fix: Upgrade transformers to 4.56.0 to fix NV-Reason-CXR dtype error

Root cause: We were using transformers 4.51.3, but NVIDIA's official
demo uses 4.56.0. The older version has a bug where the 'dtype' parameter
causes JSON serialization errors.

Solution:
- Upgrade transformers from 4.51.3 to 4.56.0 (matches NVIDIA demo)
- Use NVIDIA's EXACT loading code with dtype=torch.bfloat16
- This is the same version and code that works in NVIDIA's official Space

This should finally resolve the persistent NV-Reason-CXR loading error.

Reference: https://huggingface.co/spaces/nvidia/nv-reason-cxr

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

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

medrax/tools/nv_reason_cxr.py

@@ -66,19 +66,19 @@ class NVReasonCXRTool(BaseTool):
         super().__init__()
         self.device = device
 
-        # Load model following NVIDIA's official demo code
+        # Load model following NVIDIA's official demo code EXACTLY
+        # Requires transformers==4.56.0 (same as NVIDIA's demo)
         try:
             print(f"Loading NV-Reason-CXR model from {model_path}...")
             print(f"Using device: {self.device}")
-            print("Following NVIDIA's exact loading pattern from official demo")
+            print("Using NVIDIA's exact loading pattern with transformers 4.56.0")
 
-            # NVIDIA's approach but adapted for our environment
-            # The 'dtype' parameter works in their Gradio Space environment
-            # but causes JSON serialization issues in our setup.
-            # Solution: Load with default dtype, then convert to bfloat16 manually
+            # Match NVIDIA's demo exactly - requires transformers 4.56.0
+            # The dtype parameter works correctly in newer transformers versions
             self.model = AutoModelForImageTextToText.from_pretrained(
-                model_path,
-            ).to(dtype=torch.bfloat16).eval().to(self.device)
+                pretrained_model_name_or_path=model_path,
+                dtype=torch.bfloat16,
+            ).eval().to(self.device)
 
             self.processor = AutoProcessor.from_pretrained(
                 model_path,

requirements.txt

@@ -21,7 +21,7 @@ Pillow>=8.0.0
 PyPDF2>=3.0.0
 pdfplumber>=0.10.0
 torchxrayvision>=0.0.37
-transformers==4.51.3
+transformers==4.56.0
 datasets>=2.15.0
 tokenizers>=0.21,<0.22
 sentencepiece>=0.1.95