Update app.py

app.py

@@ -1,15 +1,31 @@
+
 import subprocess
 subprocess.run(['sh', './spaces.sh'])
+import spaces
 
 import os
 
 os.environ['PYTORCH_NVML_BASED_CUDA_CHECK'] = '1'
 os.environ['TORCH_LINALG_PREFER_CUSOLVER'] = '1'
-os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True,pinned_use_background_threads:True'
+os.environ['PYTORCH_ALLOC_CONF'] = 'expandable_segments:True,pinned_use_background_threads:True'
 os.environ["SAFETENSORS_FAST_GPU"] = "1"
 os.environ['HF_HUB_ENABLE_HF_TRANSFER'] = '1'
 
-import spaces
+import torch
+
+torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = False
+torch.backends.cuda.matmul.allow_fp16_reduced_precision_reduction = False
+torch.backends.fp32_precision = "ieee"
+torch.backends.cuda.matmul.fp32_precision = "ieee"
+torch.backends.cudnn.fp32_precision = "ieee"
+torch.backends.cudnn.conv.fp32_precision = "ieee"
+torch.backends.cudnn.rnn.fp32_precision = "ieee"
+torch.backends.cudnn.deterministic = False
+torch.backends.cudnn.benchmark = False
+torch.backends.cuda.preferred_blas_library="cublas"
+torch.backends.cuda.preferred_linalg_library="cusolver"
+torch.set_float32_matmul_precision("highest")
+
 import gradio as gr
 import numpy as np
 import random