Scott Roy

Test fix

3

Summary: Depending on torchao target is breaking internal test due to cuda compile error (T222166791): ``` bcode/pytorch/ao/torchao/csrc/cuda/fp6_llm/utils_parallel_dequant.cuh(74): error: identifier "__nv_bfloat16" is undefined 1 error detected in the compilation of "fbcode/pytorch/ao/torchao/csrc/cuda/fp6_llm/fp6_linear.cu"....

Update utils_parallel_dequant.cuh

1

bfloat16 header is needed to compile MultScale with __nv_bfloat16 argument

Add split copy

3

This adds coremltools support for split_copy, the copy variant of split. Many view ops in coremltools already register the copy variant as an alias, e.g., view: https://github.com/apple/coremltools/blob/main/coremltools/converters/mil/frontend/torch/ops.py#L2316 permute: https://github.com/apple/coremltools/blob/main/coremltools/converters/mil/frontend/torch/ops.py#L1011 But...

CoreML segfaults on torch.nn.Conv1d

3

## 🐞Describing the bug CoreML segfaults when running torch.ops.aten.conv1d.default. ## To Reproduce ``` import torch class Model(torch.nn.Module): def __init__(self): super().__init__() self.conv = torch.nn.Conv1d(16, 4, 6, stride=8, padding=0, dilation=2, groups=2, bias=False)...

CoreML BatchNorm3d crashes process

1

## 🐞Describing the bug CoreML BatchNorm3d crashes CoreML process ## To Reproduce ``` import torch class Model(torch.nn.Module): def __init__(self): super().__init__() self.norm = torch.nn.BatchNorm3d(3) def forward(self, x): return self.norm(x) model =...

## 🐞Describing the bug torch.acosh and torch.asinh have registrations in torch that do not lower. ## To Reproduce ``` import torch class Model(torch.nn.Module): def __init__(self): super().__init__() def forward(self, x): return...

CoreML fails at runtime with int32 torch.mm

5

## 🐞Describing the bug torch.mm fails at runtime in CoreML. ## Stack Trace ``` /opt/miniconda3/envs/op-et/lib/python3.10/site-packages/coremltools/models/model.py:560: RuntimeWarning: You will not be able to run predict() on this Core ML model. Underlying...