Graph methods¶

Any graph must start with functionstart.It is usually following byvin to add an input.

• bring (Var.bring,Vars.bring):assembles multiple results into a set before calling an operator taking mulitple inputs,

• cst (Var.cst,Vars.cst):adds a constant tensor to the graph,

• rename (Var.rename,Vars.rename):renames or give a name to a variable in order to call it later.

• vin (Var.vin,Vars.vin):adds an input to the graph,

• vout (Var.vout,Vars.vout):declares an existing result as an output.

These methods are implemented in classonnx_array_api.light_api.var.BaseVar

Operator methods¶

They are described inONNX Operators and redefined in a stable APIso that the definition should not change depending on this opset.onnx_array_api.light_api.Var defines all operators taking only one input.onnx_array_api.light_api.Vars defines all other operators.

Numpy methods¶

Numpy users expect methods such asreshape, propertyshape oroperator+ to be available as well and that the case. They aredefined in classVar orVars depending on the number ofinputs they require. Their name starts with a lower letter.

Other domains¶

The following example uses operatorNormalizer from domainai.onnx.ml. The operator name is called with the syntax<domain>.<operator name>. The domain may have dots in its namebut it must follow the python definition of a variable.The operatorNormalizer becomesai.onnx.ml.Normalizer.

<<<

importnumpyasnpfromonnx_array_api.light_apiimportstartfromonnx_array_api.plotting.text_plotimportonnx_simple_text_plotmodel=(start(opset=19,opsets={"ai.onnx.ml":3}).vin("X").reshape((-1,1)).rename("USE").ai.onnx.ml.Normalizer(norm="MAX").rename("Y").vout().to_onnx())print(onnx_simple_text_plot(model))

>>>

opset:domain=''version=19opset:domain='ai.onnx.ml'version=3input:name='X'type=dtype('float32')shape=Noneinit:name='r'type=int64shape=(2,)--array([-1,1])Reshape(X,r)->USENormalizer(USE,norm=b'MAX')->Youtput:name='Y'type=dtype('float32')shape=None