Contents

• Tutorial
  Toggle navigation of Tutorial
  • Many ways to implement a custom graph in ONNX
  • GraphBuilder: common API for ONNX
  • Light API for ONNX: everything in one line
  • Numpy API for ONNX
    Toggle navigation of Numpy API for ONNX
    • First examples with onnx-array-api
    • First examples with onnxruntime
  • Tools
  • Benchmarks
    Toggle navigation of Benchmarks
    • Benchmark of TreeEnsemble implementation
• API
  Toggle navigation of API
  • onnx_array_api.array_api
    Toggle navigation of onnx_array_api.array_api
    • onnx_array_api.array_api.onnx_numpy
    • onnx_array_api.array_api.onnx_ort
    • onnx_array_api.npx.npx_array_api
  • onnx_array_api.graph_api
  • onnx_array_api.light_api
  • onnx_array_api.translate_api
  • npx_core_api
  • npx.npx_functions
  • npx_jit_eager
  • npx.npx_numpy_tensors
  • npx_tensors
  • npx.npx_types
  • Shortcuts
  • npx.npx_var
  • onnx tools
  • ort
  • plotting
  • reference
  • tools
  • profiling
  • Float 8
  • validation.docs
• Technical Details
  Toggle navigation of Technical Details
  • Difficulty to implement an Array API for ONNX
• command lines
• Example gallery
  Toggle navigation of Example gallery
  • About float 8
  • Compares the conversions of the same model with different options
  • First examples with onnx-array-api
  • First examples with onnxruntime
  • Optimization with onnxruntime
  • Profiling with onnxruntime
  • Benchmark of TreeEnsemble implementation

More

• Change Logs
• License
• Long outputs uneasy to see

fromtypingimportDict,OptionalfromonnximportModelProtofrom..annotationsimportdomainfrom.modelimportOnnxGraph,ProtoTypefrom.varimportVar,Vars
[docs]defstart(opset:Optional[int]=None,opsets:Optional[Dict[str,int]]=None,ir_version:Optional[int]=None,)->OnnxGraph:"""    Starts an onnx model.    :param opset: main opset version    :param opsets: others opsets as a dictionary    :param ir_version: specify the ir_version as well    :return: an instance of :class:`onnx_array_api.light_api.OnnxGraph`    A very simple model:    .. runpython::        :showcode:        from onnx_array_api.light_api import start        onx = start().vin("X").Neg().rename("Y").vout().to_onnx()        print(onx)    Another with operator Add:    .. runpython::        :showcode:        from onnx_array_api.light_api import start        onx = (            start()            .vin("X")            .vin("Y")            .bring("X", "Y")            .Add()            .rename("Z")            .vout()            .to_onnx()        )        print(onx)    """returnOnnxGraph(opset=opset,opsets=opsets,ir_version=ir_version)
[docs]defg()->OnnxGraph:"""    Starts a subgraph.    :return: an instance of :class:`onnx_array_api.light_api.OnnxGraph`    """returnOnnxGraph(proto_type=ProtoType.GRAPH)