NotImplementedError: Could not run 'aten::index.Tensor' on MPS #77794

philspence · 2022-05-18T21:49:21Z

🐛 Describe the bug

I see that there are other NotImplementedErrors being report but wanted to add this one to the list too:

import torch

t = torch.tensor([0, 1, 2], device='mps')
t[t == 1]

NotImplementedError: Could not run 'aten::index.Tensor' with arguments from the 'MPS' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'aten::index.Tensor' is only available for these backends: [Dense, Negative, UNKNOWN_TENSOR_TYPE_ID, QuantizedXPU, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, SparseCPU, SparseCUDA, SparseHIP, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, SparseXPU, UNKNOWN_TENSOR_TYPE_ID, SparseVE, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, NestedTensorCUDA, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID, UNKNOWN_TENSOR_TYPE_ID].

Versions

Collecting environment information...
PyTorch version: 1.12.0.dev20220518
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: macOS 12.3.1 (arm64)
GCC version: Could not collect
Clang version: 13.1.6 (clang-1316.0.21.2.5)
CMake version: version 3.22.3
Libc version: N/A

Python version: 3.10.4 | packaged by conda-forge | (main, Mar 24 2022, 17:42:03) [Clang 12.0.1 ] (64-bit runtime)
Python platform: macOS-12.3.1-arm64-arm-64bit
Is CUDA available: False
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] numpy==1.22.3
[pip3] torch==1.12.0.dev20220518
[pip3] torchaudio==0.11.0
[pip3] torchvision==0.12.0
[conda] numpy 1.22.3 py310h99fd38b_2 conda-forge
[conda] torch 1.12.0.dev20220518 pypi_0 pypi
[conda] torchaudio 0.11.0 pypi_0 pypi
[conda] torchvision 0.12.0 pypi_0 pypi

The text was updated successfully, but these errors were encountered:

albanD · 2022-05-18T22:01:28Z

Hi,

Thanks for the report.
Added to the tracking issue to prioritize adding it #77764

alin256 · 2022-08-08T13:37:20Z

I am also missing the 'aten::index.Tensor' feature.

kulinseth · 2022-08-22T19:26:32Z

Fixed in latest nightly.

a19grey · 2022-09-04T15:58:17Z

I ran pip install torch on my M2 system this morning, but still get this error. Despite, thank you, fixing it 13 days ago. Do I need to pull a newer release than whatever comes with "pip install torch"?

Oh! and I'm on python 3.8 so that may be the issue...

Joe-Ecust · 2022-09-06T12:43:18Z

I ran "conda install pytorch torchvision torchaudio -c pytorch" on my M1 Mac today, but still get "NotImplementedError: Could not run 'aten::index.Tensor' with arguments from the 'MPS' backend. " error. Is it fixed yet?

mkagenius · 2022-09-08T12:50:48Z

Fixed in latest nightly.

@kulinseth I think only index.Tensor_out is fixed and not index.Tensor . Correct me if I am wrong.

kulinseth · 2022-09-09T06:56:07Z

@mkagenius tensor_out is indeed fixed . Tensor.tensor_put is ready and I will create a PR soon

Ekberjan · 2022-10-11T18:59:01Z

Hi @kulinseth,
Any progress regarding Tensor.tensor_put?
The original problem still exists even in the latest release.

kulinseth · 2022-10-11T20:08:57Z

Hi @kulinseth,
Any progress regarding Tensor.tensor_put?
The original problem still exists even in the latest release.

@Ekberjan The support for the Op already exists in the Torch Nightly. Can you provide a specific example where you are still seeing this issue?

Ekberjan · 2022-10-12T10:37:24Z

Thanks for your reply @kulinseth.
I'm using the latest release, not the nightly; thought it's included in the latest version.
The below code throws that error (which is from yolov7 official implementation):

    na, nt = self.na, targets.shape[0]  # number of anchors, targets
    indices, anch = [], []
    gain = torch.ones(7, device=targets.device).long()  # normalized to gridspace gain
    ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
    targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices

    g = 0.5  # bias
    off = torch.tensor([[0, 0],
                        [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
                        # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
                        ], device=targets.device).float() * g  # offsets

    for i in range(self.nl):
        anchors = self.anchors[i]
        gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain

        # Match targets to anchors
        t = targets * gain
        if nt:
            # Matches
            r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
            j = torch.max(r, 1. / r).max(2)[0] < self.hyp['anchor_t']  # compare
            # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
            t = t[j]  # filter

