requirements.txt

# Requirements for Hugging Face Spaces deployment
# Visual-CoT Demo with Zero GPU

# Core dependencies
torch==2.1.2
torchvision==0.16.2
transformers==4.37.2
tokenizers==0.15.1
sentencepiece==0.1.99

# Gradio and Spaces (use latest version)
gradio  # Latest version with all security updates
spaces>=0.19.4

# Model dependencies
accelerate==0.21.0
einops==0.6.1
einops-exts==0.0.4
timm==0.6.13

# Flash Attention via Kernels (pre-built, Spaces-compatible)
kernels>=0.0.1  # For fast attention without compilation

# Utilities
Pillow>=10.0.0
numpy>=1.24.0
requests>=2.31.0

# Optional optimizations (comment out if causing issues)
# xformers>=0.0.22  # For memory-efficient attention
# flash-attn>=2.3.0  # For faster attention (may need CUDA)