[FX][Conformance] Enable Conformance Test for FX Backend

[anzr299]

Changes

1. Save torch exported compressed model instead of OV coverted model. Save OV model which is obtained using torch.compile and dumping cache. The OV model obtained by dumping cache is used for counting FQ nodes.
2. Save torch exported model using torch.export.save instead of converting to OV and saving OV model when saving fp32 model for debugging.
3. Make torch compile validation default path for torch fx models.

Above changes were made to obtain the validation path:

Related tickets

163422

Tests

Category	Job	Status	Job Number	Notes
Manual	post_training_quantization	Failed	627	Expected Regression of 0.6% in Swin with FX backend
GH WC	WC test on GHA	Pass	36
Example	Example	Pass	240

[daniil-lyakhov]

Please do not remove the self.torch_compile_validation option and made it True by default

[anzr299]

but then the default path for FX backend models will be using OV validation

[daniil-lyakhov]

Typo, my bad, I meant True by default

[daniil-lyakhov]

What if there are several submodels in this dir?

[anzr299]

submodels of the same model? or other models?
I can save in a different folder by the model name if the problem is the latter like this:
torch.compile(exported_model.module(), backend="openvino", options = {"model_caching" : True, "cache_dir": str(self.output_model_dir / self.model_name)})
instead of just "cache_dir": str(self.output_model_dir)

[daniil-lyakhov]

I mean one model is being cut on several part, as it was with the Yolo 11 model. As far as I remember, this means there are several IRs generated for one model which are run sequentially

[anzr299]

hm, but the models with graph break should not be supported right?

[daniil-lyakhov]

They should have one graph, but it is possible due to bugs in ov/nncf that there are several IRs. I wonder, what will be the result? Perhaps we shouldn't analyze the parts of the model separately

[anzr299]

Then Maybe I can raise an error after a check for multiple bin and xml files in the location. The expected behavior would be to simple rename and replace the files.

[alexsu52]

Could you update the description with more details? I mean, it's not entirely clear what exactly was done.

[daniil-lyakhov]

@anzr299, the issue 162009 was fixed in the openvino 2025.1, please consider it