Description
🐛 Describe the bug
torch.linalg.svd produces an error for some sufficiently large inputs. As the error message says this is most likely a bug in the implementation calling the backend library.
Example (run on CPU):
import torch
A = torch.randn(172032,30000)
U,S,V = torch.linalg.svd(A,full_matrices=False)
yields
Intel MKL ERROR: Parameter 12 was incorrect on entry to SGESDD.
Traceback (most recent call last):
File "<my_path>/lib/python3.8/site-packages/spyder_kernels/py3compat.py", line 356, in compat_exec
exec(code, globals, locals)
File "<my_home_path>/PythonFiles/pytorch_test.py", line 14, in <module>
U,S,V = torch.linalg.svd(A,full_matrices=False)
RuntimeError: false INTERNAL ASSERT FAILED at "../aten/src/ATen/native/LinearAlgebraUtils.h":289, please report a bug to PyTorch. linalg.svd: Argument 12 has illegal value. Most certainly there is a bug in the implementation calling the backend library.
Versions
PyTorch version: 1.12.0+cu102
Is debug build: False
CUDA used to build PyTorch: 10.2
ROCM used to build PyTorch: N/A
OS: Ubuntu 20.04.5 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Clang version: Could not collect
CMake version: version 3.16.3
Libc version: glibc-2.31
Python version: 3.8.10 (default, Nov 14 2022, 12:59:47) [GCC 9.4.0] (64-bit runtime)
Python platform: Linux-5.15.0-56-generic-x86_64-with-glibc2.29
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to:
GPU models and configuration: GPU 0: NVIDIA RTX A6000
Nvidia driver version: 515.86.01
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True
Versions of relevant libraries:
[pip3] mpi4torch==0.1.0
[pip3] mypy-extensions==0.4.3
[pip3] numpy==1.23.0
[pip3] numpydoc==1.4.0
[pip3] torch==1.12.0
[pip3] torchvision==0.13.0
[conda] No relevant packages