[Device] Catch WebGPU OOM error #402
Merged
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
Member
Author
|
Example of trying to allocate a KV cache with 900k context length (should be similar for trying to load a model that is too large): |
Member
Author
|
Marked as a draft for now as it depends on apache/tvm#17005 |
CharlieFRuan
added a commit
that referenced
this pull request
May 21, 2024
### Changes Main changes include: - New model `Hermes-2-Pro-Mistral-7B` in `prebuiltAppConfig` via: - #390 - Various `index.js` and `index.js.map` post-processings to resolve frontend compatibility issues with `require()` and `perf_hoooks` - #397 - #406 - Catch WebGPU OOM error upon `reload()` and `CreateEngine()`: - #402 - Service Worker support (in addition to Extension Service Worker): - #395 - #400 - #401 ### WASM Version v0_2_34 as no change is required. ### TVMjs TVMjs compiled at apache/tvm@a5862a5, with only one change in `tvm/web`: apache/tvm#17005
This was referenced May 21, 2024
atebites-hub
pushed a commit
to atebites-hub/web-llm
that referenced
this pull request
Oct 4, 2025
Prior to this PR, when users `createEngine()` or call `reload()` with a model that is too large for the device, likely the device would keep generating, ignoring OOM issue and correctness. See mlc-ai#356 and mlc-ai#209. This PR catches such error with `device.lost.then()`, depending on tvmjs to call `device.destroy()` upon detecting error in `createBuffer()` via apache/tvm#17005. We have only observed `createBuffer()` errors and hence will only process such kind of errors for now. Besides, since most OOM errors occur in `reload()`, we make the error handling synchronous despite using `.then()` by throwing the error at the end of `reload()` if there is one.
atebites-hub
pushed a commit
to atebites-hub/web-llm
that referenced
this pull request
Oct 4, 2025
### Changes Main changes include: - New model `Hermes-2-Pro-Mistral-7B` in `prebuiltAppConfig` via: - mlc-ai#390 - Various `index.js` and `index.js.map` post-processings to resolve frontend compatibility issues with `require()` and `perf_hoooks` - mlc-ai#397 - mlc-ai#406 - Catch WebGPU OOM error upon `reload()` and `CreateEngine()`: - mlc-ai#402 - Service Worker support (in addition to Extension Service Worker): - mlc-ai#395 - mlc-ai#400 - mlc-ai#401 ### WASM Version v0_2_34 as no change is required. ### TVMjs TVMjs compiled at apache/tvm@a5862a5, with only one change in `tvm/web`: apache/tvm#17005
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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 users
createEngine()or callreload()with a model that is too large for the device, likely the device would keep generating, ignoring OOM issue and correctness. See #356 and #209.This PR catches such error with
device.lost.then(), depending on tvmjs to calldevice.destroy()upon detecting error increateBuffer()via apache/tvm#17005.We have only observed
createBuffer()errors and hence will only process such kind of errors for now. Besides, since most OOM errors occur inreload(), we make the error handling synchronous despite using.then()by throwing the error at the end ofreload()if there is one.