fixing the path of concept model's bin file

app.py

@@ -3,6 +3,8 @@ import torch
 from torch import autocast
 from diffusers import StableDiffusionPipeline
 import random
+from huggingface_hub import hf_hub_download
+import os
 
 # Initialize the model
 model_id = "CompVis/stable-diffusion-v1-4"
@@ -17,6 +19,19 @@ concepts = [
     "sd-concepts-library/wu-shi"
 ]
 
+def download_concept_embedding(concept_name):
+    try:
+        # Download the learned_embeds.bin file from the Hub
+        embed_path = hf_hub_download(
+            repo_id=concept_name,
+            filename="learned_embeds.bin",
+            repo_type="model"
+        )
+        return embed_path
+    except Exception as e:
+        print(f"Error downloading {concept_name}: {str(e)}")
+        return None
+
 def load_learned_embed_in_clip(learned_embeds_path, text_encoder, tokenizer):
     loaded_learned_embeds = torch.load(learned_embeds_path, map_location="cpu")
     
@@ -43,36 +58,45 @@ def generate_images(prompt):
     ).to(device)
     
     for concept in concepts:
-        # Load concept embedding
-        token = load_learned_embed_in_clip(
-            f"{concept}/blob/main/learned_embeds.bin",
-            pipe.text_encoder,
-            pipe.tokenizer
-        )
-       
-        # Generate random seed
-        seed = random.randint(1, 999999)
-        generator = torch.Generator(device=device).manual_seed(seed)
-        
-        # Add concept token to prompt
-        concept_prompt = f"{token} {prompt}"
-        
-        # Generate image
-        with autocast(device):
-            image = pipe(
-                concept_prompt,
-                num_inference_steps=50,
-                generator=generator,
-                guidance_scale=7.5
-            ).images[0]
-        
-        images.append(image)
-        
-        # Clear concept from pipeline
-        pipe.tokenizer.remove_tokens([token])
-        pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer))
+        # Download and load concept embedding
+        embed_path = download_concept_embedding(concept)
+        if embed_path is None:
+            continue
+            
+        try:
+            token = load_learned_embed_in_clip(
+                embed_path,
+                pipe.text_encoder,
+                pipe.tokenizer
+            )
+            
+            # Generate random seed
+            seed = random.randint(1, 999999)
+            generator = torch.Generator(device=device).manual_seed(seed)
+            
+            # Add concept token to prompt
+            concept_prompt = f"{token} {prompt}"
+            
+            # Generate image
+            with autocast(device):
+                image = pipe(
+                    concept_prompt,
+                    num_inference_steps=50,
+                    generator=generator,
+                    guidance_scale=7.5
+                ).images[0]
+            
+            images.append(image)
+            
+            # Clear concept from pipeline
+            pipe.tokenizer.remove_tokens([token])
+            pipe.text_encoder.resize_token_embeddings(len(pipe.tokenizer))
+            
+        except Exception as e:
+            print(f"Error processing concept {concept}: {str(e)}")
+            continue
     
-    return images
+    return images if images else [None] * 5
 
 # Create Gradio interface
 iface = gr.Interface(

requirements.txt

@@ -1,4 +1,5 @@
 torch
 diffusers
 transformers
-gradio 
+gradio
+huggingface_hub