JonTriebenbach

Single layer Keras LSTM converted to loop instead of ONNX LSTM op

7

The attached example shows a Keras model with a single layer LSTM that is converted to a loop instead of the expected ONNX LSTM op. Test system configuration: Using `tensorflow/tensorflow:2.7.1`...

Trying to compile the MLPerf model 224_224_160_dyanmic_bs.onnx fails with the message: ``` onnx-mlir: /onnx-mlir/src/Dialect/ONNX/ONNXOps.cpp:2137: mlir::LogicalResult mlir::ONNXConvTransposeOp::inferShapes(std::function): Assertion `!outputShape.hasValue() && "unhandled option in ConvTranspose"' failed. ``` The model is here: https://github.com/mlcommons/inference/tree/master/vision/medical_imaging/3d-unet-brats19...

Trying to compile the MLPerf model 224_224_160.onnx fails with the message: ``` error: failed to legalize operation 'onnx.NoValue' ``` The model is here: https://github.com/mlcommons/inference/tree/master/vision/medical_imaging/3d-unet-brats19 https://zenodo.org/record/3928973#.Yu022_HMIYE The compile was with onnx-mlir...

Multiple ONNX optimizer tests fail on big-endian

4

These tests were run on s390x. s390x is big-endian architecture. Example failure log from pytest: ``` _________________________________________________ TestOptimizer.test_fuse_pad_into_conv_1d _________________________________________________ self = def test_fuse_pad_into_conv_1d(self): # type: () -> None pad =...

fuse_bn_into_conv optimization fails on big-endian architecture

1

# Bug Report This test was run on s390x. s390x is big-endian architecture. Failure log from pytest: ``` _______________________________________________ TestOptimizer.test_fuse_bn_into_conv_simple ________________________________________________ self = def test_fuse_bn_into_conv_simple(self): # type: () -> None...