mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-21 23:12:07 +00:00
b2b5df0e94
This PR adds support for AMD Instinct MI210 & MI250 GPUs, with paged attention and FAv2 support. Remaining items to discuss, on top of possible others: * Should we have a `ghcr.io/huggingface/text-generation-inference:1.1.0+rocm` hosted image, or is it too early? * Should we set up a CI on MI210/MI250? I don't have access to the runners of TGI though. * Are we comfortable with those changes being directly in TGI, or do we need a fork? --------- Co-authored-by: Felix Marty <felix@hf.co> Co-authored-by: OlivierDehaene <olivier@huggingface.co> Co-authored-by: Your Name <you@example.com>
16 lines
743 B
Plaintext
16 lines
743 B
Plaintext
flash_att_commit := 3a9bfd076f98746c73362328958dbc68d145fbec
|
|
|
|
flash-attention:
|
|
# Clone flash attention
|
|
pip install -U packaging ninja --no-cache-dir
|
|
git clone https://github.com/HazyResearch/flash-attention.git
|
|
|
|
build-flash-attention: flash-attention
|
|
cd flash-attention && git fetch && git checkout $(flash_att_commit)
|
|
cd flash-attention && python setup.py build
|
|
cd flash-attention/csrc/rotary && python setup.py build
|
|
cd flash-attention/csrc/layer_norm && python setup.py build
|
|
|
|
install-flash-attention: build-flash-attention
|
|
pip uninstall flash_attn rotary_emb dropout_layer_norm -y || true
|
|
cd flash-attention && python setup.py install && cd csrc/layer_norm && python setup.py install && cd ../rotary && python setup.py install |