This project implements efficient inference of Mixtral-8x7B models.
# docker run --gpus 0 -p 5000:5000 -v ~/.cache/huggingface/hub:/root/.cache/huggingface/hub ogmiladyloki/mixtral-offloader
You can give a gpu to each container, there's logic included to decide how many experts to load depending on how much VRAM the GPU has
Original work below:
In summary, we achieve efficient inference of Mixtral-8x7B models through a combination of techniques:
- Mixed quantization with HQQ. We apply separate quantization schemes for attention layers and experts to fit the model into the combined GPU and CPU memory.
- MoE offloading strategy. Each expert per layer is offloaded separately and only brought pack to GPU when needed. We store active experts in a LRU cache to reduce GPU-RAM communication when computing activations for adjacent tokens.
For more detailed information about our methods and results, please refer to our tech-report.
Some techniques described in our technical report are not yet available in this repo. However, we are actively working on adding support for them in the near future.
Some of the upcoming features are:
- Support for other quantization methods
- Speculative expert prefetching