-
Notifications
You must be signed in to change notification settings - Fork 256
[Weight Compression] Rework WC Algorithm to Return WC Params #3636
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Weight Compression] Rework WC Algorithm to Return WC Params #3636
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Aamir, I think it would be beneficial to split this function on two: get_weight_compression_parameters
and get_weight_compression_statistics
as it seem like the statistic collection does not relate to the collection of weight parameters in any way and could be split without any effort. This way it would be clear: statistic points and datasets have nothing to do with the choice of the WC configuration.
Other thing - what a processed
parameters means? What is an unprocessed parameter than?
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
Sure, that sounds good. |
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
src/nncf/quantization/algorithms/weight_compression/algorithm.py
Outdated
Show resolved
Hide resolved
Co-authored-by: Daniil Lyakhov <daniil.lyakhov@intel.com>
Changes
Re-worked the
apply
method in WC algorithm to use an extra method to return weights compression params such that the apply method is more concise and only contains algorithm, quantization logic etc.Reason for changes
This is done so that OpenVINO quantizer can obtain the final collection of weights compression parameters for all the nodes so that are to be compressed.