Repository navigation
fix(embedding): resolve ONNX dtype from model declaration or cache - #393
Merged
Merged
Conversation
pipeline() without an explicit dtype resolves device=cpu on Node/Bun, and cpu is missing from transformers.js DEFAULT_DEVICE_DTYPE_MAPPING, so it falls back to fp32 and requests onnx/model.onnx. Quantized-only repos do not ship that file, so the local lookup 404s, the remote fallback runs, and offline hosts fail warmup with 'fetch failed'. Resolve the dtype in two tiers before calling pipeline(): first the repo's own transformers.js_config.dtype declaration, then a probe of the cached ONNX variants in descending precision. When neither applies, dtype is omitted and transformers.js keeps its existing defaults.
Expand the cache probe to the full transformers.js suffix map, skip pipeline dtype overrides when config uses a per-file object map, and cover resolveLocalOnnxDtype with unit tests.
EyJunge1
added a commit
to EyJunge1/opencode-mem
that referenced
this pull request
Oct 10, 2026
Prefer CONFIG.embeddingDtype over resolveLocalOnnxDtype from tickernelz#393 so explicit config still wins after the upstream merge.
2 of 9 tasks
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
pipeline() without an explicit dtype resolves device=cpu on Node/Bun, and cpu is missing from transformers.js DEFAULT_DEVICE_DTYPE_MAPPING, so it falls back to fp32 and requests onnx/model.onnx. Quantized-only repos do not ship that file, so the local lookup 404s, the remote fallback runs, and offline hosts fail warmup with 'fetch failed'.
Resolve the dtype in two tiers before calling pipeline(): first the repo's own transformers.js_config.dtype declaration, then a probe of the cached ONNX variants in descending precision. When neither applies, dtype is omitted and transformers.js keeps its existing defaults.
Summary
Test plan
Checklist
main(do not open PRs from commits already onmain)bun testpasses (or N/A)bun run typecheckpasses (or N/A)bun run checkpasses (format + lint + typecheck; or N/A)