The basic workflow is loading a JSON configuration script containing NNCF-specific parameters determining the compression to be applied to your model, and then passing your model along with the configuration script to the create_compressed_model function. The NNCF is organized as a regular Python package that can be imported in your target training pipeline script. Support for Accuracy-Aware model training pipelines via the Adaptive Compression Level Training and Early Exit Training. Git patches for prominent third-party repositories ( huggingface-transformers) demonstrating the process of integrating NNCF into custom training pipelinesĮxporting PyTorch compressed models to ONNX* checkpoints and TensorFlow compressed models to SavedModel or Frozen Graph format, ready to use with OpenVINO™ toolkit. GPU-accelerated layers for faster compressed model fine-tuning.Ĭonfiguration file examples for each supported compression algorithm. The models created using Sequential or Keras Functional API are only supported.Ĭommon interface for compression methods. NOTE: Limited support for TensorFlow models. Support of various compression algorithms, applied during a model fine-tuning process to achieve a better performance-accuracy trade-off: Compression algorithmĪutomatic, configurable model graph transformation to obtain the compressed model. The frameworkĪrchitecture is unified to make it easy to add different compression algorithms for both PyTorch and TensorFlow deep The framework is organized as a Python* package that can be built and used in a standalone mode. TensorFlow models and datasets: Image Classification, Object Detection and Semantic Segmentation.Ĭompression results achievable with the NNCF-powered samples can be found in a table at NNCF provides samples that demonstrate the usage of compression algorithms for three different use cases on public PyTorch and NNCF is designed to work with models from PyTorch and TensorFlow. NNCF provides a suite of advanced algorithms for Neural Networks inference optimization in OpenVINO™ with minimal accuracy drop. Neural Network Compression Framework (NNCF) Upgrade TensorFlow from 2.4.2 to 2.4.3 (#974)Īdd release notes for release v2.0 (#846) Update configuration mgmt owner files (#896) Introduce load_state and get_state methods for compression controller + for compression loss + put compression ctrl state into checkpoint (#679) Moved examples to examples/torch dir and tests to tests/torch dir (#725) Update gitattributes for *.jpg files (#942) Third_party_integration/huggingface_transformersĪdjust README.md for transformers patch (#943) Knowledge Distillation scale and temperature features (#987) The docker file was added for NNCF TF (#837)Ĭhange the default pruning schedule to exponential (#961) Test cases for test_knowledge_distillation_loss_types were added Temperature and scale features were added to KD algorithmĮfficient distillation of some models require logits softening (temperature) and loss scaling Knowledge Distillation scale and temperature features ( #987 ) # Changes
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |