# Server image already provides PyTorch; install LaKun and its other packages.
-e .
transformers==4.57.3
accelerate>=0.26,<2
safetensors>=0.4
huggingface-hub>=0.34,<1
Pillow>=10
numpy>=1.24
