Skip to content

Training job conflict error when installing local src #3721

@cirofdo

Description

@cirofdo

I'm setting up a simple pipeline with preprocessing, training and inference.
But when it gets on the training job I get an error when trying to install local src lib withing the requirements.txt. This is the error:
ERROR: Cannot install src 0.0.0 (from /opt/ml/code) because these package versions have conflicting dependencies.

    ├── pipeline.py    <- file that initiates the pipeline
    ├── requirements.txt 
    │
    └── src 
        │
        ├── data
        │   └── make_dataset.py
        │
        └─ models
            ├── predict_model.py
            └── train_model.py

On the pipeline.py I first run the make_dataset.py and it runs without error.
Then on the training job I pass entry_point="src/models/train_model.py" and source_dir=".". Then it runs pip install with the requirements.txt (that contains a -e . to install src)

The full error:
image

Weird because why it gives conflict error even when there is no other src installed.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions