[Lane 2 · kotlin-core] ONNX importer — import ONNX ConvTranspose as a real op #1406
Copy link
Copy link
Open
Labels
codingImplementation task (DARC: C)Implementation task (DARC: C)size:sA few hoursA few hoursskill:kotlin-coreKotlin implementation in commonMainKotlin implementation in commonMainsub-issueSub-issue of a tracking issueSub-issue of a tracking issue
Description
Sub-issue of #1400 ([Feature]: ONNX importer — replace placeholder ops with real operations, stream tensors in OnnxModelReader).
Lane: 2 · Kotlin core
Skill needed: Kotlin. No ONNX-protobuf knowledge needed — the node is already parsed; you only map it.
Size: s (a few hours)
Blocked by: nothing
What to do
- In
skainet-io/skainet-io-onnx/src/commonMain/kotlin/sk/ainet/io/onnx/OnnxImport.kt(toOperation()),ConvTransposecurrently maps toOnnxPlaceholderOperation, which throws when the graph is executed. Map it to the convTranspose1d operation (and convTranspose2d once the TransposedConv2d feature lands) instead, copying howRELU/CONVmap to real operations. - Read the ONNX operator spec for its attributes and defaults: https://onnx.ai/onnx/operators/
- Add a one-node import + execute test next to the existing importer tests in
skainet-io/skainet-io-onnx/src/commonTest. - Run
./gradlew :skainet-io:skainet-io-onnx:jvmTest.
Acceptance
- A one-node
ConvTransposegraph imports and executes with the expected values - Result reported back on the parent issue
Notes
If something here turns out to be wrong or much bigger than the size label, comment on this issue or open a specific bug instead of blocking the parent.
Coverage matrix: docs/modules/ROOT/pages/reference/ops-status-matrix.adoc · How lanes work: Issue taxonomy
Metadata
Metadata
Assignees
Labels
codingImplementation task (DARC: C)Implementation task (DARC: C)size:sA few hoursA few hoursskill:kotlin-coreKotlin implementation in commonMainKotlin implementation in commonMainsub-issueSub-issue of a tracking issueSub-issue of a tracking issue
Activity