waldburger/farood-pilot / install_and_probe.sh
waldburger's picture
download
raw
899 Bytes
#!/bin/bash
set -x
pip install -q --no-cache-dir evo2 || exit 1
ABI=$(python -c 'import torch; print(torch._C._GLIBCXX_USE_CXX11_ABI)')
echo "torch cxx11 ABI: $ABI"
if [ "$ABI" = "True" ]; then W1=TRUE; W2=FALSE; else W1=FALSE; W2=TRUE; fi
pip install -q --no-cache-dir "https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.6cxx11abi${W1}-cp311-cp311-linux_x86_64.whl" \
|| pip install -q --no-cache-dir "https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.6cxx11abi${W2}-cp311-cp311-linux_x86_64.whl" \
|| pip install -q --no-cache-dir "https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abi${W1}-cp311-cp311-linux_x86_64.whl" \
|| pip install -q --no-cache-dir flash-attn --no-build-isolation \
|| exit 1
python /workspace/probe_evo2.py

Xet Storage Details

Size:
899 Bytes
·
Xet hash:
b759d276c0ae06c8f0a95fb72233db9ab74ad488de4c3f5a07bbaae3e3ebd238

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.