jinchen

Math ops lowering and runtime issue

21

### What happened? MLIR is [here](https://github.com/nod-ai/SHARK-TestSuite/blob/main/iree_tests/onnx/node/generated/test_acos/model.mlir). Compilation failed after `iree-convert-to-llvm` pass. > error: failed to legalize operation 'math.acos' that was explicitly marked illegal I added theses [changes](https://github.com/iree-org/iree/commit/87edd2c230b549ffa1d2d6bdbe834bce86775911) and it compiled,...

Number of dims and results of reindexed AffineMap doesn't match on Vectorization

12

### What happened? dispatch: https://gist.github.com/jinchen62/5e2af98f9b5bfc3b55e949f964459815 error log: https://gist.github.com/jinchen62/df2038b5a43ed4680804a3d7d0647d95 The failing op dumped at https://github.com/iree-org/iree/blob/main/compiler/src/iree/compiler/Codegen/Common/GenericVectorization.cpp#L336 is > %10 = linalg.generic {indexing_maps = [affine_map (d0, d1)>, affine_map (0, d0)>], iterator_types = ["parallel",...

Missing dtype support for OnnxToTorch

3

Current support: https://github.com/llvm/torch-mlir/blob/main/lib/Conversion/TorchOnnxToTorch/Utils.cpp#L64 According to [onnx dtype](https://onnx.ai/onnx/intro/concepts.html) and [torch dtype](https://github.com/llvm/torch-mlir/blob/main/include/torch-mlir/Dialect/Torch/Utils/TorchUpstream.h#L88), missing support: - 4: onnx.TensorProto.UINT16 - 8: onnx.TensorProto.STRING - 12: onnx.TensorProto.UINT32 - 13: onnx.TensorProto.UINT64 - 17: onnx.TensorProto.FLOAT8E4M3FN - 18: onnx.TensorProto.FLOAT8E4M3FNUZ...

'util.initializer' op failed to inline into combined initializer

17

### What happened? The further error after https://github.com/iree-org/iree/issues/18232 ``` repro.mlir:107:12: error: 'util.initializer' op failed to inline into combined initializer %414 = affine.apply #map32()[%408, %409, %247] ^ repro.mlir:7:3: note: called from...

### What happened? ``` repro.mlir:72:11: error: 'linalg.conv_2d_nchw_fchw' op inferred input/output operand #1 has shape's dimension #1 to be 4, but found 3 %24 = linalg.conv_2d_nchw_fchw {dilations = dense : vector,...

### What happened? error log: https://gist.github.com/jinchen62/fd0fed08617084a34b704465a5a1abad repro: https://gist.github.com/jinchen62/74ecce88bfa97c37ed76e15bca7cfd52 The repro is a dispatch dumped from model `model--opt-125m-finetuned-squad-assignment--Isente`. ### Steps to reproduce your issue `iree-compile --iree-hal-target-backends=llvm-cpu dispatch_6.mlir -o test.vmfb` ### What...

### What happened? crash log: https://gist.github.com/jinchen62/b5c114d227a0ab4089111be0e4b25468 repro: https://gist.github.com/jinchen62/59af5b1a61bcb78c8ec018af4629258d The repro is a dispatch dumped from model `model--opt-125m-finetuned-squad-assignment--Isente`. ### Steps to reproduce your issue `iree-compile --iree-hal-target-backends=llvm-cpu dispatch_7.mlir -o test.vmfb` ### What...

Support max_unpool2d lowering

10

Support torch->linalg lowering for max_unpool2d. Fixes https://github.com/nod-ai/SHARK-ModelDev/issues/718

### What happened? I'm trying to use transform spec mlir to plug in the topk hip kernel, but got error before `IPOPass`. > error: 'util.call' op operation destroyed but still...