nn.functional.pad(negative_prompt_embeds, (0, 0, 0, max_length - negative_prompt_embeds.shape[1]))
-
-device = "cuda" if torch.cuda.is_available() else "cpu"
 
-model_id_default = "sd-legacy/stable-diffusion-v1-5"
-model_dropdown = ['stabilityai/sdxl-turbo', 'CompVis/stable-diffusion-v1-4', 'sd-legacy/stable-diffusion-v1-5']
+    pipe = DiffusionPipeline.from_pretrained(base_model_name_or_path, torch_dtype=dtype).to(device)
 
-model_lora_default = "lora"
+    if os.path.exists(unet_sub_dir):
+        pipe.unet = PeftModel.from_pretrained(pipe.unet, unet_sub_dir, adapter_name=adapter_name)
+        print(f"LoRA adapter loaded: {adapter_name}")
 
-if torch.cuda.is_available():
-    torch_dtype = torch.float16
-else:
-    torch_dtype = torch.float32
+    if os.path.exists(text_encoder_sub_dir):
+        pipe.text_encoder = PeftModel.from_pretrained(pipe.text_encoder, text_encoder_sub_dir, adapter_name=adapter_name)
 
-MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
+    return pipe
 
 
-@spaces.GPU #[uncomment to use ZeroGPU]
 def infer(
     prompt,
-    negative_prompt,
-    randomize_seed,
+    negative_prompt="",
+    randomize_seed=False,
     width=512,
     height=512,
-    model_repo_id=model_id_default,
+    model_repo_id="sd-legacy/stable-diffusion-v1-5",
     seed=42,
     guidance_scale=7,
     num_inference_steps=20,
-    model_lora_id=model_lora_default,
+    model_lora_id="lora",
     lora_scale=0.5,
-    progress=gr.Progress(track_tqdm=True),
-    ):
-        
+):
     if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-
-    generator = torch.Generator().manual_seed(seed)
-    
-    # добавляем обновление pipe по условию
-    if model_repo_id != model_id_default:
-        pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype).to(device)
-        prompt_embeds = process_prompt(prompt, pipe.tokenizer, pipe.text_encoder)
-        negative_prompt_embeds = process_prompt(negative_prompt, pipe.tokenizer, pipe.text_encoder)
-        prompt_embeds, negative_prompt_embeds = align_embeddings(prompt_embeds, negative_prompt_embeds)
+        seed = random.randint(0, np.iinfo(np.int32).max)
+
+    generator = torch.manual_seed(seed)
+
+    # Загружаем основную модель или с LoRA
+    if model_lora_id != "none":
+        pipe = get_lora_sd_pipeline(ckpt_dir=f"./{model_lora_id}", base_model_name_or_path=model_repo_id)
     else:
-        # добавляем lora
-        pipe = get_lora_sd_pipeline(ckpt_dir='./' + model_lora_id, base_model_name_or_path=model_id_default, dtype=torch_dtype).to(device)
-        prompt_embeds = process_prompt(prompt, pipe.tokenizer, pipe.text_encoder)
-        negative_prompt_embeds = process_prompt(negative_prompt, pipe.tokenizer, pipe.text_encoder)
-        prompt_embeds, negative_prompt_embeds = align_embeddings(prompt_embeds, negative_prompt_embeds)
-        print(f"LoRA adapter loaded: {pipe.unet.active_adapters}")
-        print(f"LoRA scale applied: {lora_scale}")
+        pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch.float16).to(device)
+
+    # Применяем LoRA, если он есть
+    if hasattr(pipe.unet, "fuse_lora"):
         pipe.fuse_lora(lora_scale=lora_scale)
 
-    # на вызов pipe с эмбеддингами
-    params = {
-        'prompt_embeds': prompt_embeds,
-        'negative_prompt_embeds': negative_prompt_embeds,
-        'guidance_scale': guidance_scale,
-        'num_inference_steps': num_inference_steps,
-        'width': width,
-        'height': height,
-        'generator': generator,
-    }
-    
-    return pipe(**params).images[0], seed    
-
-
-examples = [
-    "A Elon Mask lady in a Russian embroidered kaftan is sitting on a beautiful carved veranda, holding a cup to her mouth and drinking tea from the cup. With her other hand, the girl holds a saucer. The cup and saucer are painted with gzhel. Next to the girl on the table stands a samovar, and steam can be seen above it.",
-    "Elon Mask in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An Elon Mask astronaut riding a green horse",
-    "A delicious Elon Mask ceviche cheesecake slice",
-]
+    # Генерируем изображение
+    image = pipe(
+        prompt=prompt,
+        negative_prompt=negative_prompt,
+        width=width,
+        height=height,
+        guidance_scale=guidance_scale,
+        num_inference_steps=num_inference_steps,
+        generator=generator,
+    ).images[0]
+
+    return image, seed
+
 
 css = """
 #col-container {
@@ -139,20 +82,25 @@ css = """
 
 with gr.Blocks(css=css) as demo:
     with gr.Column(elem_id="col-container"):
-        gr.Markdown(" # Text-to-Image")
-
-        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, variant="primary")
-
-        result = gr.Image(label="Result", show_label=False)
+        gr.Markdown(" # Text-to-Image with LoRA Support")
+
+        prompt = gr.Text(label="Prompt", placeholder="Enter your prompt")
+        negative_prompt = gr.Text(label="Negative Prompt", placeholder="Optional")
+        width = gr.Slider(256, 1024, value=512, step=64, label="Width")
+        height = gr.Slider(256, 1024, value=512, step=64, label="Height")
+        num_steps = gr.Slider(1, 50, value=20, step=1, label="Steps")
+        guidance = gr.Slider(1, 15, value=7, step=0.1, label="Guidance Scale")
+        lora_scale = gr.Slider(0, 1, value=0.5, step=0.1, label="LoRA Strength")
+        randomize_seed = gr.Checkbox(label="Randomize Seed")
+        result = gr.Image(label="Generated Image")
+
+        run_button = gr.Button("Generate")
+
+        run_button.click(
+            fn=infer,
+            inputs=[prompt, negative_prompt, randomize_seed, width, height, num_steps, guidance, lora_scale],
+            outputs=[result],
+        )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()