Ma, Guokai

> @delock, is this PR still actively developed for merging? Hi @tjruwase, We have validated this branch in our environment. The latest Intel extension for DeepSpeed works with this PR...

Hi @tjruwase , want to know whether this PR is in the merge queue or still need some changes. Currently DeepSpeed engine is already integrated with accelerator abstraction and the...

> @delock it feels like we could add a pre-commit hook to ensure that our formatter fails if someone tries to use `torch.cuda` outside the new get_accelerator api. Similar to...

https://github.com/microsoft/DeepSpeed/pull/2981 is created for pre-commit check. @jeffra

Hi @ezyang, this PR needs approval from maintainer to be merged. Can you help review this PR? Thanks!

I used the following commands and it still works. I don't have full detail about your issue. But some suggestions: 1. Makesure gstuil is installed for python2 rather than python3...

Google had reorgnized the directory to accomondate training v0.7 submission, thus the directory structure no longer compatible with previous submission code for v0.6. The directory path for v0.6 checkpoint is...

Hi, thanks for this example! When I keep talking with it I got this error, is there a way to avoid this error? ``` ValueError: Requested tokens (527) exceed context...

@awan-10 @lekurile Thanks for start this thread. I met this error when I tried to run this example on Xeon server with CPU. I suspect this is a configuration issue....

Hi @lekurile Now I can start the server from seperate command line and run benchmark on this server with reduced test size (max batch 128, avg prompt128) to start with....