ONE Release 1.10.0

Release Note 1.10.0

ONE Compiler

Compiler supports more operations

• Dequantize, UnidirectionalSequenceLSTM

Changes

• New --fold_dequantize option in one-optimize
• New --fuse_add_with_tconv option in one-optimize
• Support int16 quantization in one-quantize
• Test scripts are added for basic testing of one-cmds command line tools
• Bug fixes for one-cmds command line tools

ONE Runtime

Runtime backend operation support

• ACL-CL backend: OneHot
• CPU backend: FullyConnected for Float32 16x1 Block Sparsity

Optimization

• Speed up for ReduceSum, StrideSlice and BinaryArithmetic in CPU backend

ONE Release 1.9.1

Release Note 1.9.1

ONE Compiler

Changes

• tf2nnpkg now supports to import TensorFlow model which includes BCQ information.
• Minor change for preserving BCQ information.
• Fix invalid input arguments and add more error handles for one-cmds

ONE Release 1.9.0

Release Note 1.9.0

ONE Compiler

Compiler supports more operations

• NonMaxSuppressionV4, NonMaxSuppressionV5, PadV2, Unique

Changes

• Quantization enhancements: channel wise UINT8 quantization(Conv2D, DepwiseConv, TransposeConv, FullyConnected)
• Experimental requantization from INT8 to UINT8
• Adding more operator value tests
• tf2tfliteV2 supports conversion from Keras model, saved model
• Refactoring for better maintenance long Class codes using visitor patterns
• Introducing optimization pass that fuses batch normalization with Transposed Convolution.

ONE Runtime

Runtime backend operation support

• CPU backend: RANK
• CPU backend qasymm uint8: LOG_SOFTMAX
• ACL-CL backend: LEAKY_RELU, RESIZE_NEAREST_NEIGHBOR

Optimization

• Copy Elimination between compatible backends

Operation Implementation

• Operations with same parameters are unified

Change

• CPU backend qasymm uint8 performance enhancement: arithmetic operations

ONE Release 1.8.0

Release Note 1.8.0

Feature Highlights

• ONE Compiler

  • Support new command line interface

• ONE Runtime

  • CPU backend supports 7 more operations
  • CPU backend supports 9 more quant8 operations

ONE Compiler

New command line interface for user interface consistancy

• one-import-bcq : import BCQ(Binary coding quantized) TensorFlow model
• Commands now support --version option to show version number

Changes

• Experimental support for TensorFlow 2.x has updated to 2.3.0 (TensorFlow 1.3.2 is our official support version)
• Support more operators in luci-interpreter
• Enhancing one-quantizer

ONE Runtime

Rename headers

• Rename nnfw_dev.h to nnfw_experimental.h

Optimization

• Remove copies for model input/outputs whenever possible

Support CPU backend operation

• BatchToSpaceND, L2Normalization, ReLU6, ResizeBilinear, SpaceToDepth, SplitV, StatelessRandomUniform

Support CPU backend quant8 operation

• BatchToSpaceND, L2Normalization, Pad, PadV2, ResizeBilinear, Slice, Quantize, SpaceToDepth, Sum

ONE Release 1.7.0

ONE Release Note 1.7.0

Feature Highlights

• ONE Compiler
  • Compiler supports more operations
  • New command line interface for user interface consistancy
• ONE Runtime
  • Runtime CPU backend supports more operations
  • Runtime CPU backend supports more quant8 operations
  • API changes
  • New optimization

ONE Compiler

Compiler supports more operations

• MatrixDiag, MatrixSetDiag, ReverseSequence, ReverseV2, SegmentSum, SelectV2, SparseToDense, Where

New command line interface for user interface consistancy

• one-import: imports conventional model files to circle
  • one-import-tf: imports TensorFlow model to circle
  • one-import-tflite: imports TensorFlow lite model to circle
• one-optimize: circle optimize command
• one-quantize: circle quantize command
  • supports float32 to uint8, layer wise (for Conv series)
• one-pack: package command
• one-prepare-venv: prepares python virtual environment for importing TensorFlow model
• one-codegen: backend(if available) code generator

ONE Runtime

Runtime CPU backend supports more operations

• LogSoftmax, SpaceToBatchND

Runtime CPU backend supports more quant8 operations

• Logistic, Mul, Tanh, SpaceToBatchND, Transpose, Sub, Max, Min, Less, Greater, GreaterEqual, LessEqual, Equal, NotEqual

API changes

• Introduce basic asynchronous execution API

New optimization

• Remove dynamic tensor overhead from static models

ONE Release 1.6.1

Release Note 1.6.1

ONE Compiler

Bug patch

• Update tf2nnpkg to latest preset; 1.6.0 contained old version.
  • 1.6.0 tf2nnpkg called old compiler binaries that are not packaged.

