The human-centric platform for production ML & AI

Outerbounds

Hello team,

We have a user trying out the bleeding edge decorator and using `named_env` . But they're experiencing some issues. Any insights?

This is the `conda env` they're using:

```name: llm_sft
channels:
  - conda-forge
  - pytorch
dependencies:
  - python==3.10
  - pip
  - metaflow
  - boto3
  - pytorch=2.1
  - pip:
      - setuptools==59.5.0
      - python-dotenv
      - black
      - torchvision
      - transformers==4.40
      - datasets
      - accelerate
      - evaluate
      - bitsandbytes
      - huggingface_hub
      - trl
      - peft
      - wandb
      - pytest
      - metaflow-netflixext```

Nvm - got it resolved. Discrepancy between Amazon Linux 2 (where the virtual environment was created) and Ubuntu (the docker container of AWS Batch)

cool — sorry am out so didn’t see this but glad you got it figured out. As a quick note, you typically don’t need to include metaflow (or any extension) in the requirements.txt (or env.yml) — it gets auto included.