Ekberjan · 2022-10-12T18:09:54Z

I tried the nightly as well, and got this error:
NotImplementedError: The operator 'aten::remainder.Tensor_out' is not currently implemented for the MPS device. If you want this op to be added in priority during the prototype phase of this feature, please comment on https://github.com/pytorch/pytorch/issues/77764. As a temporary fix, you can set the environment variable PYTORCH_ENABLE_MPS_FALLBACK=1 to use the CPU as a fallback for this op. WARNING: this will be slower than running natively on MPS.

kulinseth · 2022-10-13T20:07:16Z

I tried the nightly as well, and got this error: NotImplementedError: The operator 'aten::remainder.Tensor_out' is not currently implemented for the MPS device. If you want this op to be added in priority during the prototype phase of this feature, please comment on https://github.com/pytorch/pytorch/issues/77764. As a temporary fix, you can set the environment variable PYTORCH_ENABLE_MPS_FALLBACK=1 to use the CPU as a fallback for this op. WARNING: this will be slower than running natively on MPS.

Thanks @Ekberjan , can you please upvote on the General MPS Op coverage issue page, if you haven;t already?

Also this is a good first issue, if you are interested in contributing to MPS backend.

sandeepwadhwa-ai · 2022-12-25T13:25:12Z

Tried the latest release version and the nightly build. Both give out the same error on M1.

XiangkunFang · 2022-12-26T04:55:27Z

Same here. Have tried both the 1.13.1 and nightly version, but still got "NotImplementedError: The operator 'aten::remainder.Tensor_out' is not currently implemented for the MPS device. If you want this op to be added in priority during the prototype phase of this feature, please comment on #77764. As a temporary fix, you can set the environment variable PYTORCH_ENABLE_MPS_FALLBACK=1 to use the CPU as a fallback for this op. WARNING: this will be slower than running natively on MPS“

leedrake5 · 2023-01-17T21:12:39Z

I've got a truly bizarre variant of this bug - I can train a model just fine, but the NotImplementedError: The operator 'aten::roll' is not currently implemented for the MPS device. shows up when I try to predict a model trained on the same Mac. Is there some reason prediction would fail when training would succeed? Doesn't a model have to predict the validation set while training?

GF-Huang · 2025-01-14T12:55:03Z

Pytorch 2.5.1 issue still.

GPU: RTX 4060ti 8G

> conn_matrix
tensor([[0., 0., 0.,  ..., 1., 0., 1.],
        [1., 1., 0.,  ..., 1., 1., 0.],
        [0., 0., 0.,  ..., 0., 0., 0.],
        ...,
        [0., 1., 0.,  ..., 0., 0., 1.],
        [1., 1., 1.,  ..., 0., 0., 0.],
        [1., 0., 1.,  ..., 1., 1., 1.]], device='cuda:0')

> indexes
tensor([True, True, True,  ..., True, True, True], device='cuda:0')

> coo_conn_matrix = conn_matrix.to_sparse_coo()
> coo_conn_matrix[indexes]
---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
Cell In[7], [line 2](vscode-notebook-cell:?execution_count=7&line=2)
      [1](vscode-notebook-cell:?execution_count=7&line=1) coo_conn_matrix = conn_matrix.to_sparse_coo()
----> [2](vscode-notebook-cell:?execution_count=7&line=2) coo_conn_matrix[indexes]

File c:\Users\jokeh\miniconda3\envs\mlagents\lib\site-packages\torch\utils\_device.py:106, in DeviceContext.__torch_function__(self, func, types, args, kwargs)
    [104](file:///C:/Users/jokeh/miniconda3/envs/mlagents/lib/site-packages/torch/utils/_device.py:104) if func in _device_constructors() and kwargs.get('device') is None:
    [105](file:///C:/Users/jokeh/miniconda3/envs/mlagents/lib/site-packages/torch/utils/_device.py:105)     kwargs['device'] = self.device
--> [106](file:///C:/Users/jokeh/miniconda3/envs/mlagents/lib/site-packages/torch/utils/_device.py:106) return func(*args, **kwargs)

