You can now experiment & iterate on your model in a Kaggle notebook, and once it's ready to train at scale, launch a training job on Google Cloud using as much compute as you want -- just by adding one line.
-
-
Show this thread
-
New conversation -
-
-
@Aclopez16 Hay que revisar Kaggle.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
ML + the Cloud is a powerful combo
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
This is good news, but my assumption is you must first have a GCP account activated for billing. Is this service in the same billing tier as running it in GCP notebooks directly or is this a new service with a new billing tier and is this more economic for small businesses?
-
Yes, it requires an active and billable GCP account. The cost of using TFC is the same as the cost of the underlying compute instances being used -- TFC is simply an automation layer that saves you from having to manually launch and shut down these instances.
- Show replies
New conversation -
-
-
As an end-user who struggles with infrastructure management (i.e. kubernetes) and distributed training this has been one of my favorite releases. Great to see it integrated with Kaggle now.https://twitter.com/mchrestkha/status/1313002599948324865?s=20 …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Any chance there will be support for preemptible instances? This was the biggest reason we rolled our own GKE setup. 3x cost savings.
-
Possibly in the future, yes. Not in the immediate roadmap though
End of conversation
New conversation -
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
