BigDL v0.10.0 Release Notes

Release Date: 2019-11-05 // 11 months ago
  • Highlights

    ๐ŸŽ Continue RNN optimization. We support both LSTM and GRU integration with MKL-DNN which acheives ~3x performance

    ๐Ÿ‘ ONNX support. We support loading third party framework models via ONNX

    ๐Ÿ‘ Richer data preprocssing support and segmentation inference pipeline support

    Details

    • ๐Ÿ‘ [New Feature] Full MaskRCNN model support with data processing
    • ๐Ÿ‘ [New Feature] Support variable-size Resize
    • ๐Ÿ‘ [New Feature] Support batch input for region proposal
    • ๐Ÿ‘ [New Feature] Support samples of different size in one minibatch
    • [New Feature] MAP validation method implementation
    • ๐Ÿ‘ [New Feature] ROILabel enhancement to support both object detection and segmentation
    • ๐Ÿ‘ [New Feature] Grey image support for segmentation
    • ๐Ÿ‘ [New Feature] Add TopBlocks support for Feature Pyramid Networks (FPN)
    • ๐Ÿ‘ [New Feature] GRU integration with MKL-DNN support
    • ๐Ÿ‘ [New Feature] MaskHead support for MaskRCNN
    • ๐Ÿ‘ [New Feature] BoxHead support for MaskRCNN
    • ๐Ÿ‘ [New Feature] RegionalProposal support for MaskRCNN
    • ๐Ÿ‘ [New Feature] Shape operation support for ONNX
    • ๐Ÿ‘ [New Feature] Gemm operation support for ONNX
    • ๐Ÿ‘ [New Feature] Gather operation support for ONNX
    • ๐Ÿ‘ [New Feature] AveragePool operation support for ONNX
    • ๐Ÿ‘ [New Feature] BatchNormalization operation support for ONNX
    • ๐Ÿ‘ [New Feature] Concat operation support for ONNX
    • ๐Ÿ‘ [New Feature] Conv operation support for ONNX
    • ๐Ÿ‘ [New Feature] MaxPool operation support for ONNX
    • ๐Ÿ‘ [New Feature] Reshape operation support for ONNX
    • ๐Ÿ‘ [New Feature] Relu operation support for ONNX
    • ๐Ÿ‘ [New Feature] SoftMax operation support for ONNX
    • ๐Ÿ‘ [New Feature] Sum operation support for ONNX
    • ๐Ÿ‘ [New Feature] Squeeze operation support for ONNX
    • ๐Ÿ‘ [New Feature] Const operation support for ONNX
    • [New Feature] ONNX model loader implementation
    • ๐Ÿ‘ [New Feature] RioAlign layer support
    • [Enhancement] Align batch normalization layer between mklblas and mkl-dnn
    • ๐Ÿ‘ [Enhancement] Python API enhancement to support nested list input
    • ๐Ÿ‘ [Enhancement] Multi-model training/inference support with MKL-DNN
    • [Enhancement] BatchNormalization fusion with Scale
    • ๐Ÿ‘ [Enhancement] SoftMax companion object support no argument initialization
    • ๐Ÿ‘ [Enhancement] Python support for training with MKL-DNN
    • ๐Ÿ“„ [Enhancement] Docs enhancement
    • [Bug Fix] Fix model version comparison
    • [Bug Fix] Fix graph backward bug for ParallelTable
    • [Bug Fix] Fix memory leak for training with MKL-DNN
    • ๐ŸŽ [Bug Fix] Fix performance caused by denormal values during training
    • [Bug Fix] Fix SoftMax segment fault issue under MKL-DNN
    • [Bug Fix] Fix TimeDistributedCriterion python API inconsistent with Scala

Previous changes from v0.9.0

  • Highlights

    โœจ Continue VNNI acceleration support, we add optimization for more CNN models including object detection models, enhance model scales generation support for VNNI.

    โž• Add attention based model support, we add Transformer implementation for both lanuage model and translation model.

    ๐ŸŽ RNN optimization, We support LSTM integration with MKL-DNN which acheives ~3x performance speedup.

    Details

    • ๐Ÿ‘ [New Feature] Add attention layer support
    • ๐Ÿ‘ [New Feature] Add FeedForwardNetwork layer support
    • ๐Ÿ‘ [New Feature] Add ExpandSize layer support
    • ๐Ÿ‘ [New Feature] Add TableOperation layer to support table calculation with different input sizes
    • ๐Ÿ‘ [New Feature] Add LayerNormalizaiton layer support
    • ๐ŸŒ [New Feature] Add Transformer support for both language and translation models
    • ๐Ÿ‘ [New Feature] Add beam search support in Transformer model
    • [New Feature] Add Layer-wise adaptve rate scaling optim method
    • ๐Ÿ‘ [New Feature] Add LSTM integration with MKL-DNN support
    • ๐Ÿ‘ [New Feature] Add dilated convolution integration with MKL-DNN support
    • [New Feature] Add parameter process for LarsSGD optim method
    • ๐Ÿ‘ [New Feature] Support Affinity binding option with mkl-dnn
    • ๐Ÿ— [Enhancement] Document enhancement for configuration and build
    • 0๏ธโƒฃ [Enhancement] Reflection enhancement to get default values for constructor parameters
    • [Enhhancement] User one AllReducemParameter for multi-optim method training
    • ๐Ÿ‘ [Enhancement] CAddTable layer enhancement to support input expansion along specific dimension
    • [Enhancement] Resnet-50 preprocessing pipeline enhancement to replace RandomCropper with CenterCropper
    • [Enhancement] Calculate model scales for arbitrary mask
    • [Enhancment] Enable global average pooling
    • [Enhancement] Check input shape and underlying MKL-DNN layout consistency
    • ๐Ÿ‘ป [Enhancement] Threadpool enhancement to throw proper exception at executor runtime
    • ๐Ÿ‘ [Enhancement] Support mkl-dnn format conversion from ntc to tnc
    • [Bug Fix] Fix backward graph generation topology ordering issue
    • [Bug Fix] Fix MemoryData hash code calculation
    • ๐ŸŒฒ [Bug Fix] Fix log output for BCECriterion
    • [Bug Fix] Fix setting mask for container quantization
    • ๐Ÿ‘ท [Bug Fix] Fix validation accuracy issue when multi-executor running with the same worker
    • [Bug Fix] Fix INT8 layer fusion between conlution with multi-group masks and BatchNormalization
    • [Bug Fix] Fix JoinTable scales generation issue
    • [Bug Fix] Fix CMul forward issue with special input format
    • [Bug Fix] Fix weights change issue after model fusion issue
    • [Bug Fix] Fix SpatinalConvolution primitives initializaiton issue