NotImplementedError: Could not run 'aten::index.Tensor' with arguments from the 'SparseCUDA' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'aten::index.Tensor' is only available for these backends: [CPU, CUDA, HIP, MPS, IPU, XPU, HPU, VE, MTIA, PrivateUse1, PrivateUse2, PrivateUse3, Meta, FPGA, MAIA, Vulkan, Metal, QuantizedCPU, QuantizedCUDA, QuantizedHIP, QuantizedMPS, QuantizedIPU, QuantizedXPU, QuantizedHPU, QuantizedVE, QuantizedMTIA, QuantizedPrivateUse1, QuantizedPrivateUse2, QuantizedPrivateUse3, QuantizedMeta, CustomRNGKeyId, MkldnnCPU, SparseCsrCPU, SparseCsrCUDA, SparseCsrHIP, SparseCsrMPS, SparseCsrIPU, SparseCsrXPU, SparseCsrHPU, SparseCsrVE, SparseCsrMTIA, SparseCsrPrivateUse1, SparseCsrPrivateUse2, SparseCsrPrivateUse3, SparseCsrMeta, BackendSelect, Python, FuncTorchDynamicLayerBackMode, Functionalize, Named, Conjugate, Negative, ZeroTensor, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradHIP, AutogradXLA, AutogradMPS, AutogradIPU, AutogradXPU, AutogradHPU, AutogradVE, AutogradLazy, AutogradMTIA, AutogradPrivateUse1, AutogradPrivateUse2, AutogradPrivateUse3, AutogradMeta, AutogradNestedTensor, Tracer, AutocastCPU, AutocastXPU, AutocastMPS, AutocastCUDA, FuncTorchBatched, BatchedNestedTensor, FuncTorchVmapMode, Batched, VmapMode, FuncTorchGradWrapper, PythonTLSSnapshot, FuncTorchDynamicLayerFrontMode, PreDispatch, PythonDispatcher].

Undefined: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
CPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCPU.cpp:30476 [kernel]
CUDA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCUDA.cpp:44679 [kernel]
HIP: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
MPS: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
IPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
XPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
HPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
VE: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
MTIA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
PrivateUse1: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
PrivateUse2: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
PrivateUse3: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
Meta: registered at /dev/null:241 [kernel]
FPGA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
MAIA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
Vulkan: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
Metal: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedCPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterQuantizedCPU.cpp:954 [kernel]
QuantizedCUDA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedHIP: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedMPS: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedIPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedXPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedHPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedVE: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedMTIA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedPrivateUse1: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedPrivateUse2: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedPrivateUse3: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
QuantizedMeta: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
CustomRNGKeyId: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
MkldnnCPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrCPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrCUDA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrHIP: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrMPS: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrIPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrXPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrHPU: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrVE: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrMTIA: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrPrivateUse1: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrPrivateUse2: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrPrivateUse3: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
SparseCsrMeta: registered at C:\cb\pytorch_1000000000000\work\build\aten\src\ATen\RegisterCompositeExplicitAutogradNonFunctional.cpp:21615 [default backend kernel]
BackendSelect: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\BackendSelectFallbackKernel.cpp:3 [backend fallback]
Python: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\PythonFallbackKernel.cpp:153 [backend fallback]
FuncTorchDynamicLayerBackMode: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\functorch\DynamicLayer.cpp:497 [backend fallback]
Functionalize: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\FunctionalizeFallbackKernel.cpp:349 [backend fallback]
Named: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\NamedRegistrations.cpp:7 [backend fallback]
Conjugate: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\ConjugateFallback.cpp:17 [backend fallback]
Negative: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\native\NegateFallback.cpp:18 [backend fallback]
ZeroTensor: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\ZeroTensorFallback.cpp:86 [backend fallback]
ADInplaceOrView: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\VariableFallbackKernel.cpp:96 [backend fallback]
AutogradOther: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradCPU: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradCUDA: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradHIP: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradXLA: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradMPS: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradIPU: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradXPU: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradHPU: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradVE: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradLazy: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradMTIA: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradPrivateUse1: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradPrivateUse2: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradPrivateUse3: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradMeta: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
AutogradNestedTensor: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\VariableType_1.cpp:16894 [autograd kernel]
Tracer: registered at C:\cb\pytorch_1000000000000\work\torch\csrc\autograd\generated\TraceType_1.cpp:16106 [kernel]
AutocastCPU: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\autocast_mode.cpp:321 [backend fallback]
AutocastXPU: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\autocast_mode.cpp:463 [backend fallback]
AutocastMPS: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\autocast_mode.cpp:209 [backend fallback]
AutocastCUDA: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\autocast_mode.cpp:165 [backend fallback]
FuncTorchBatched: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\functorch\BatchRulesScatterOps.cpp:1230 [kernel]
BatchedNestedTensor: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\functorch\LegacyBatchingRegistrations.cpp:758 [backend fallback]
FuncTorchVmapMode: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\functorch\VmapModeRegistrations.cpp:27 [backend fallback]
Batched: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\LegacyBatchingRegistrations.cpp:1075 [backend fallback]
VmapMode: fallthrough registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\VmapModeRegistrations.cpp:33 [backend fallback]
FuncTorchGradWrapper: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\functorch\TensorWrapper.cpp:207 [backend fallback]
PythonTLSSnapshot: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\PythonFallbackKernel.cpp:161 [backend fallback]
FuncTorchDynamicLayerFrontMode: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\functorch\DynamicLayer.cpp:493 [backend fallback]
PreDispatch: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\PythonFallbackKernel.cpp:165 [backend fallback]
PythonDispatcher: registered at C:\cb\pytorch_1000000000000\work\aten\src\ATen\core\PythonFallbackKernel.cpp:157 [backend fallback]

