Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix no worker node exception for remote embedding model #1482

Merged
merged 2 commits into from
Oct 11, 2023
Merged

fix no worker node exception for remote embedding model #1482

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
99e34a0
fix no worker node exception for remote embedding model
ylwu-amzn Oct 11, 2023
ef8da7b
only add model info to cache if model cache exist
ylwu-amzn Oct 11, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion plugin/src/main/java/org/opensearch/ml/action/prediction/TransportPredictionTaskAction.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,9 +137,16 @@ private void executePredict(
String requestId = mlPredictionTaskRequest.getRequestID();
log.debug("receive predict request " + requestId + " for model " + mlPredictionTaskRequest.getModelId());
long startTime = System.nanoTime();
// For remote text embedding model, neural search will set mlPredictionTaskRequest.getMlInput().getAlgorithm() as
// TEXT_EMBEDDING. In ml-commons we should always use the real function name of model: REMOTE. So we try to get
// from model cache first.
FunctionName functionName = modelCacheHelper
.getOptionalFunctionName(modelId)
.orElse(mlPredictionTaskRequest.getMlInput().getAlgorithm());
mlPredictTaskRunner
.run(
mlPredictionTaskRequest.getMlInput().getAlgorithm(),
// This is by design to NOT use mlPredictionTaskRequest.getMlInput().getAlgorithm() here
functionName,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should fine-tune this part either in neural search side or in ml-commons side, more and more twisted code could cause more and more issues.

mlPredictionTaskRequest,
transportService,
ActionListener.runAfter(wrappedListener, () -> {
Expand Down
6 changes: 4 additions & 2 deletions plugin/src/main/java/org/opensearch/ml/model/MLModelCacheHelper.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -431,8 +431,10 @@ public boolean getDeployToAllNodes(String modelId) {
}

public void setModelInfo(String modelId, MLModel mlModel) {
MLModelCache mlModelCache = getExistingModelCache(modelId);
mlModelCache.setModelInfo(mlModel);
MLModelCache mlModelCache = modelCaches.get(modelId);
if (mlModelCache != null) {
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@zane-neo , we should check if model cache exist or not

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we also need to fine tuning the error message in getExistingModelCache method, when model is not in cache, it should be a not deployed model, we should return the should deploy model info, so that in the future other method can use this method safely.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sounds good, we can do that later. One note: not all models needs to be deployed first

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For current code, I think either local or remote needs to be deployed first, are you saying some feature will be implemented in the future?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

no , kmeans doesn't need deploy

mlModelCache.setModelInfo(mlModel);
}
}

public MLModel getModelInfo(String modelId) {
Expand Down
Loading