# Runtime dependency set for the local OCI image.
# Native packages listed here are built in the wheel stage and installed from /opt/wheels.

-r /tmp/requirements.txt

# CUDA/quantization extras that are intentionally optional in source installs but useful
# in the full Linux container distribution. flash-attn and causal-conv1d stay out of
# the base image because they are CUDA-arch-sensitive and fragile to build generally.
bitsandbytes>=0.41.0
sentencepiece>=0.1.99
tiktoken>=0.5.0
tokenizers>=0.15.0
protobuf>=3.20.0
optimum-quanto>=0.2.0
gguf>=0.9.0

# Native wheels are built by packaging/linux/Dockerfile.oci with CUDA+Vulkan enabled.
llama-cpp-python>=0.2.0
stable-diffusion-cpp-python>=0.2.0
whispercpp>=0.0.17
