# tensorplay.onnx API

Source: https://www.tensorplay.cn/docs/api/tensorplay.onnx.html

## Functions 6

[#](#api-tensorplay.onnx.export)

### export

function[Full reference ↗](/docs/generated/tensorplay.onnx.export.html)

```python
tensorplay.onnx.export(exported_program: ExportedProgram | Any, f: Any = None, *, input_names: Sequence[str] | None = None, output_names: Sequence[str] | None = None, opset_version: int | None = None, dynamic_axes: Mapping[str, Mapping[int, str] | Sequence[int]] | None = None, do_constant_folding: bool = True, verify: bool = False, rtol: float = 0.0001, atol: float = 1e-05, external_data: bool | None = None, external_data_location: str | None = None, check_model: bool = True) → Any
```

Export a TensorPlay model to ONNX.

Parameters:

- exported_program – an [ExportedProgram](/docs/generated/tensorplay.export.ExportedProgram.html#tensorplay.export.ExportedProgram), or a (model, *args, kwargs) sequence captured on the fly.

- f – file path or writable binary file object. When omitted the ModelProto is returned instead of being written.

- input_names – names for the graph inputs, in placeholder order.

- output_names – names for the graph outputs.

- opset_version – target ONNX opset (default 18, minimum 13).

- dynamic_axes – {value_name: {axis: axis_name}} (or a list of axis indices) marking dimensions that vary at runtime. Applies to both inputs and outputs.

- do_constant_folding – fold subgraphs whose inputs are all constants.

- verify – run the exported model under onnxruntime and compare against eager execution of exported_program.

- rtol/atol – tolerances used by verify.

- external_data – store initializers in a side-car file. None decides from the model size (models at or above the 2 GiB protobuf limit).

- external_data_location – side-car file name for external_data.

- check_model – run onnx.checker over the finished model.

Returns:

The onnx.ModelProto when f is None, else None.

[#](#api-tensorplay.onnx.is_supported)

### is_supported

function[Full reference ↗](/docs/generated/tensorplay.onnx.is_supported.html)

```python
tensorplay.onnx.is_supported(target: Any) → bool
```

Whether a captured call_function target has an ONNX lowering.

[#](#api-tensorplay.onnx.lookup_function_handler)

### lookup_function_handler

function[Full reference ↗](/docs/generated/tensorplay.onnx.lookup_function_handler.html)

```python
tensorplay.onnx.lookup_function_handler(module: str, name: str) → tuple[Callable[[OpContext], Any], list[str]] | None
```

[#](#api-tensorplay.onnx.lookup_method_handler)

### lookup_method_handler

function[Full reference ↗](/docs/generated/tensorplay.onnx.lookup_method_handler.html)

```python
tensorplay.onnx.lookup_method_handler(name: str) → tuple[Callable[[OpContext], Any], list[str]] | None
```

[#](#api-tensorplay.onnx.tp_export)

### tp_export

function[Full reference ↗](/docs/generated/tensorplay.onnx.tp_export.html)

```python
tensorplay.onnx.tp_export(model: Callable[[...], Any], *args: Any, dynamic_shapes: Any = None, strict: bool = False, preserve_module_call_signature: Any = (), **kwargs: Any) → ExportedProgram
```

Capture a callable and return an executable graph program.

Parameters:

- model – an nn.Module or plain callable; child modules are inlined.

- args/kwargs – example inputs binding argument defaults.

- dynamic_shapes – dimension specification per argument (dict, sequence, ShapesCollection, or AdditionalInputs).

- strict – reserved for callers of the strict capture contract; capture validation is identical in both modes.

- preserve_module_call_signature – submodule paths whose call metadata is recorded in module_call_graph for module-level tooling.

[#](#api-tensorplay.onnx.verify_model)

### verify_model

function[Full reference ↗](/docs/generated/tensorplay.onnx.verify_model.html)

```python
tensorplay.onnx.verify_model(model: Any, *, expected: Any, input_names: Sequence[str], example_inputs: dict[str, Any], rtol: float = 0.0001, atol: float = 1e-05, raise_on_mismatch: bool = True) → VerificationResult
```

Compare model under onnxruntime against the eager result.

Parameters:

- model – the built ModelProto.

- expected – what eager TensorPlay produced for example_inputs — a tensor, or an arbitrarily nested tuple/list of them.

- input_names – ONNX graph input names, in placeholder order.

- example_inputs – placeholder name -> example tensor.

- rtol/atol – tolerances forwarded to numpy.allclose().

- raise_on_mismatch – raise [VerificationError](/docs/generated/tensorplay.onnx.VerificationError.html#tensorplay.onnx.VerificationError) instead of returning a failed result.

