[Good First Issue][NNCF]: [TorchFX] Test PTQ MinMax parameters

[daniil-lyakhov]

Greetings🐱! As a part of #2766 TorchFX PTQ backend support, we are gladly presenting to you following issue:

Implement TemplateTestPTQParams, TemplateTestQuantizerConfig as it done for other backends(example: https://github.com/openvinotoolkit/nncf/blob/develop/tests/torch/ptq/test_ptq_params.py)

Example Pull Requests

#2856

Resources

• Contribution guide - start here!
• Intel DevHub Discord channel - engage in discussions, ask questions and talk to OpenVINO developers
• How to link your Pull Request to an issue

Contact points

@daniil-lyakhov

Ticket

#2766

[rk119]

.take

[github-actions]

Thank you for looking into this issue! Please let us know if you have any questions or require any help.

[daniil-lyakhov]

Hi, @rk119 , do you still plan to contribute?

[rk119]

Hi, @rk119 , do you still plan to contribute?

Yes. I do.

[mlukasze]

@rk119 do you need any help with task? do not hesitate to ask :)

[rk119]

@rk119 do you need any help with task? do not hesitate to ask :)

Hi @mlukasze, I will try to look into this issue over the weekend as I am still eager to continue. If I fail to do so due to lack of availability, then I will unassign myself. I hope it is alright to do so? :)

[mlukasze]

if you want to continue we can wait for you, thanks :)

[rk119]

if you want to continue we can wait for you, thanks :)

Oh alright, thank you so much!

[rk119]

Hi @daniil-lyakhov,

I am facing an issue trying to implement test_unified_scales_command_creation . I am unable to verify and access the quantizer parameters in check_unified_scale_layout, since the create_unified_scales_quantizers_insertion_commands in Torch FX backend returns transformations with only params priority and transformation_fn. I would appreciate some guidance in this. Thank you!

[daniil-lyakhov]

Hi @rk119, please use the __closure__ attribute of the transformation_fn callable object, it will return cells for each captured value. To access cells value please use cell_contents attribute.
Ref: https://docs.python.org/3/reference/datamodel.html#special-read-only-attributes
https://stackoverflow.com/a/55320885/9183892

[rk119]

Hi @rk119, please use the __closure__ attribute of the transformation_fn callable object, it will return cells for each captured value. To access cells value please use cell_contents attribute. Ref: https://docs.python.org/3/reference/datamodel.html#special-read-only-attributes https://stackoverflow.com/a/55320885/9183892

Ah yes! Thank you. :)

[rk119]

Hi @daniil-lyakhov,

I noticed that in the example reference test for this issue, the test_range_estimator_per_channel parameter in the test_params is defined but not used anywhere in the test cases in TemplateTestPTQParams or I could be mistaken. How would you suggest I address this?

[daniil-lyakhov]

Hi @rk119, please just skip this key in the FX tests implementation and keep other tests as it is

[rk119]

Hi @daniil-lyakhov, I opened a PR #2989 for this issue.

[alexsu52]

@rk119, thanks for the contribution!