No such media/table/ downloading media table via api

Hello, This a reply from our support bot designed to assist you with your Weights & Biases related queries. To reach a human please reply to this message.

To download a media table using the Weights & Biases API, you’re on the right track by accessing the sweep and iterating through its runs. However, it seems like there might be a misunderstanding or a missing step in how you’re trying to access the table data from the run’s summary.

Here’s a more detailed approach to access and download a table logged with wandb.log:

  1. Initialize the W&B API and Access the Run: First, ensure you have the correct project and run ID to access the specific run that contains the table you want to download.

    import wandb
    import json
    
    api = wandb.Api()
    run_path = "your_username/project_name/run_id"  # Replace with your actual path
    run = api.run(run_path)
    
  2. Access the Table from the Run Summary: If you’ve logged the table using wandb.log, it should be accessible through the run’s summary. However, the method you’ve described seems to be missing some steps or might contain inaccuracies in accessing the table’s path.

    Instead, you should directly access the table artifact if it’s logged as an artifact. If it’s logged in the summary, you would typically access it via the summary attributes directly, but for complex structures like tables, they are usually logged as artifacts.

  3. Download the Table Artifact: If the table was logged as an artifact, you can access and download it using the following approach:

    artifact = run.use_artifact('artifact_name:version')  # Replace 'artifact_name:version' with your table artifact name and version
    table = artifact.get('table_name')  # Replace 'table_name' with the name of your table
    

    If the table was not logged as an artifact but directly in the summary in a way that’s not standard, you might need to adjust your approach based on how the data is stored.

  4. Alternative Method: If you’re facing issues with the direct approach, consider logging your tables as artifacts for easier access and versioning. Here’s how you can log a table as an artifact:

    import wandb
    import pandas as pd
    
    # Assuming times_df is your DataFrame
    times_df = pd.DataFrame(...)  # Your DataFrame code here
    
    wandb.init(project='your_project_name', job_type='data-logging')
    table = wandb.Table(dataframe=times_df.describe())
    artifact = wandb.Artifact('times_table_describe', type='dataset')
    artifact.add(table, 'times_table_describe')
    wandb.log_artifact(artifact)
    wandb.finish()
    

    This way, your table is stored as an artifact, making it easier to version, share, and download.

Common Issues When Downloading Media Tables via the W&B API:

  • Incorrect path or artifact name.
  • Accessing summary data that’s not directly downloadable.
  • Permissions or API key issues.

Remember: Always ensure your W&B API key is correctly set up in your environment to authenticate your requests.

Unfortunately, the provided context does not include specific methods for downloading media tables directly via the API, especially if they are logged in a manner not standard or expected by the API’s design. If the issue persists, consider reaching out to W&B support or the community forums for more targeted assistance.

To reach a human please reply to this message.

-WandBot :robot:

HI, unfortunately in the previous projects I was logging wandb tables directly and not as artefacts. Is there support to recover/download tables logged in this manner? I appreciate artefacts now seems to be the preferred approach but my I can’t make the retrospective amendments to my experiments. Many thanks!

Dear James, thank you for reaching out to Weights and Biases! Could you please provide the name of the entity and project you are trying to retrieve your data from?
Please also provide the specific name of the run as well.

Also, as a heads up, every time you log something as a table, in our end it gets logged also as an artifact type.

You can find that under Artifacts > run_table > v0 latest > Files > click on the json file. You can manually download it there from the UI.

Let me know if that works on your end!

Hey James,

Let us know if you were able to get your json files with the suggestions above.

Best,
Vitoria

Hey James, since we have not heard back from you we are going to close this request.

If you would like to re-open the conversation, please let us know!

Best,
Vitoria