Add debugging info and Zero GPU hardware suggestion

- Add GPU availability debug prints
- Show GPU name and memory when available
- Add note about Zero GPU settings when running on CPU
- Add suggested_hardware: zero-a10g in README.md
- This helps diagnose why GPU is not being allocated

README.md

@@ -8,6 +8,7 @@ sdk_version: 5.34.2
 app_file: app.py
 pinned: false
 license: mit
+suggested_hardware: zero-a10g
 models:
   - PascalNotin/Tranception_Small
   - PascalNotin/Tranception_Medium

app.py

@@ -268,16 +268,21 @@ def score_and_create_matrix_all_singles_impl(sequence,mutation_range_start=None,
     model = tranception.model_pytorch.TranceptionLMHeadModel.from_pretrained(pretrained_model_name_or_path=model_path)
   
   # Device selection - Zero GPU will provide CUDA when decorated with @spaces.GPU
+  print(f"GPU Available: {torch.cuda.is_available()}")
   if torch.cuda.is_available():
     device = torch.device("cuda")
     model = model.to(device)
-    print(f"Inference will take place on {torch.cuda.get_device_name(0)}")
+    gpu_name = torch.cuda.get_device_name(0)
+    gpu_memory = torch.cuda.get_device_properties(0).total_memory / 1024**3
+    print(f"Inference will take place on {gpu_name}")
+    print(f"GPU Memory: {gpu_memory:.2f} GB")
     # Increase batch size for GPU inference
     batch_size_inference = min(batch_size_inference, 50)
   else:
     device = torch.device("cpu")
     model = model.to(device)
     print("Inference will take place on CPU")
+    print("Note: If you expected GPU, ensure Zero GPU is enabled in Space settings")
     # Reduce batch size for CPU inference
     batch_size_inference = min(batch_size_inference, 10)