Dev by nroope · Pull Request #30 · cern-nextgen/PQuantML

Commits on Dec 19, 2025

Update README.md
nroope
authored

Commits on Jan 12, 2026

Commits on Feb 22, 2026

Add minor modifications to the core functionality

nastiapetrovych
authored and
nroope
committed
Add minor modifications to the core functionality

nastiapetrovych
authored and
nroope
committed
Add parameter

nastiapetrovych
authored and
nroope
committed
Add loss function modifications

nastiapetrovych
authored and
nroope
committed
Add support for different fixed granularities

nastiapetrovych
authored and
nroope
committed
Add weight and biasses parameters registration and hgq quantization limitation

nastiapetrovych
authored and
nroope
committed
Modified hgq constraint comment

nastiapetrovych
authored and
nroope
committed
Modified default values for pruning_first and use_relu_multiplier

nastiapetrovych
authored and
nroope
committed
Add overflow modes for data and parameters seperately; fixed if self.bias condition error; fix fine-tuning module if mlflow disabled and allow fine-tuning of parameters in quantization/pruning/fitc…

nastiapetrovych
authored and
nroope
committed
Fixed model logging error

nastiapetrovych
authored and
nroope
committed
Add Torch HGQ serialization. Add Keras serialization for dense
nroope
committed
fixed dynamic bit calculation bug, added torch fixed point quantizer
nroope
committed
fix fitcompress, make pdp structured do output channel pruning, rename tuningconfig to PQConfig, finetuning parameters to hpo_parameters, fix layer replacement
nroope
committed
rename rewind options to match paper
nroope
committed
remove extras from configs, fix pdp bug with Keras, add input_shape back to add_compression_layers as an option
nroope
committed
add hpo_params to all configs, make hpo return best_trials in case of multiobjective optimization
nroope
committed

Commits on Feb 23, 2026

add missing add to cuda
nroope
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Dev#30

Dev#30
nroope wants to merge 20 commits intomainfrom
dev

Commits on Dec 19, 2025

Update README.md

Commits on Jan 12, 2026

Add removed property at training model (#22)

Modified 'post_round' function condition (#23)

Commits on Feb 22, 2026

Add minor modifications to the core functionality

Add minor modifications to the core functionality

Add parameter

Add loss function modifications

Add support for different fixed granularities

Add weight and biasses parameters registration and hgq quantization limitation

Modified hgq constraint comment

Modified default values for pruning_first and use_relu_multiplier

Add overflow modes for data and parameters seperately; fixed if self.bias condition error; fix fine-tuning module if mlflow disabled and allow fine-tuning of parameters in quantization/pruning/fitc…

Fixed model logging error

Add Torch HGQ serialization. Add Keras serialization for dense

fixed dynamic bit calculation bug, added torch fixed point quantizer

fix fitcompress, make pdp structured do output channel pruning, rename tuningconfig to PQConfig, finetuning parameters to hpo_parameters, fix layer replacement

rename rewind options to match paper

remove extras from configs, fix pdp bug with Keras, add input_shape back to add_compression_layers as an option

add hpo_params to all configs, make hpo return best_trials in case of multiobjective optimization

Commits on Feb 23, 2026

add missing add to cuda