Casper

Hi @ghntd, at the moment, data parallelism is not implemented. I welcome any help on implementing this that demonstrates a speedup.

@robertgshaw2-neuralmagic could you take a look? I think Triton implementations could be interesting to expand access to other hardware.

LiteLLM should adhere to the responses API, making delegation there an impractical option. The parsing in DSPy incorporates specific assumptions about the structure of a response, which unfortunately fail to...

I am still getting errors running the script I referenced in the issue after installing from your branch. It fixes some initial issue, but it seems not fully compatible with...

New example is out for this. https://github.com/deepspeedai/DeepSpeedExamples/blob/592d28fa45c12613f39ed388e043be760707237c/training/tensor_parallel/train.py

@winglian yes, but I seem to remember upgrading to latest accelerate fixed it

> Hi, I ran your code but didn't hit the exact error. Can you share the full stack trace? I just tried making a new environment and reinstalled things, but...

I accidentally deleted this when uploading my script and forgot about it. This line of code is the sole reason the script crashes immediately, but a little while later the...

@bhaskargurram-ai Appreciate the dev experience getting better, but I am not sure this fixes the issue raised here. I am more concerned with not being able to use the web_search...

@TomeHirata looks like the PR for this issue got abandoned. any plans to fix this issue for now?