Use gesvd as solver

[andompesta]

What does this PR do?

Type of change: ?

Bug fix: default svd solver does not work for ill-conditioned matrixes. Switch to gesvd solver as it is more accurate (as per documentation) as does not fails for ill-conditioned matrixes

Overview: ?

Usage

# Add a code snippet demonstrating how to use this

Testing

Before your PR is "Ready for review"

• Make sure you read and follow Contributor guidelines and your commits are signed.
• Is this change backward compatible?: Yes
• Did you write any new necessary tests?: No. as we might not want to test all matrix shapes
• Did you add or update any necessary documentation?: No need to change the documentation
• Did you update Changelog?: No

Additional Information

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[jingyu-ml]

LGTM. As a reference, could you point me to the documentation?

[kevalmorabia97]

/ok to test c513c47

[kevalmorabia97]

@andompesta seems like our SVD Quant tests fail on CPU: https://github.com/NVIDIA/Model-Optimizer/actions/runs/20025754823/job/58410748645?pr=661

FAILED tests/unit/torch/quantization/test_calib.py::test_svdquant_lora_weights
FAILED tests/unit/torch/quantization/test_quantize_cpu.py::test_quantize[config4-SimpleLinear]
FAILED tests/unit/torch/quantization/test_quantize_cpu.py::test_quantize[config4-SimpleConvLinear]
FAILED tests/unit/torch/quantization/test_quantize_cpu.py::test_save_restore[SimpleLinear-quant_config3]

RuntimeError: torch.linalg.svd: keyword argument `driver=` is only supported on CUDA inputs with cuSOLVER backend.

[andompesta]

I see, I gived it for granted that svd runs on cuda devices

[kevalmorabia97]

Thanks. One more thing:

Commit sha: 84fc8fb, Author: scavallari, Committer: scavallari; The sign-off is missing.

Please use git commit -s -S for signing-off. More details in https://github.com/NVIDIA/TensorRT-Model-Optimizer/blob/main/CONTRIBUTING.md#%EF%B8%8F-signing-your-work

[kevalmorabia97]

Still failing