Open daniil-lyakhov opened 1 month ago
@rk119
.take
Thank you for looking into this issue! Please let us know if you have any questions or require any help.
Hi @daniil-lyakhov,
I am encountering an issue with the test_depthwise_conv_default_quantizer_config
test case in test_quantizer_config.py. The test is failing because the length of q_setup
is 1 instead of the expected 2.
Upon further investigation, I found that the get_weight_nodes function for Torch FX min_max does not include the PTDepthwiseConv2dSubtype
metatype in its list of recognized metatypes. As a result, the weight nodes variable in _get_quantizer_setup is empty, leading to the failure.
Could you please confirm if the PTDepthwiseConv2dSubtype
metatype should be included in the list? If so, that should resolve the issue. Else, I'd appreciate some guidance.
Hi @daniil-lyakhov,
I am encountering an issue with the
test_depthwise_conv_default_quantizer_config
test case in test_quantizer_config.py. The test is failing because the length ofq_setup
is 1 instead of the expected 2.Upon further investigation, I found that the get_weight_nodes function for Torch FX min_max does not include the
PTDepthwiseConv2dSubtype
metatype in its list of recognized metatypes. As a result, the weight nodes variable in _get_quantizer_setup is empty, leading to the failure.Could you please confirm if the
PTDepthwiseConv2dSubtype
metatype should be included in the list? If so, that should resolve the issue. Else, I'd appreciate some guidance.
Hi @rk119, Thank you for your investigation! You are right, I've forgot to include all metatypes for nodes with weights, good catch! Please replace current weighted metatypes list with this list.
Thanks a lot!
Hi, @rk119 , do you still plan to contribute?
Hi, @rk119 , do you still plan to contribute?
Hi @daniil-lyakhov, yes I do, I'll be opening a PR for this issue by next week. I'm just a little busy with other commitments.
Greetings🐱! As a part of https://github.com/openvinotoolkit/nncf/issues/2766 TorchFX PTQ backend support, we are gladly presenting to you following issue:
Implement TemplateTestQuantizerConfig as it done for other backends(example: https://github.com/openvinotoolkit/nncf/blob/develop/tests/torch/ptq/test_quantizer_config.py)
Example Pull Requests
https://github.com/openvinotoolkit/nncf/pull/2856
Resources
Contact points
@daniil-lyakhov
Ticket
2766