add in gpu milvus #166
Merged
add in gpu milvus #166
Conversation
This file contains 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
jperez999
commented
Oct 15, 2024
| # # Turn on to leverage the `vdb_upload` task | ||
| # restart: always | ||
| # container_name: milvus-standalone | ||
| # image: milvusdb/milvus:v2.4.9-gpu |
There was a problem hiding this comment.
This is the new version of milvus that we would need to use for GPU indexing to work.
jperez999
commented
Oct 15, 2024
| # ports: | ||
| # - "19530:19530" | ||
| # - "9091:9091" | ||
| # deploy: |
There was a problem hiding this comment.
Added GPU access to the container for functionality support.
jperez999
commented
Oct 15, 2024
| @@ -37,10 +37,11 @@ def build_default_milvus_config(embedding_size: int = 1024) -> typing.Dict[str, | |||
| "index_conf": { | |||
| "field_name": "vector", | |||
| "metric_type": "L2", | |||
| "index_type": "HNSW", | |||
| "index_type": "GPU_CAGRA", | |||
There was a problem hiding this comment.
I am thinking it is ok to expect GPU support by default, given that the milvus image is also GPU specific. Let me know if we need to be more malleable in this regard.
randerzander
approved these changes
Oct 15, 2024
There was a problem hiding this comment.
Thanks, @jperez999 !
I'm eager to play w/ this in the context of forthcoming bulk indexing
|
@jdye64 can you give this a review/merge? |
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Outdated suggestions cannot be applied.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.
Description
This PR changes the milvus index to the GPU CAGRA index and it uses the parameters that created the best recall for a randomized index (it is possible to tweak this parameter to get better results but that is entirely based on the dataset) which is a good starting point. It also changes the method of identifying the GPUs used in docker-compose file to specify IDs instead of counts and leveraging the CUDA_VISIBLE_DEVICES env variable. I left the variable in place, but changed it to look for GPU ID 0, which is essentially a no op. If we like this new methodology we can remove that env var completely.
Checklist