ONNX v1.10.0 is now available with exciting new features! We would like to thank everyone who contributed to this release! Please visit onnx.ai to learn more about ONNX and associated projects.

Key Updates

• Added new Optional and SparseTensor types. #3407 #3398
• Added model local functions to ModelProto. #3532
• Shape inference enhancements for Reshape, Squeeze, NonZero, DynamicQuantizeLinear.
• Introduce symbolic shape inference support. #3506
• New version converter tests. #3344
• Add aarch64 wheel build support. #3414
• Update ONNX IR version to 8 and opset version to 15. #3587

IR Updates

• Added two new types to ONNX type system. Optional and SparseTensor #3407 #3398
• Extend model proto to include model local functions. #3532

Opset version 15

• New Function Operators:
  • Bernoulli #3431
  • CastLike #3558
• New Operators:
  • Optional
  • OptionalGetElement
  • OptionalHasElement
• Operator Updates:
  • Add additional type constraints in BatchNormalization. #3545
  • Add bfloat16 support for Pow. #3412
  • Extend Shape to return a slice using optional attributes start,end. #3580

API

• Symbolic shape inference support. #3506
  • Symbol generation #3518
  • Data propagation #3551 #3593
• Shape inference enhancements
  • Add shape inference for NonZero. #3364
  • Add shape inference for Dynamic QuantizeLinear. #3539
  • Update Reshape shape inference. #3592
  • Fix shape inference for Squeeze. #3516
    • Fix shape inference for Squeeze without axes. #3465
• Expose model parser API in Python (onnx.parser). #3540
• Extend model proto to include model local functions.

Infrastructure

• Update protobuf version to 3.16. #3571
• Add README contents to package description. #3376
• Add requirements.txt to onnx repo. #3448
• Add aarch64 wheel build support. #3414
• Version converter support for recursion into subgraphs. #3474
• Update ONNX examples to python3. #3450

Bug fixes

• Spec clarification for MatMulInteger and QLinearMatMul. #3585
• Extend strict_model for ONNX checker. #3348
• Always set the output of Shape to be rank-1. #3394
• BatchNormalization outputs updated for training mode. #3379
• Bugfix for proto utils and update checker error messages. #3373
• Fix compilation warnings. #3616

Installation

You can upgrade to the latest release using pip install onnx --upgrade or build from source following the README instructions.

Notes

• Beware of the protobuf version gap issue (building onnx with protobuf>=3.12 is not compatible with older protobuf)

Contributors

Thanks to these individuals for their contributions in this release:
@jcwchen, @askhade, @gramalingam, @neginraoof, @matteosal, @postrational, @garymm, @yuslepukhin, @fdwr, @jackwish, @manbearian, @etusien, @impactaky, @rajeevsrao, @prasanthpul, @take-cheeze, @chudegao, @mindest, @yufenglee, @annajung, @hwangdeyu, @calvinmccarter-at-lightmatter, @ashbhandare, @xuzijian629, @IceTDrinker, @mrry