Closed Prasanna-iq closed 1 month ago
@Prasanna-iq optimum-neuron supports Mistral for training. If you want to do inference on Inferentia/Trainium, check out https://huggingface.co/aws-neuron/Mistral-neuron
For Mixtral, this project needs Mixtral support in transformer-neuronx, which is almost there.
@jimburtoft any news about this issue ? My goal is to export Mixtral 8x7b from pytorch format to neuron, while using Inf1 instance on AWS. I came to understand that optimum-neuronx/transformer-neuronx are dedicated to "Inf2, Trn1, Trn1n" instances, which aren't available in my AWS region.
This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread. Thank you!
As the newly launched Neuron SDK 2.18 has supported Mixtral 8x7B model (transformer_neuronx has supported this). It would be great to have this feature added in Optimum Neuron. Thanks~ 🚀
Is there a timeline when the Mixtral support will be added?
This is already merged on main if you want to give it a try. It should be included in the next release.
This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread. Thank you!
This is already merged on main if you want to give it a try. It should be included in the next release.
So this issue can be closed?
Yes, Mixtral is supported since 0.0.22.
Hi, when can we expect the support for Mixtral 8x7B ? I can see Mistral 7B is not supported yet. Thanks