Skip to content

Interface: TorchFxNode

Defined in: ml/model.ts:180

One node of a traced torch.fx graph. Produce it in Python with:

python
gm = torch.fx.symbolic_trace(model)
[{"name": n.name, "op": n.op, "target": str(n.target),
  "args": [a.name for a in n.all_input_nodes]} for n in gm.graph.nodes]

Properties

args?

ts
optional args?: string[];

Defined in: ml/model.ts:187

Names of the input nodes ([a.name for a in n.all_input_nodes]).


flops?

ts
optional flops?: number;

Defined in: ml/model.ts:192


moduleType?

ts
optional moduleType?: string;

Defined in: ml/model.ts:189

type(module).__name__ for call_module nodes — the nicest display type.


name

ts
name: string;

Defined in: ml/model.ts:181


op

ts
op: string;

Defined in: ml/model.ts:183

placeholder | call_module | call_function | call_method | get_attr | output.


params?

ts
optional params?: number;

Defined in: ml/model.ts:191


shape?

ts
optional shape?: number[];

Defined in: ml/model.ts:190


target?

ts
optional target?: string;

Defined in: ml/model.ts:185

Module path or function name.

MIT licensed. WebGL2 required.