Update app.py

app.py

@@ -27,14 +27,14 @@ from datetime import datetime
 from openai import OpenAI
 
 torch.backends.cuda.matmul.allow_tf32 = True
-torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = False
+torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = True
 torch.backends.cuda.matmul.allow_fp16_reduced_precision_reduction = False
 torch.backends.cudnn.allow_tf32 = True
 torch.backends.cudnn.deterministic = False
 
 torch.backends.cudnn.benchmark = False
 
-torch.set_float32_matmul_precision("medium")
+torch.set_float32_matmul_precision("high")
 
 # Load Hugging Face token if needed
 hf_token = os.getenv("HF_TOKEN")