Update app.py

app.py

@@ -1,6 +1,8 @@
 import gradio as gr
-from huggingface_hub import login
+from huggingface_hub import login, HfFileSystem, HfApi, ModelCard
 import os
+import spaces
+import random
 import torch
 
 is_shared_ui = True if "fffiloni/sdxl-control-loras" in os.environ['SPACE_ID'] else False
@@ -8,6 +10,9 @@ is_shared_ui = True if "fffiloni/sdxl-control-loras" in os.environ['SPACE_ID'] e
 hf_token = os.environ.get("HF_TOKEN")
 login(token=hf_token)
 
+fs = HfFileSystem(token=hf_token)
+api = HfApi()
+
 device="cuda" if torch.cuda.is_available() else "cpu"
 
 from diffusers import ControlNetModel, StableDiffusionXLControlNetPipeline, AutoencoderKL
@@ -24,22 +29,57 @@ controlnet = ControlNetModel.from_pretrained(
     torch_dtype=torch.float16
 )
 
-pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
-    "stabilityai/stable-diffusion-xl-base-1.0",
-    controlnet=controlnet,
-    vae=vae,
-    torch_dtype=torch.float16, 
-    variant="fp16",
-    use_safetensors=True
-)
+def get_files(file_paths):
+    last_files = {}  # Dictionary to store the last file for each path
+
+    for file_path in file_paths:
+        # Split the file path into directory and file components
+        directory, file_name = file_path.rsplit('/', 1)
+    
+        # Update the last file for the current path
+        last_files[directory] = file_name
+    
+    # Extract the last files from the dictionary
+    result = list(last_files.values())
 
-pipe.to(device)
+    return result
 
+def load_model(model_name):
 
+    if model_name == "":
+        gr.Warning("If you want to use a private model, you need to duplicate this space on your personal account.")
+        raise gr.Error("You forgot to define Model ID.")
 
-#pipe.enable_model_cpu_offload()
+    # Get instance_prompt a.k.a trigger word
+    card = ModelCard.load(model_name)
+    repo_data = card.data.to_dict()
+    instance_prompt = repo_data.get("instance_prompt")
 
-from PIL import Image
+    if instance_prompt is not None:
+        print(f"Trigger word: {instance_prompt}")
+    else:
+        instance_prompt = "no trigger word needed"
+        print(f"Trigger word: no trigger word needed")
+
+    # List all ".safetensors" files in repo
+    sfts_available_files = fs.glob(f"{model_name}/*safetensors")
+    sfts_available_files = get_files(sfts_available_files)
+
+    if sfts_available_files == []:
+        sfts_available_files = ["NO SAFETENSORS FILE"]
+
+    print(f"Safetensors available: {sfts_available_files}")
+
+    return model_name, "Model Ready", gr.update(choices=sfts_available_files, value=sfts_available_files[0], visible=True), gr.update(value=instance_prompt, visible=True)
+
+def custom_model_changed(model_name, previous_model):
+    if model_name == "" and previous_model == "" :
+        status_message = ""      
+    elif model_name != previous_model:
+        status_message = "model changed, please reload before any new run"
+    else:
+        status_message = "model ready"
+    return status_message
 
 def resize_image(input_path, output_path, target_height):
     # Open the input image
@@ -61,8 +101,24 @@ def resize_image(input_path, output_path, target_height):
     return output_path
 
 def infer(use_custom_model, model_name, weight_name, custom_lora_weight, image_in, prompt, negative_prompt, preprocessor, controlnet_conditioning_scale, guidance_scale, inf_steps, seed, progress=gr.Progress(track_tqdm=True)):
+
+    pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
+        "stabilityai/stable-diffusion-xl-base-1.0",
+        controlnet=controlnet,
+        vae=vae,
+        torch_dtype=torch.float16, 
+        variant="fp16",
+        use_safetensors=True
+    )
+    
+    pipe.to(device)
+    
     prompt = prompt
     negative_prompt = negative_prompt
+
+    if seed < 0 :
+        seed = random.randint(0, 423538377342)
+    
     generator = torch.Generator(device=device).manual_seed(seed)
 
     if image_in == None:
@@ -88,7 +144,20 @@ def infer(use_custom_model, model_name, weight_name, custom_lora_weight, image_i
         custom_model = model_name
 
         # This is where you load your trained weights
-        pipe.load_lora_weights(custom_model, weight_name=weight_name, use_auth_token=True)
+        if weight_name == "NO SAFETENSORS FILE": 
+            pipe.load_lora_weights(
+                custom_model,     
+                low_cpu_mem_usage = True,
+                use_auth_token = True
+            )
+    
+        else:
+            pipe.load_lora_weights(
+                custom_model,
+                weight_name = weight_name,        
+                low_cpu_mem_usage = True,
+                use_auth_token = True
+            )
     
         lora_scale=custom_lora_weight
 
@@ -115,7 +184,7 @@ def infer(use_custom_model, model_name, weight_name, custom_lora_weight, image_i
 
     images[0].save(f"result.png")
 
