alankelly

Thanks for finding and fixing this. Let me know when it's ready for review

This looks like a cpuinfo issue to me. Please ask on https://github.com/pytorch/cpuinfo Could also be something to do with cross compilation.

@Tessil The newly added tests in batch_to_space_nd.py are failing: { "dtype": [tf.float32], "input_shape": [[1, 3, 3, 1]], "block_shape": [[1, 1]], "crops": [[[0, 0], [0, 0]], [[1, 1], [1, 1]]], "constant_block_shape":...

Thanks for the fix! This is now merged. https://github.com/tensorflow/tensorflow/commit/21c982484feed43a5368bad4cac274cc4323769c

Please fix tensorflow/lite/kernels:transpose_conv_test: ```W0802 00:49:23.534827 2118086 thread.cc:2019] --- Thread 7f1003e31d00 (name: main/2118086) stack: --- W0802 00:49:23.544801 2118086 thread.cc:2019] @ 0x7f100470fd0a ./third_party/tensorflow/lite/kernels/test_util.h:93 tflite::Quantize() W0802 00:49:23.545504 2118086 thread.cc:2019] @ 0x7f0f2f81ed68 base/process_state.cc:1281 FailureSignalHandler()...

Hi Saoirse, The issue can be recreated using sanitizers: `bazel run -c dbg --config=asan tensorflow/lite/kernels:transpose_conv_test` I'm not sure that whether you use bazel or not or whether sanitizers will work...

I have given this a performance review, re-assigning this to @abattery for a quantization review

Cab you please rebase this and I will land first thing on Monday?

Thanks, importing now.

Our bot which handles inporting/exporting PRs to github was temporarily broken and I think this PR got mixed up somehow with this. I have added created a new PR with...