Woosuk Kwon
Woosuk Kwon
@concretevitamin Got it. Makes sense. Thanks!
@concretevitamin Thanks for the pointer to the related issue. `make -j$(nproc)` and `python train.py --num_workers $(nproc)` were real examples from my personal experience. I believe this should be at least...
> Future TODO: We may want to consider adding the sky storages with object store to the input automatically so that the optimizer can take the data movement of the...
Thanks for quick review @concretevitamin! Addressed all comments. PTAL. BTW, I realized there is a potential problem in this approach. While this is NOT the case for our current catalogs,...
I worked well on local clusters as well. @concretevitamin Can I merge this?
@istoica Could you add more? When do you think we might want to allow different names for the same accelerator?
> BTW, I realized there is a potential problem in this approach. While this is NOT the case for our current catalogs, the clouds may add the same accelerator with...
Thanks for your review @romilbhardwaj . I've addressed your comments. PTAL.
Hey @iojw, thanks for submitting this PR! Just curious: what's the expected usage? Is it for real-time inference or offline inference? Is the inference request repeated over time? And why...
@concretevitamin > RE (2): This is a great example workflow where people just debug their requirements.txt (Justin et al). We should come up with a way to make this work....