-    return f"result.png"
+    return f"result.png", seed
 
 css="""
 #col-container{
@@ -143,6 +212,12 @@ div#warning-duplicate .actions a {
     display: inline-block;
     margin-right: 10px;
 }
+button#load_model_btn{
+    height: 46px;
+}
+#status_info{
+    font-size: 0.9em;
+}
 """
 
 with gr.Blocks(css=css) as demo:
@@ -170,28 +245,100 @@ with gr.Blocks(css=css) as demo:
         """)
         
         image_in = gr.Image(source="upload", type="filepath")
+        
         with gr.Row():
+            
             with gr.Column():
                 prompt = gr.Textbox(label="Prompt")
                 negative_prompt = gr.Textbox(label="Negative prompt", value="extra digit, fewer digits, cropped, worst quality, low quality, glitch, deformed, mutated, ugly, disfigured")
                 guidance_scale = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=10.0, step=0.1, value=7.5)
                 inf_steps = gr.Slider(label="Inference Steps", minimum="25", maximum="50", step=1, value=25)
+            
             with gr.Column():
-                preprocessor = gr.Dropdown(label="Preprocessor", choices=["canny"], value="canny", interactive=False, info="For the moment, only canny is available")  
-                controlnet_conditioning_scale = gr.Slider(label="Controlnet conditioning Scale", minimum=0.1, maximum=0.9, step=0.01, value=0.5)
-                seed = gr.Slider(label="seed", minimum=0, maximum=500000, step=1, value=42)
+                with gr.Group():
+                    preprocessor = gr.Dropdown(label="Preprocessor", choices=["canny"], value="canny", interactive=False, info="For the moment, only canny is available")  
+                    controlnet_conditioning_scale = gr.Slider(label="Controlnet conditioning Scale", minimum=0.1, maximum=0.9, step=0.01, value=0.5)
+                with gr.Group():
+                    seed = gr.Slider(
+                        label="Seed",
+                        info = "-1 denotes a random seed",
+                        minimum=-1,
+                        maximum=423538377342,
+                        step=1,
+                        value=-1
+                    )
+                    last_used_seed = gr.Number(
+                        label = "Last used seed",
+                        info = "the seed used in the last generation",
+                    )
+        
         use_custom_model = gr.Checkbox(label="Use a public custom model ?(optional)", value=False, info="To use a private model, you'll prefer to duplicate the space with your own access token.")
-        with gr.Row():
-            model_name = gr.Textbox(label="Custom Model to use", placeholder="username/my_custom_public_model")
-            weight_name = gr.Textbox(label="Specific safetensor", value="pytorch_lora_weights.safetensors",info="specify which one if model has several .safetensors files" )
-        custom_lora_weight = gr.Slider(label="Custom model weights", minimum=0.1, maximum=0.9, step=0.1, value=0.9)
+        
+        with gr.Box():
+            with gr.Row():
+                with gr.Column():
+                    if not is_shared_ui:
+                        your_username = api.whoami()["name"]
+                        my_models = api.list_models(author=your_username, filter=["diffusers", "stable-diffusion-xl", 'lora'])
+                        model_names = [item.modelId for item in my_models]
+    
+                    if not is_shared_ui:
+                        custom_model = gr.Dropdown(
+                            label = "Your custom model ID",
+                            info="You can pick one of your private models",
+                            choices = model_names,
+                            allow_custom_value = True
+                            #placeholder = "username/model_id"
+                        )
+                    else:
+                        custom_model = gr.Textbox(
+                            label="Your custom model ID", 
+                            placeholder="your_username/your_trained_model_name", 
+                            info="Make sure your model is set to PUBLIC"
+                        )
+                    
+                    weight_name = gr.Dropdown(
+                        label="Safetensors file", 
+                        #value="pytorch_lora_weights.safetensors", 
+                        info="specify which one if model has several .safetensors files",
+                        allow_custom_value=True,
+                        visible = False
+                    )
+                with gr.Column():
+                    with gr.Group():
+                        load_model_btn = gr.Button("Load my model", elem_id="load_model_btn")
+                        previous_model = gr.Textbox(
+                            visible = False
+                        )
+                        model_status = gr.Textbox(
+                            label = "model status",
+                            show_label = False,
+                            elem_id = "status_info"
+                        )
+                    trigger_word = gr.Textbox(label="Trigger word", interactive=False, visible=False)
+        
+            custom_lora_weight = gr.Slider(label="Custom model weights", minimum=0.1, maximum=0.9, step=0.1, value=0.9)
+        
         submit_btn = gr.Button("Submit")
+        
         result = gr.Image(label="Result")
 
+    custom_model.blur(
+        fn=custom_model_changed,
+        inputs = [custom_model, previous_model],
+        outputs = [model_status],
+        queue = False
+    )
+    load_model_btn.click(
+        fn = load_model,
+        inputs=[custom_model],
+        outputs = [previous_model, model_status, weight_name, trigger_word],
+        queue = False
+    )
     submit_btn.click(
         fn = infer,
-        inputs = [use_custom_model, model_name, weight_name, custom_lora_weight, image_in, prompt, negative_prompt, preprocessor, controlnet_conditioning_scale, guidance_scale, inf_steps, seed],
-        outputs = [result]
+        inputs = [use_custom_model, custom_model, weight_name, custom_lora_weight, image_in, prompt, negative_prompt, preprocessor, controlnet_conditioning_scale, guidance_scale, inf_steps, seed],
+        outputs = [result, last_used_seed]
     )
 
 demo.queue(max_size=12).launch()