The premise: you should be able to go from a local script (or a Colab notebook) to a distributed job on GCP with zero friction. Just add one line to your code and get your results 10 minutes later after training on 8 GPUs. Or on any hardware configuration you need, really.
-
-
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
You folks are producing goodies faster than my organism can digest them! Moving this to the front of the queue.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Someone on the Tensorflow team should do a demo/tutorial with all the process (Creating GCP project, enabling AI services, etc...) for Linux/Windows... One line code is attractive, but the set up process is not... adding video tutorial will also help in marketing
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Sounds promising, I’ll sure try it out by next week (too bad that GCP ai platform is not compatible with TF > 2.1 btw)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
This is pretty cool
@fchollet , is it also possible to integrate horovod with this? is it part of the roadmap?Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Not many use GCP. I used to use it in one department in previous company but the rest business all use AWS. Current place all use Microsoft. GCP is still relative niche share.
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.