We have created a cleaml pipeline using PipelineController.
We use add_function_step to add tasks to the pipeline.
So as to upload artifacts etc. directly to S3 bucket we assign
bucket path to the output_uri parameter in the add_function_step.
We want to totally avoid using fileserver, however even after passing
the S3 bucket path, it first saves the artifact to fileserver and then
uploads it to the bucket.
We also tried specifying the bucket path in the Clearml configuration file