Add future model support for Fouroversix (#44286)
* Add fouroversix integration and docs
* Small fix
* Update quantized modules outside of FourOverSixQuantize
* Delete high-precision parameters during conversion and before weight loading
* Update year
Co-authored-by: Mohamed Mekkouri <93391238+MekkCyber@users.noreply.github.com>
* Replace FourOverSixLinear check with QuantizedModule check
* Run fix-repo
* Update fouroversix
* Update check_docstrings
* Fix keep_master_weights
* Update get_quantized_parameters API
* added fouroversix gpt oss support
* changed high_precision_parameters to parameters_to_quantize
* generalize weight conversions for future model support
* rename config arg and add checks
* clean up
* clean up
* fix parameter deletion
* update comment for weight_conversions
* fix model_config_type logic
* remove model_config_type config argument
* only checking if model_config_type exists instead of None
* Fix param_element_size
* Small update
* Small update
* Fix param_element_size for the last time
* Style fix
---------
Co-authored-by: Jack Cook <hello@jackcook.com>
Co-authored-by: Jack Cook <cookj@mit.edu>
Co-authored-by: Jack Cook <jbcook418@gmail.com>
Co-authored-by: Mohamed Mekkouri <93391238+MekkCyber@users.noreply.github.com>
Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>
