Sean Silva
A gotcha I recently learned is that Value::getUsers() doesn't dedupe the users if there are multiple uses by one operation. There should be an easy way to get this functionality...
(ignore the "Rework how global slot initializers work." commit -- this one doesn't trivially commute with it, but is still logically separate) This introduces a new pass LowerToBackendContract (better name...
We try to avoid using #ifdefs because they make it impossible to test all the configurations easily. I noticed two in TorchToMhlo that can be removed. TORCH_MLIR_ENABLE_MHLO_STATIC_SHAPE TORCH_MLIR_ENABLE_MHLO_TRUNC_DIMSIZE_TO_I32 https://github.com/llvm/torch-mlir/blob/9d6ee48661cefa677950bcd262fe2824a4fa9d17/lib/Conversion/TorchToMhlo/Basic.cpp#L412 We...
At the next Community Meeting we will be asking folks their thoughts on project health. I am creating this issue as a place for folks that watch the recording to...
https://github.com/llvm/torch-mlir/blob/06750815d14371a7b8f0e964f2380517583e745f/lib/Conversion/TorchToLinalg/Linear.cpp#L242 Right now we implement this with an open-coded linalg.generic, but backends want to see this as a batch matmul named op. We should implement this with a collapsing reshape,...
### Description of the problem / feature request: Bazel doesn't support cc_embed_data ### Feature requests: what underlying problem are you trying to solve with this feature? I would like to...
For months, we have been lucky to have @ashay taking care of the LLVM update, but we need to start sharing the load here as a community. In this issue,...
Hey folks, custom ops has been a common feature request, and I'd like to share a design that we've been thinking about. Let's talk first about the functionality we want...
In IREE-Torch, there is one test that segfaults recently. The segfault is on memory access at address 0 before it falls off into ud2's in `printType`. Any ideas about build...
This lowers them into ml_program.global Example use like this: ``` import torch import torch_mlir import torchvision resnet18 = torchvision.models.resnet18(pretrained=True) resnet18.eval() example_inputs = torch.randn(1, 2) mlir_module = torch_mlir.compile( resnet18, torch.ones(1, 3,...