webmachinelearning/webnn Issues

Last updated May 18, 2024, 5:58:29 AM UTC.

This repository doesn't have the Priority: Eventually label that's used to mark an issue as triaged without giving it an SLO. Until that's added, this summary uses heuristics to guess if each issue has been triaged.

Untriaged

Try to triage issues within . [ More Info ]

Issue Title Within SLO On maintainers' plates for Time left Time past SLO
#21 (Computer) network-related use cases ❌
#24 Backwards compatibility of the initial set of ops ❌
#93 Add QuantizeLinear and DequantizeLinear for mixed precision ❌
#157 Clarify at which point weights are used in the compilation ❌
#165 Add the first-wave models to the spec appendix ❌
#189 Request for BatchToSpaceND and SpaceToBatchND ops ❌
#228 Support for LRN operation ❌
#252 Support for non-IEEE 754 float point types ❌
#270 Support coordinate transformation modes for Resample2d ❌
#342 Clarify where GPU related states are stored for WebGPU interop (MLContext, MLCommandEncoder and MLGraph) ❌
#346 WebNN and StableHLO opset compatibility ❌
#350 Need to understand how WebNN supports implementation that involves multiple devices and timelines ❌
#374 Simplify `MLPool2dOptions` by removing the `outputSizes` option ❌
#474 Simplify `resample2d` op ❌
#477 API lacks handling for async ML device errors on the context ❌
#496 Bitwise operators and logical operators ❌
#529 Specify WebNN timelines ❌
#537 Vision transformers ❌
#543 [MLBuffer] Uploading/downloading tensor data ❌
#544 [MLBuffer] Support for MLBuffer in graph execution ❌
#559 Control flow operations: if, while ❌
#653 Consider changing output type of ArgMax/Argmin to int32, or allow passing output_type ❌
#669 Rename `MLOperandDescriptor`'s `dimensions` to `shape` ❌
#688 [MLBuffer] Support interop with WebGPU ✔
#689 Consider removing `lstm` and `gru` operators ✔