# \[MPS Crash\] Assertion failed in \_getLSTMGradKernelDAGObject using LSTM on macOS with MPS backend **URL:** \ **Category:** mps\ **Created:** [July 26, 2025, 7:30am UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821 "2025-07-26T07:30:15Z")\ **Posts on this page:** 9\ **Page:** 1
**Author:** ![HarishAi](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/h/df705f/32.png) [@HarishAi](https://discuss.pytorch.org/u/HarishAi)\ **Post date:** [July 26, 2025, 7:30am UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/1 "2025-07-26T07:30:15Z")
Running an Bidirectional LSTM model on macOS with MPS backend crashes with the following error during training/backpropagation: Assertion failed: (shape4.size() \>= 3), function \_getLSTMGradKernelDAGObject, file GPURNNOps.mm, line 2417. zsh: abort python3 translation\_model.py resource\_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown ---
**Author:** ![HarishAi](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/h/df705f/32.png) [@HarishAi](https://discuss.pytorch.org/u/HarishAi)\ **Post date:** [July 27, 2025, 5:22pm UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/3 "2025-07-27T17:22:03Z")
Hello Jack, PyTorch version: Version: 2.7.1 Mac OS version: 15.5 (24F74) I am feeding the input as dimension (Sequence ,Batch, length) example: (N,1,27), I am not using the batch first Model definition: self.input\_size=input\_size self.hidden\_size=hidden\_size self.output\_size=output\_size self.encoder=nn.LSTM(self.input\_size,self.hidden\_size,2,bidirectional=True) self.u=nn.Linear(self.hidden\_size_2,self.hidden\_size) self.w1=nn.Linear(self.hidden\_size,self.hidden\_size) self.w2=nn.Linear(self.hidden\_size,self.hidden\_size) self.w3=nn.Linear(self.hidden\_size,self.hidden\_size) self.w4=nn.Linear(self.hidden\_size,self.hidden\_size) self.w=nn.Linear(self.hidden\_size,self.hidden\_size) self.atten=nn.Linear(self.hidden\_size,1) self.output2catin=nn.Linear(self.output\_size,self.hidden\_size_2) self.decoder=nn.LSTM(self.hidden\_size\*4,self.hidden\_size,4) self.hidden\_output=nn.Linear(self.hidden\_size,self.output\_size) self.softmax=nn.LogSoftmax(dim=2) yes the model works fine on CPU, even in GPU we can inference model, only the training procedure is killed. ---
**Author:** ![HarishAi](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/h/df705f/32.png) [@HarishAi](https://discuss.pytorch.org/u/HarishAi)\ **Post date:** [July 28, 2025, 11:14am UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/5 "2025-07-28T11:14:49Z")
Hello Jack, I am mean model fails while training with the GPU, do you have idea why its failing, if so kindly help with the root cause? Thanks, Harish. ---
**Author:** ![Vinnie](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/v/71e660/32.png) [@Vinnie](https://discuss.pytorch.org/u/Vinnie)\ **Post date:** [July 30, 2025, 8:35am UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/6 "2025-07-30T08:35:29Z")
Hi all, Getting this exact same error when training a RecurrentPPO model on mps, which also uses a LSTM PyTorch model underneath. Running on the M3 Ultra. Just updated to MacOS 15.6 and still the same error. Running Torch version 2.7.1 ---
**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\ **Post date:** [July 30, 2025, 12:23pm UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/7 "2025-07-30T12:23:10Z")
Could you create an issue on GitHub describing this use case in case the most recent nightly binary also fails? ---
**Author:** ![Vinnie](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/v/71e660/32.png) [@Vinnie](https://discuss.pytorch.org/u/Vinnie)\ **Post date:** [July 30, 2025, 1:04pm UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/8 "2025-07-30T13:04:25Z")
@HarishAi , will you raise that issue? ---
**Author:** ![Vinnie](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/v/71e660/32.png) [@Vinnie](https://discuss.pytorch.org/u/Vinnie)\ **Post date:** [August 1, 2025, 12:38pm UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/10 "2025-08-01T12:38:35Z")
Having exact same problem with the latest nightly build (2.9.0.dev20250801). Did you already raise that issue @HarishAi ? ---
**Author:** ![HarishAi](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/h/df705f/32.png) [@HarishAi](https://discuss.pytorch.org/u/HarishAi)\ **Post date:** [August 2, 2025, 12:26pm UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/11 "2025-08-02T12:26:24Z")
@Vinnie , Yes, I have raised it. ---
**Author:** ![Vinnie](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/v/71e660/32.png) [@Vinnie](https://discuss.pytorch.org/u/Vinnie)\ **Post date:** [August 26, 2025, 9:37pm UTC](https://discuss.pytorch.org/t/mps-crash-assertion-failed-in-getlstmgradkerneldagobject-using-lstm-on-macos-with-mps-backend/221821/12 "2025-08-26T21:37:03Z")
I am seeing no activity on the GitHub issue. How are they normally being picked up?