nncase v1.5.0

Major changes

• Support TFLite HardSwish (#539)
• Support TCN

Bug fixes

#538 Fix k210 .data section is too large.
#556 Fix overhead of dbg macro

nncase v1.4.0

Major changes

• Support Bidirection/Reverse LSTM (#525, #529)
• Optimize binary ops for RVV (#516)

Bug fixes

#514, #515, #526

nncase v1.3.0

Major changes

• Support TFLite ops (Split) (#491)
• Support ONNX ops (Equal, Not) (#495, #498)

Bug fixes

#496, #504, #505

nncase v1.2.0

Major changes

• Support ONNX ops (TopK, Trilu, RoiAlign) (#466 , #477 , #493)
• Minor bug fixes

nncase v1.1.0

Major changes

• K210 can accelerate Conv2DTranspose (#421)
• Support ONNX ops (QuantizeLinear, Random*, ReduceLogSum/ReduceLogSumExp/ReduceSumSquare, Celu/Selu/Softplus/Softsign, LogSoftmax, Size, SpaceToDepth, Sign, ReduceProd, Where, Tile, ReverseSequence, Lstm) (#424 , #427 , #429 , #437 , #439 , #441 , #442 , #444 , #445 , #446 , #447 , #449, #425 )
• Make letterbox preprocess optional (#448 )
• Add python interface document
• Minor bug fixes

nncase v1.0.0

Major changes

• Support ONNX ops (ConstantOfShape, DepthToSpace, And, ArgMax/ArgMin, DequantizeLinear, CumSum, HardMax) (#393, #403, #398, #407, #411, #414, #416)
• New Functional API (#384)
• Support dump imported graph's output range (#418)
• Minor bug fixes

nncase v1.0.0 rc1

Major changes

• Support ONNX ops (HardSwish, ReduceL1, Elu, Sinh/Cosh/Asin/Acos/Asinh/Acosh/Tanh) (#355, #360, #372, #376)
• Update YOLOX demo (#364)
• Support dump quantize errors (#366)
• MSE quantization for DepthwiseConv2D weights (#375)
• New preprocess API (#370, #379)
• Minor bug fixes

nncase v1.0.0 beta2

Bugs fixes

• Enable remove_exclusive_copy_to_output (#353)
• fix preprocess (#354)

nncase v1.0.0 beta1

Major changes

• Upgrade to kmodel v5
• Python compiler interface
• Support preprocess methods: normalize letter box
• More PTQ methods: l2 kld_m0 kld_m1 kld_m2 cdf
• More TFLite ONNX & Caffe Ops
• Plugable compiler backends
• Ready for K510

New examples

YoloX

NNCase v0.2.0 Beta4

Major changes

• Add --weights-quantize-threshold, --output-quantize-threshold and --no-quantized-binary (See usage)

Bugs fixes

• Fix a bug in binary ops optimization
• Fix a bug in k210 pooling optimization
• Some bug fixes in tflite and onnx importer