TaylorHo · 2025-05-04T04:38:15Z

I'm still getting this error on a macOS M3 device.

Output says:

NotImplementedError: Could not run 'aten::_sparse_coo_tensor_with_dims_and_tensors' with arguments from the 'SparseMPS' backend.

Can we reopen this issue, since more people are still having the same problem running it on MPS?

albanD added feature A request for a proper, new feature. triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module module: mps Related to Apple Metal Performance Shaders framework labels May 18, 2022

DenisVieriu97 mentioned this issue Jul 8, 2022

Add suppport for dispatch stub; Register index.Tensor_out kulinseth/pytorch#53

Merged

kulinseth closed this as completed Aug 22, 2022

stared mentioned this issue Mar 13, 2023

CPU and MPS (Apple Silicon) backends gligen/GLIGEN#14

Merged

4 tasks

ringohoffman mentioned this issue Apr 25, 2025

NotImplementedError: Could not run 'aten::index.Tensor' with arguments from the 'SparseCUDA' backend. #152226

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

NotImplementedError: Could not run 'aten::index.Tensor' on MPS #77794

NotImplementedError: Could not run 'aten::index.Tensor' on MPS #77794

philspence commented May 18, 2022

albanD commented May 18, 2022

alin256 commented Aug 8, 2022

kulinseth commented Aug 22, 2022

a19grey commented Sep 4, 2022 •

edited

Loading

Joe-Ecust commented Sep 6, 2022

mkagenius commented Sep 8, 2022

kulinseth commented Sep 9, 2022

Ekberjan commented Oct 11, 2022 •

edited

Loading

kulinseth commented Oct 11, 2022 •

edited

Loading

Ekberjan commented Oct 12, 2022 •

edited

Loading

Ekberjan commented Oct 12, 2022 •

edited

Loading

kulinseth commented Oct 13, 2022

sandeepwadhwa-ai commented Dec 25, 2022

XiangkunFang commented Dec 26, 2022

leedrake5 commented Jan 17, 2023

GF-Huang commented Jan 14, 2025

TaylorHo commented May 4, 2025

NotImplementedError: Could not run 'aten::index.Tensor' on MPS #77794

NotImplementedError: Could not run 'aten::index.Tensor' on MPS #77794

Comments

philspence commented May 18, 2022

🐛 Describe the bug

Versions

albanD commented May 18, 2022

alin256 commented Aug 8, 2022

kulinseth commented Aug 22, 2022

a19grey commented Sep 4, 2022 • edited Loading

Joe-Ecust commented Sep 6, 2022

mkagenius commented Sep 8, 2022

kulinseth commented Sep 9, 2022

Ekberjan commented Oct 11, 2022 • edited Loading

kulinseth commented Oct 11, 2022 • edited Loading

Ekberjan commented Oct 12, 2022 • edited Loading

Ekberjan commented Oct 12, 2022 • edited Loading

kulinseth commented Oct 13, 2022

sandeepwadhwa-ai commented Dec 25, 2022

XiangkunFang commented Dec 26, 2022

leedrake5 commented Jan 17, 2023

GF-Huang commented Jan 14, 2025

TaylorHo commented May 4, 2025

a19grey commented Sep 4, 2022 •

edited

Loading

Ekberjan commented Oct 11, 2022 •

edited

Loading

kulinseth commented Oct 11, 2022 •

edited

Loading

Ekberjan commented Oct 12, 2022 •

edited

Loading

Ekberjan commented Oct 12, 2022 •

edited

Loading