Ramiro Leal-Cavazos

Hi @jxhekang, Please take a look at https://github.com/llvm/torch-mlir/issues/1042, and let me know if that fixes it.

Hi @ZihengJiang, This error means that the `MaximizeValueSemantics` pass was not able to successfully convert the IR graph into one that only uses `!torch.vtensor`s (tensors with value semantics) and does...

> What's the difference between above two return types Hi @Tengxu-Sun, torch-mlir currently does not have support for returning lists. If you need to return several tensors, the only supported...

Hi @KangHe000, Torch-mlir currently does not have support for data-dependent control flow, so the first code snippet failing is expected. The reason you were not getting an error with the...

@silvasean, the main concern here is that for every element in the `offsets` tensor, the entire `output_tensor` gets iterated over. Do you know if there is a different approach using...

This seems related to the error @gpetters94 is also having https://discord.com/channels/636084430946959380/742573221882364009/1009709228950294529. It would be useful to dump the MLIR. It's possible that the issue is that we don't have a...

A correction on the update I gave about the differences between `bert-base-uncased` in Shark vs. the one used by torchbench. Both models do actually use the same vocabulary size, 30522...

All configuration parameters used between the two models are the same: Shark: ``` "attention_probs_dropout_prob": 0.1, "classifier_dropout": null, "gradient_checkpointing": false, "hidden_act": "gelu", "hidden_dropout_prob": 0.1, "hidden_size": 768, "initializer_range": 0.02, "intermediate_size": 3072, "layer_norm_eps":...

Can you split the changes into 3 PRs? They are all quite independent from one another. It would also make the commit titles a lot more descriptive, since each PR...

Is the error not reproducible locally?