Automatic update of fbcode/onnx to 9005291283e943f1a91da5f0acf218bc4e8eb2ca (#21057)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/21057
Previous import was cc2333a3f929caca7223b98699237f19388dd585
Included changes:
- **[90052912](https://github.com/onnx/onnx/commit/90052912)**: Fix wrong condition and add --user in update_doc.sh (#2050) <daquexian>
- **[a4f44a20](https://github.com/onnx/onnx/commit/a4f44a20)**: Add bit-shift operators for supporting hashing (#1931) <Wei-Sheng Chin>
- **[0098752c](https://github.com/onnx/onnx/commit/0098752c)**: Add shape inference logic for Expand op (#2041) <Hariharan Seshadri>
- **[fbe8addb](https://github.com/onnx/onnx/commit/fbe8addb)**: update qops tests (#2040) <Ashwini Khade>
- **[874fb37c](https://github.com/onnx/onnx/commit/874fb37c)**: Fix torchvision installation (#2054) <bddppq>
- **[1f5f6582](https://github.com/onnx/onnx/commit/1f5f6582)**: Fix bug that kernel_shape rather than effective_kernel_shape is used in dilated conv (#2043) <daquexian>
- **[38b6c44e](https://github.com/onnx/onnx/commit/38b6c44e)**: Changes done internally at Facebook (#2035) <Lu Fang>
- **[5c51f0db](https://github.com/onnx/onnx/commit/5c51f0db)**: Explicitly specify type of integers in the input tensor. (#2034) <Dmitri Smirnov>
Reviewed By: benoitsteiner
Differential Revision: D15534241
fbshipit-source-id: 8d2b78a986e5b7fbeb248f2d7b80c1a07230654e