[CI] Skip onnx_ops hip rdna3 tests

[amd-eochoalo]

These tests have been failing since #22813 . An issue has been filed in #23189

[amd-eochoalo]

Relevant tests are passing. Merging.

[MaheshRavishankar]

Just to clarify, all of these have some sort of control flow in the generated IR on input?

[amd-eochoalo]

I can't answer that at the moment since I am building IREE for the integrate. I'll double check once I have a few cycles and revert the necessary changes. I understand now what you meant when you said "disable only [the tests without any dispatches]"

[MaheshRavishankar]

I can't answer that at the moment since I am building IREE for the integrate. I'll double check once I have a few cycles and revert the necessary changes. I understand now what you meant when you said "disable only [the tests without any dispatches]"

Lets take these one by one... we can have a quick sync where we can just check what the IR looks like after global opt phase and I can let you know what to do with these.

[amd-eochoalo]

@MaheshRavishankar sounds good, let me revert and tomorrow I can look into this in more detail.

[amd-eochoalo]

@MaheshRavishankar I've reverted them here #23199

[MaheshRavishankar]

Its OK, if you are going to look at it soon, no need to revert.. you can just enable them as we look at it.

[amd-eochoalo]

Here is the list of reasons why tests are failing

test	reason
test_gru_seq_length	hangs
test_rnn_seq_length	hangs
test_scan9_sum	hangs
test_stft_with_window	hangs
test_tfidfvectorizer_tf_batch_onlybigrams_skip0	Does not generate any dispatches
test_tfidfvectorizer_tf_batch_onlybigrams_skip5	Does not generate any dispatches
test_tfidfvectorizer_tf_batch_uniandbigrams_skip5	Does not generate any dispatches

Based on this I'll look a little bit closer into the first four, keep the last three as skips.

[MaheshRavishankar]

Here is the list of reasons why tests are failing

test reason
test_gru_seq_length hangs
test_rnn_seq_length hangs
test_scan9_sum hangs
test_stft_with_window hangs
test_tfidfvectorizer_tf_batch_onlybigrams_skip0 Does not generate any dispatches
test_tfidfvectorizer_tf_batch_onlybigrams_skip5 Does not generate any dispatches
test_tfidfvectorizer_tf_batch_uniandbigrams_skip5 Does not generate any dispatches
Based on this I'll look a little bit closer into the first four, keep the last three as skips.

If you post the input after global optimizations somewhere I can help make a decision on this.

[amd-eochoalo]

I'll have this later today. Thanks!