Restore GPU version - remove CPU optimizations and restore GPU-compatible dependencies

app.py

@@ -15,15 +15,13 @@ print(f"Using device: {device}")
 
 model = Qwen2VLForConditionalGeneration.from_pretrained(
     MODEL_PATH,
-    torch_dtype=torch.float32 if device == "cpu" else torch.bfloat16,
-    device_map="auto" if device == "cuda" else None,
-    low_cpu_mem_usage=True if device == "cpu" else False,
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
 )
 processor = AutoProcessor.from_pretrained(MODEL_PATH)
 
 # Move model to device
-if device == "cpu":
-    model = model.to(device)
+model = model.to(device)
 
 temp_generation_config = GenerationConfig(
     max_new_tokens=1024,

requirements.txt

@@ -1,28 +1,13 @@
-# gradio==5.42.0
-# transformers==4.36.0
-# --find-links https://download.pytorch.org/whl/torch_stable.html
-# torch==2.1.0+cpu
-# torchvision==0.16.0+cpu
-# torchaudio==2.1.0+cpu
-# requests>=2.31.0
-# Pillow>=10.0.0
-# accelerate>=0.20.0
-# safetensors>=0.3.0
-# tokenizers>=0.15.0
-# numpy<2.0.0
-# scipy>=1.10.0
-# qwen-vl-utils
-# ipython>=8.0.0
-# Core dependencies
-torch>=2.1.0
-transformers>=4.41.0
-accelerate>=0.31.0
-gradio>=4.29.0
-requests
-
-# Qwen VL utilities
-git+https://github.com/QwenLM/Qwen-VL.git
-
-# Optional but recommended for HF Spaces stability
-sentencepiece
-protobuf
+gradio==5.42.0
+transformers>=4.40.0
+torch>=2.0.0
+torchvision>=0.15.0
+requests>=2.31.0
+Pillow>=10.0.0
+accelerate>=0.20.0
+safetensors>=0.3.0
+tokenizers>=0.15.0
+numpy>=1.24.0
+scipy>=1.10.0
+qwen-vl-utils
+ipython>=8.0.0