0.3.4¶

0.3.3¶

• #104: add code rendering when conveting a model into code

• #103: fix import issue with the latest onnx version

0.3.2¶

• #101: fix as_tensor in onnx_text_plot_tree

0.3.1¶

• #100: updates requirements, add 3.12

• #96: supports local functions in translator

• #95: improves translation to GraphBuilder

0.3.0¶

• #93: fixes evaluator type incompare_onnx_execution

• #92: avoids recursion errors in profiling

• #87: adds command line to replace contant by ConstantOfShape

• #79: first draft to export to GraphBuilder

• #77: supports ConcatOfShape and Slice with the light API

0.2.0¶

• #76,#79: add a mode to compare models without execution

• #75: add QuickGelu to ExtendedReferenceEvaluator

• #71: adds tools to compare two onnx graphs

• #61: adds function to plot onnx model as graphs

• #60: supports translation of local functions

• #59: add methods to update nodes in GraphAPI

0.1.3¶

• #57: implements GraphBuilder

• #49: adds command line to export a model into code

• #48: support for subgraph in light API

• #47: extends export onnx to code to support inner API

• #46: adds an export to convert an onnx graph into light API code

• #45: fixes light API for operators with two outputs

0.1.2¶

• #42: first sketch for a very simple API to create onnx graph in one or two lines

• #27: add function from_array_extended to convertan array to a TensorProto, including bfloat16 and float 8 types

• #24: add ExtendedReferenceEvaluator to support scenariofor the Array API onnx does not support

• #22: support OrtValue in functionort_profile

• #17: implements ArrayAPI

• #3: fixes Array API with onnxruntime and scikit-learn