foundation-model-stack / fms-hf-tuning

🚀 Collection of tuning recipes with HuggingFace SFTTrainer and PyTorch FSDP.
Apache License 2.0
28 stars 48 forks source link

build: Set triton environment variables #370

Closed willmj closed 1 month ago

willmj commented 1 month ago

Description of the change

Add triton environment variables for QLoRA tuning if random user in cluster.

Related issue number

closes #367

How to verify the PR

Use image docker-na-public.artifactory.swg-devops.com/wcp-ai-foundation-team-docker-virtual/sft-trainer:triton_env_vars_ubi9_py311.triton_env_vars and run tuning as a random user without setting these variable manually.

Was the PR tested

github-actions[bot] commented 1 month ago

Thanks for making a pull request! 😃 One of the maintainers will review and advise on the next steps.

willmj commented 1 month ago

@anhuong Triton creates them so they do not have to already be created, but it seems like most don't end up having anything written to them anyways.