fix: Fix for Ray deployment in nemo-run script #266
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Abhishree <[email protected]>
…ialization errors Signed-off-by: Abhishree Thittenamane <[email protected]>
Signed-off-by: Abhishree <[email protected]>
athitten
commented
Nov 4, 2025
| --tensor_parallelism_size {tensor_model_parallel_size} \ | ||
| --pipeline_parallelism_size {pipeline_model_parallel_size} \ | ||
| --tensor_model_parallel_size {tensor_model_parallel_size} \ | ||
| --pipeline_model_parallel_size {pipeline_model_parallel_size} \ |
There was a problem hiding this comment.
Use the same name between Pytriton and Ray script for TP, PP. PR in ED for this: NVIDIA-NeMo/Export-Deploy#501
athitten
commented
Nov 4, 2025
scripts/evaluation_with_nemo_run.py
Outdated
| "max_batch_size": args.batch_size, | ||
| "devices": args.devices, | ||
| "max_batch_size": args.batch_size, #TODO check in llama 405B run | ||
| "num_gpus": args.devices if args.serving_backend == "pytriton" else args.devices * args.nodes, |
There was a problem hiding this comment.
Ray requires num_gpus to be total num of gpus in case of multi node.
Signed-off-by: Abhishree Thittenamane <[email protected]>
Signed-off-by: Abhishree Thittenamane <[email protected]>
athitten
changed the title
|
/ok to test 522994a |
Signed-off-by: Abhishree <[email protected]>
|
/ok to test a4d4956 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adds fixes for nemo-run script to enable nemo-fw checkpoint deployment with Ray backend. Works for single node, multiple replicas and multi node, single replica. Does not work for multi node, multi replica case. This will be addressed in a follow up PR.
Some accuracy numbers (GSM8k, 10%) attained using this script for ray serve backend: