fix ray install (#1630)

more fixes to work with runpod + skypilot (#1629)

cloud image w/o tmux (#1628)

install rsync too (#1627)

fix symlinks for axolotl outputs (#1625)

bump versions of deps (#1621)

update outputs path so that we can mount workspace to /workspace/data (#1623)

make sure to install causal_conv1d in docker (#1459)

support galore once upstreamed into transformers (#1409)

deprecate py 3.9 support, set min pytorch version (#1343) [skip ci]

make mlflow optional (#1317)

add a helpful motd for cloud image (#1235) [skip ci]

Revert "run PR e2e docker CI tests in Modal" (#1220) [skip ci]

run PR e2e docker CI tests in Modal (#1217) [skip ci]

upgrade deepspeed to 0.13.1 for mixtral fixes (#1189) [skip ci]

jupyter lab fixes (#1139) [skip ci]

Dockerfile cloud ports (#1148)

Agnostic cloud gpu docker image and Jupyter lab (#1097)

misc fixes from #943 (#1086) [skip ci]

fix: warn user to install mamba_ssm package (#1019)

attempt to also run e2e tests that needs gpus (#1070)

Add tests to Docker (#993)

Dockerfile torch fix (#987)

fix for build for nccl in dockerfile (#970)

update to latest nccl in docker image (#965)

Mixtral multipack (#928)

don't compile deepspeed or bitsandbytes from source (#837)

add deepspeed-kernels dependency for deepspeed>=0.12.0 (#827)

fix pytorch 2.1.0 build, add multipack docs (#722)

apex not needed as amp is part of pytorch (#696)

fix multiline for docker (#694)

Feat: Set WORKDIR to /workspace/axolotl (#679)

tweak: improve base builder for smaller layers (#500)

let MAX_JOBS use the default since we're not resource constrained on our self-hosted runners (#427)

update dockerfile to not build evoformer since it fails the build (#607)

update readme to point to direct link to runpod template, cleanup install instrucitons (#532)

Add support for GPTQ using native transformers/peft (#468)

remove --force-reinstall from Dockerfile to ensure correct pytorch version (#492)

flash attn pip install (#426)

bump flash-attn to 2.0.4 for the base docker image (#382)

add peft install back since it doesn't get installed by setup.py (#331)

pin accelerate so it works with llama2 (#330)

Prune cuda117 (#327)

add runpod envs to .bashrc, fix bnb env (#316)

pin flash attention 2 to the fix for backwards pass

flash attention 2

explicitly pin flash attention 1 to v1.0.9

misc fixes

set transformers cache env var in docker image

git fetch fix for docker