Tensorboard: Support environment variable S3_ENDPOINT in tensorboard without tensorflow

[lgeo3]

When using tensorboard with tensorflow installed, we are able to set the s3_endpoint using S3_ENDPOINT environment variable. It is particularly usefull when running a local s3 instance like minio.

But when using the tensorboard package (without tensorflow), it uses the boto3 client, and don't use the S3_ENDPOINT variable.

A simple modification would be to replace boto3.client/session call with the same call but with the parameter: endpoint_url=os.environ.get('S3_ENDPOINT', None)

What do you think ?

[stephanwlee]

That sounds reasonable. Would you like to propose the change via PR?

@nfelt @orionr FYI

[thealphacod3r]

@lgeo3 @stephanwlee Is this closed???

[lgeo3]

I didn't manage to do the pull request [no cla on my emloyer side], but the code is really simple (i.e A simple modification would be to replace boto3.client/session call with the same call but with the parameter: endpoint_url=os.environ.get('S3_ENDPOINT', None)) so if someone could write the code it could be nice. Otherwise we could close the issue, not sure a lot of people need it.
sorry.

[thealphacod3r]

@lgeo3 Please give a look to this!

[lgeo3]

Seems good to me :+1:. Thanks.