To do

• lora (torchtune?)
• workflow (metaflow?)
• lr scheduling
• freezing
• mlflow dataset tracking

Running

mlflow ui --waitress-opts '--threads=6 --send-bytes=9000 --max-request-body-size=2000000000 --expose-tracebacks' --backend-store-uri file:mlruns --artifacts-destination mlartifacts

Dagster

dagster dev