Torch debugging

app.py

@@ -26,6 +26,12 @@ vardecoder_model = AutoModelForCausalLM.from_pretrained(
     device_map="cuda",
 )
 print("Loaded vardecoder model successfully.")
+
+# Check memory after first model
+print(f"GPU memory after vardecoder:")
+print(f"Allocated: {torch.cuda.memory_allocated() / 1024**3:.2f} GB")
+print(f"Reserved: {torch.cuda.memory_reserved() / 1024**3:.2f} GB")
+
 try:
     fielddecoder_model = AutoModelForCausalLM.from_pretrained(
         "ejschwartz/resym-fielddecoder",