ONE 1.6.0 Release

Release Note 1.6.0

Feature Highlights

• ONE Compiler

  • Compiler supports 22 more operations including 1 custom operation.

• ONE Runtime

  • CPU backend supports 8 more operations
  • Support dynamically shaped tensors
  • Support Control Flow operations
  • API updates

• Add ONE Runtime package for x86_64 linux (Experimental)

ONE Compiler

Compiler supports more operations

• AddN, ArgMin, Custom(BatchMatmul), Ceil, DepthToSpace, Floor, InstanceNormalize,
  L2Normalization, L2Pool, LessEqual, Log, LogSoftmax, PReLU, Rank, ReduceMin(Min),
  ResizeBilinear, Round, ScatterND, Sqrt, TransposeConv, BCQGather,
  BCQFullyConnected

ONE Runtime

CPU backend supports more operations

• BatchMatMul, BroadcastTo, Einsum, FusedBatchNorm, MatrixBandPart, Range, ReduceAll, Add(quant8)

Support dynamically shaped tensors

• Support static shape inference (input resizing)
• Support dynamic shape inference (general resizing)

Support Control Flow operations

• IF and WHILE
• Fully support static and dynamic tensors

API updates

• Introduce nnfw_set_input_tensorinfo for input resizing
• nnfw_input_tensorinfo and nnfw_output_tensorinfo behavior have changed to return tensorinfo according to the session state

Release 1.5.0

ONE 1.5.0 Release Note

Feature Highlights

• ONE Compiler

  • Compiler supports more operations

• ONE Runtime

  • CPU backend supports more operations

ONE Compiler

Compiler supports more operations

The following operations are supported :

• Abs, Add, ArgMax, AvgPool2D, BatchToSpaceND, Cast, Concat, Const, Conv2D, Cos,
  Custom, DepthwiseConv2D, Div, Elu, Equal, Exp, ExpandDims, Fill, FloorDiv,
  FloorMod, FullyConnected, Gather, GatherNd, Greater, GreaterEqual, If,
  LeakyRelu, Less, LocalResponseNormalize, LogicalAnd, LogicalNot, LogicalOr,
  Logistic, Maximum, MaxPool2D, Mean, Minimum, MirrorPad, Mul, Neg, NotEqual,
  OneHot, Pack, Pad, Pow, Range, ReduceAny(Any), ReduceMax(Max), ReduceProd,
  ReduceSum(Sum), ReLU, RELU_N1_TO_1, Reshape, ResizeNearestneighbor, Rsqrt,
  Select, Shape, Sin, Slice, Softmax, SpaceToBatchND, SpaceToDepth, Split, SplitV,
  Square, SquaredDifference, Squeeze, StridedSlice, Sub, Tanh, Tile, TopKV2,
  Transpose, Unpack(Unstack), While, ZerosLike

ONE Runtime

CPU backend supports more operations

The following operations are supported on CPU backend :

• ArgMax, Cos, ExpandDims, Fill, Log, LogicalNot, LogicalOr, Mean, Neg, Pow,
  ReLU, ReduceAny, ReduceProd, Reverse, Round, Select, SquaredDifference, Tile,
  ZerosLike

Release 1.4.0

ONE 1.4.0 Release Note

Feature Highlights

• ONE Runtime(a.k.a onert)
  • CPU backend supports more operations
  • (Experimental) Android package is released

ONE Runtime

CPU backend supports more operations

The following operations are supported on CPU backend :

• Rsqrt
• Abs
• Sin
• StrideSlice
• Shape

(Experimental) Android package is released

• For building onert from scratch, read CrossBuildForAndroid.md.
• For developing application with onert, please read HowToUseONEAPI.md.