Fix: Requirements

requirements.txt

@@ -9,7 +9,11 @@ gradio>=3.20.0
 numpy>=1.22.0
 Pillow>=9.0.0
 tqdm>=4.64.0
-huggingface-hub>=0.12.0
+huggingface-hub>=0.12.0,<0.20.0
+
+# HF Spaces specific
+gradio-client>=0.2.5
+spaces>=0.19.4
 
 # Optional dependencies for better performance
 scipy>=1.9.0

utils.py

@@ -7,10 +7,13 @@ Date: Feb 26, 2025
 
 import torch
 import gc
+import os
 from PIL import Image, ImageDraw, ImageFont
 from diffusers import StableDiffusionPipeline
 from transformers import CLIPTokenizer, CLIPTextModel
-import os
+
+# Disable HF transfer to avoid download issues
+os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "0"
 
 def load_models(device="cuda"):
     """
@@ -33,14 +36,14 @@ def load_models(device="cuda"):
     print(f"Loading models on {device}...")
     
     # Load the autoencoder model which will be used to decode the latents into image space
-    vae = AutoencoderKL.from_pretrained("CompVis/stable-diffusion-v1-4", subfolder="vae")
+    vae = AutoencoderKL.from_pretrained("CompVis/stable-diffusion-v1-4", subfolder="vae", use_safetensors=False)
     
     # Load the tokenizer and text encoder to tokenize and encode the text
     tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")
     text_encoder = CLIPTextModel.from_pretrained("openai/clip-vit-large-patch14")
     
     # The UNet model for generating the latents
-    unet = UNet2DConditionModel.from_pretrained("CompVis/stable-diffusion-v1-4", subfolder="unet")
+    unet = UNet2DConditionModel.from_pretrained("CompVis/stable-diffusion-v1-4", subfolder="unet", use_safetensors=False)
     
     # The noise scheduler
     scheduler = LMSDiscreteScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", num_train_timesteps=1000)
@@ -48,7 +51,8 @@ def load_models(device="cuda"):
     # Load the full pipeline for concept loading
     pipe = StableDiffusionPipeline.from_pretrained(
         "runwayml/stable-diffusion-v1-5",
-        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32
+        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+        use_safetensors=False
     )
     
     # Move models to device