[Fix][Grammar] Detach token table to prevent disposing it #571
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
Co-authored-by: SMarioMan <[email protected]>
This was referenced Sep 23, 2024
Closed
CharlieFRuan
added a commit
that referenced
this pull request
Sep 23, 2024
No breaking changes or API-related changes. ### Changes - #571 - #570 - This is the major change in this version, where we now depend on the packaged TVMjs: `@mlc-ai/web-runtime`. This only changes how users build WebLLM from source - Since `@mlc-ai/web-runtime` is not an official package from `apache/tvm`, we record the commit that it is built upon in the WebLLM version bump PR, like below ### TVMjs - Version `0.18.0-dev0` - Built at apache/tvm@30fb16a
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Outdated suggestions cannot be applied.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.
Prior to this PR, when reusing the same engine but for different schema will run into error "Module has already been disposed". An example to reproduce this is included in #560.
This is because
this.tokenTableis atvmjs.TVMObjectand will be disposed after the scope ends.We fix this by wrapping the call with
this.tvm.detachFromCurrentScope(), and only disposethis.tokenTablewhen we dispose the entireLLMChatPipeline.