fix: Add safetensors support and improve model loading

**Model Loading Improvements:**
- Add safetensors>=0.4.0 to requirements for modern model format support
- Add accelerate>=0.20.0 for optimized model loading
- Enable trust_remote_code for tokenizer and model loading
- Add low_cpu_mem_usage=True to reduce memory footprint
- Fix torch_dtype deprecation warning (use dtype instead)

**Technical Changes:**
- Support both safetensors and pytorch_model.bin formats
- Better memory management for large models
- Enable remote code execution for special tokenizers
- Improved compatibility with HuggingFace Hub

This fixes the model loading error on Hugging Face Spaces:
"Can't load the model for 'microsoft/DialoGPT-small'"

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

app.py

@@ -73,18 +73,21 @@ def load_model(model_name):
             tokenizer = AutoTokenizer.from_pretrained(
                 model_name,
                 token=HF_TOKEN,
-                padding_side='left'
+                padding_side='left',
+                trust_remote_code=True
             )
 
             # Add pad token if missing
             if tokenizer.pad_token is None:
                 tokenizer.pad_token = tokenizer.eos_token
 
-            # Load model
+            # Load model with safetensors support
             model = AutoModelForCausalLM.from_pretrained(
                 model_name,
                 token=HF_TOKEN,
-                torch_dtype=torch.float32,
+                dtype=torch.float32,
+                low_cpu_mem_usage=True,
+                trust_remote_code=True
             )
             model.to(device)
             model.eval()

requirements.txt

@@ -1,3 +1,5 @@
 gradio>=5.0.0
 transformers>=4.30.0
 torch>=2.0.0
+safetensors>=0.4.0
+accelerate>=0.20.0