Why do I need metadata? (TensorFlow Extended)

On today’s episode of TensorFlow Extended hosted by TensorFlow Developer Advocate, Robert Crowe, we’re asking the question, “Why do I need metadata?” Learn more on Episode 3 of the 5 part series on Real World Machine Learning in Production and a preview for Episode 4!

This video is also subtitled in Chinese, Indonesian, Italian, Japanese, Korean, Portuguese, and Spanish.

[dr_flunks]

really appreciate all this Google Team.

[armandt]

TFX metadata are great independent layered components and its schema allows isolation operations on every single component!

[oliverli]

yay! we don't need to hire a VPS to run the entirety of the training in one go, but use our expensive GPUs locally and save it upon computer shutdown and load it back tomorrow to continue heehee

[harishreddy]

lineage meaning 'direct descent from an ancestor; ancestry or pedigree'

[siddharthkotwal]

Great tutorial. I don't understand the first functionality, why would you need to track forward-backwards?

[huzhixiang]

Sounds quite useful for all kinds of pipeline. Do you plan to publish API documents about ML Metadata?

[jallaaswini]

Hi, Can you please tell me what are the libraries for ML Metadata connecting to PostgreSQL.

Will tfx metadata provide storage for PostgreSQL? Or only to SQLite and MySQL