Set model status using torchserve api #1878
Merged
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
Codecov Report
@@ Coverage Diff @@
## master #1878 +/- ##
==========================================
+ Coverage 41.67% 44.95% +3.28%
==========================================
Files 55 63 +8
Lines 2282 2609 +327
Branches 1 56 +55
==========================================
+ Hits 951 1173 +222
- Misses 1331 1436 +105
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
@jagadeeshi2i @byeongjokim @maaquib Do you know if this PR will be taken to complete state? Even we have hit a scenario that would need readiness check from the model side and this PR implements it. |
|
@gavrishp @jagadeeshi2i @maaquib |
|
Logs with llama model |
agunapal
approved these changes
Jan 11, 2024
auto-merge was automatically disabled
January 24, 2024 11:11
Head branch was pushed to by a user without write access
|
@agunapal |
|
Thank you for processing the workflows. |
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Outdated suggestions cannot be applied.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
Fixes #1773, kserve/kserve#1915
Issue:
1773 is the PR that tries to set the model status after TS ISVC is deployed. However, it causes problems.
In an initialization function of TS handler, you can do many things to prepare the model such as loading the model weight. (e.g., download tokenizer weight from online storage)
The current code may cause an error if someone calls the Kserve API before TS workers are initialized.
In particular, when the pod is scaled-out with readinessProbe(/v1/model/{mode_name} or /v2/model/{model_name}/status) set.
However, there is also a problem with the previous code.
It calls model.load() at the first request. In that case, an infinite cycle occurs between the ready status and the requests.
Therefore, to solve this problem, the model's ready status should be determined after communication with the torchserve Describe Model API.
Two types of API are used in
loadfunction.One is for checking only the worker's status. (when MODEL_LOAD_CUSTOMIZED is 'false')
It is used when the initialization of the handler does not take a long time.
The other is for checking the status of both the handler and the worker. (when MODEL_LOAD_CUSTOMIZED is 'true')
The
describe_handlefunction should be implemented in the custom handler. RelatedIssueIt is used when the initialization of the handler takes a long time.
Type of change
Feature/Issue validation/testing
log.log
Truein kserve wrapperChecklist: