Update app.py

app.py

@@ -7,13 +7,11 @@ Optimized for Hugging Face Spaces with lazy loading and fallbacks
 import os
 
 # Configure environment for CPU-only and memory optimization
-os.environ['CUDA_VISIBLE_DEVICES'] = ''  # Force CPU-only
-os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:128'
+os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:512'
 os.environ['OMP_NUM_THREADS'] = '2'  # Limit CPU threads
 os.environ['MKL_NUM_THREADS'] = '2'
 os.environ['NUMEXPR_NUM_THREADS'] = '2'
 os.environ['TOKENIZERS_PARALLELISM'] = 'false'  # Avoid tokenizer warnings
-
 # Batchalign specific settings
 os.environ['BATCHALIGN_CACHE'] = '/tmp/batchalign_cache'
 os.environ['HF_HUB_CACHE'] = '/tmp/hf_cache'  # Use tmp for model cache