KeitaW
commented
3 months ago Interim Solution for Llama3 Training
As discussed with @syedazi, we have decided to implement a temporary measure to address our current needs:
- We will incorporate the NxD codebase as an interim solution.
- This arrangement will remain in place until the release of Neuron SDK 2.19.
Post-Release Actions
Once Neuron SDK 2.19 becomes available:
- The temporary NxD codebase will be removed from our repository.
- We will update the README file to provide comprehensive guidance for users.
- This update will include instructions on how to train Llama3 using the latest NxD version.
Description of changes: This is an example for running Llama 3 8B and 70B model on SageMaker Hyperpod using Trainium instances.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.