Skip to content

Latest commit

 

History

History
89 lines (79 loc) · 4.26 KB

CHANGELOGS.rst

File metadata and controls

89 lines (79 loc) · 4.26 KB
Raw

Change Logs

0.3.0

  • :pr:`189`: use onnxruntime==1.19.2 as default, pybind11 2.13.5, MatX 0.8.0
  • :pr:`187`: Fix compilation with GCC>=13 #187
  • :pr:`185`: adds custom operator MulMulSigmoid on CUDA
  • :pr:`184`: use onnxruntime==1.18.0 as default
  • :pr:`181`: adds MaskedScatterNDOfShape custom operator
  • :pr:`175`: adds custom operator MulSub and SubMul on CUDA
  • :pr:`173`: adds custom operator AddSharedInput, MulSharedInput on CUDA
  • :pr:`170`: adds custom operator TriMatrix on CUDA
  • :pr:`169`: adds custom operator ReplaceZero on CUDA
  • :pr:`168`: adds custom operator MulSigmoid on CUDA
  • :pr:`167`: adds custom operator Rotary on CUDA
  • :pr:`166`, :pr:`178`: adds custom operators AddMul, MulAdd on CUDA
  • :pr:`165`: adds custom operators AddAddAdd, MulMulMul on CUDA
  • :pr:`163`: use onnxruntime==1.17.3 as default
  • :pr:`162`: add ScatterNDOfShape implementation on CUDA without atomics
  • :pr:`159`: add AddAdd custom operator on CUDA
  • :pr:`158`: add MulMul custom operator on CUDA
  • :pr:`157`: add ScatterNDOfShape custom operator
  • :pr:`155`: add a function to draw a timeline from a profile
  • :pr:`154`: improves ploting legend for profiling
  • :pr:`151`: refactoring of TreeEnsemble code to make them faster
  • :pr:`129`, :pr:`132`: support sparse features for TreeEnsemble

0.2.4

  • :pr:`120`: use onnxruntime==1.16.3 as default
  • :pr:`115`, :pr:`116`, :pr:`118`: adds C implementation of SVMRegressor, SVMClassifier reference operator based on it, and custom kernels for onnxruntime as well
  • :pr:`111`, :pr:`117`, :pr:`119`: adds C implementation of TfIdfVectorizer + python implementation of Tokenizer + custom kernel for onnxruntime
  • :pr:`110`: allows LEQ as an alias for BRANCH_LEQ for nodes_modes in TreeEnsemble* operators
  • :pr:`108`: improves command lines documentation, fix an issue in command line stats
  • :pr:`103`: add methods to compute statistics on TreeEnsemble and initializers

0.2.3

  • :pr:`99`: use onnxruntime==1.16.1 as default
  • :pr:`96`: implements a fonction to convert a ModelProto into string (not bytes), add a function to multiply the number of trees in a TreeEnsemble
  • :pr:`75`: add an implementation of murmurhash3 to validate some options
  • :pr:`93`: validates the wheels in CI
  • :pr:`89`: add a function to merge models and update them if both have different opsets

0.2.2

  • :pr:`87`: update the quantization tools to use a simplified dynamic linear quantization into float 8
  • :pr:`85`: add load_model, save_model to help saving with/without external data
  • :pr:`82`: fixes benchmark on multiple versions of onnxruntime

0.2.1

  • :pr:`79`: update to onnxruntime v1.16.0
  • :pr:`77`: helpers to benchmark a model
  • :pr:`74`: add a function to enumerate all intermediate results with onnxruntime
  • :pr:`71`, :pr:`72`, :pr:`73`: add function to analyse a profile produce by onnxruntime
  • :pr:`68`, :pr:`69`, :pr:`70`: add CPU implementation for CustomGemmFloat8
  • :pr:`67`: add a function to extract a subgraph of a model
  • :pr:`59`, :pr:`60`, :pr:`61`, :pr:`62`, :pr:`63`, :pr:`65`, :pr:`66`, :pr:`68`, :pr:`69`, :pr:`70`: add local functions to quantize into float 8, float 16
  • :pr:`57`: add C implementation for DynamicQuantizeLinear (for experimentation)
  • :pr:`56`: add C implementation to cast a float into float 8
  • :pr:`55`, :pr:`58`: add basic functionality to transform a graph, starts with basic quantization
  • :pr:`51`: fix optimized TreeEnsembleRegressor and adds TreeEnsembleClassifier as custom ops
  • :pr:`50`: add command line store to store intermediate outputs
  • :pr:`49`: add option to save intermediate results in CReferenceEvaluator
  • :pr:`45`: add option cuda-link to setup.py to specify how to link with CUDA library
  • :pr:`41`: implements a custom kernel for RandomForestRegressor easier to optimize
  • :pr:`34`: update to onnxruntime v1.15.1
  • :pr:`31`: implement a custom CUDA kernel (gemm)
  • :pr:`32`: update to onnxruntime v1.15.0
  • :pr:`27`: add a custom kernel with parameters to onnxruntime
  • :pr:`26`: add a custom kernel to onnxruntime
  • :pr:`24`: use Eigen to implement Conv operator
  • :pr:`23`: make pip wheel . work
  • :pr:`22`: rename cmake into _cmake to avoid warnings related to cmake package
  • :pr:`19`: minimal settings to use onnxruntime
  • :pr:`14`: minimal setting to use CUDA
  • :pr:`8`: support for C++ unit test