## Classes 5

[#](#api-tensorplay.onnx.GraphBuilder)

### GraphBuilder

class[Full reference ↗](/docs/generated/tensorplay.onnx.GraphBuilder.html)

```python
class tensorplay.onnx.GraphBuilder(opset: int, name: str = 'tensorplay_model')
```

Accumulates ONNX nodes, initializers and unique value names.

```python
constant(value: Any, dtype: Any = None, name_hint: str = 'const') → str
```

Materialize a python/tensor constant as a cached initializer.

```python
op(op_type: str, inputs: Sequence[str], *, num_outputs: int = 1, name_hint: str | None = None, outputs: Sequence[str] | None = None, **attrs: Any) → Any
```

Emit one ONNX node and return its output name (or list of names).

[#](#api-tensorplay.onnx.OpContext)

### OpContext

class[Full reference ↗](/docs/generated/tensorplay.onnx.OpContext.html)

```python
class tensorplay.onnx.OpContext(builder: GraphBuilder, node_name: str, params: Sequence[str], args: Sequence[Any], kwargs: dict[str, Any], out_shape: tuple | None = None, out_dtype: dtype | None = None)
```

Argument access plus emission helpers handed to every handler.

```python
cast_like(value: Any, reference: Any, name_hint: str = 'const') → str
```

Name for value, materialized with reference’s dtype.

```python
name(value: Any, name_hint: str = 'const') → str
```

ONNX value name for value, materializing constants on demand.

```python
property x: Any
```

First declared argument (the input tensor for nearly every op).

[#](#api-tensorplay.onnx.TensorProto)

### TensorProto

class[Full reference ↗](/docs/generated/tensorplay.onnx.TensorProto.html)

```python
class tensorplay.onnx.TensorProto
```

```python
ByteSize()
```

Returns the size of the message in bytes.

```python
Clear()
```

Clears the message.

```python
ClearExtension()
```

Clears a message field.

```python
ClearField()
```

Clears a message field.

```python
CopyFrom()
```

Copies a protocol message into the current message.

```python
DiscardUnknownFields()
```

Discards the unknown fields.

```python
Extensions
```

Extension dict

```python
FindInitializationErrors()
```

Finds unset required fields.

```python
classmethod FromString()
```

Creates new method instance from given serialized data.

```python
HasExtension()
```

Checks if a message field is set.

```python
HasField()
```

Checks if a message field is set.

```python
IsInitialized()
```

Checks if all required fields of a protocol message are set.

```python
ListFields()
```

Lists all set fields of a message.

```python
MergeFrom()
```

Merges a protocol message into the current message.

```python
MergeFromString()
```

Merges a serialized message into the current message.

```python
ParseFromString()
```

Parses a serialized message into the current message.

```python
SerializePartialToString()
```

Serializes the message to a string, even if it isn’t initialized.

```python
SerializeToString()
```

Serializes the message to a string, only for initialized messages.

```python
SetInParent()
```

Sets the has bit of the given field in its parent message.

```python
UnknownFields()
```

Parse unknown field set

```python
WhichOneof()
```

Returns the name of the field set inside a oneof, or None if no field is set.

[#](#api-tensorplay.onnx.Value)

### Value

class[Full reference ↗](/docs/generated/tensorplay.onnx.Value.html)

```python
class tensorplay.onnx.Value(name: str, shape: tuple | None = None, dtype: dtype | None = None)
```

A tensor flowing through the ONNX graph under construction.

[#](#api-tensorplay.onnx.VerificationResult)

### VerificationResult

class[Full reference ↗](/docs/generated/tensorplay.onnx.VerificationResult.html)

```python
class tensorplay.onnx.VerificationResult(matched: bool, max_abs_diff: float = 0.0, max_rel_diff: float = 0.0, mismatches: list[str] = <factory>)
```

Per-output comparison between eager TensorPlay and onnxruntime.

## Exceptions 4

[#](#api-tensorplay.onnx.OnnxExporterError)

### OnnxExporterError

exception[Full reference ↗](/docs/generated/tensorplay.onnx.OnnxExporterError.html)

```python
exception tensorplay.onnx.OnnxExporterError
```

Base class for errors raised by the ONNX exporter.

[#](#api-tensorplay.onnx.OnnxExporterWarning)

### OnnxExporterWarning

exception[Full reference ↗](/docs/generated/tensorplay.onnx.OnnxExporterWarning.html)

```python
exception tensorplay.onnx.OnnxExporterWarning
```

Warnings raised during ONNX export.

[#](#api-tensorplay.onnx.UnsupportedOperatorError)

### UnsupportedOperatorError

exception[Full reference ↗](/docs/generated/tensorplay.onnx.UnsupportedOperatorError.html)

```python
exception tensorplay.onnx.UnsupportedOperatorError(name: str, version: int | None = None, supported_version: int | None = None)
```

Raised when a captured operation has no ONNX lowering.

[#](#api-tensorplay.onnx.VerificationError)

### VerificationError

exception[Full reference ↗](/docs/generated/tensorplay.onnx.VerificationError.html)

```python
exception tensorplay.onnx.VerificationError
```

Raised when the exported graph disagrees with eager execution.