furiosa.quantizer.frontend.onnx.spec package

Submodules

furiosa.quantizer.frontend.onnx.spec.export_spec module

class furiosa.quantizer.frontend.onnx.spec.export_spec.OnnxExportSpec(model: onnx.onnx_ml_pb2.ModelProto)

Bases: furiosa.quantizer.interfaces.export_spec.ExportSpec

add(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

avgpool2d(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

clip(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

concatenation(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

conv2d(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

convtranspose2d(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

depthtospace(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

div(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

exp(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

expand(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

export() → Tuple[List[furiosa.quantizer.ir.spec.OperatorSpec], Set[str]]: Traverse graph and export nodes as specs. Returns (a list of Spec, a set of unsupported ops)

flatten(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

gelu(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

gemm(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

static get_attributes(model: onnx.onnx_ml_pb2.ModelProto) → Dict[str, Dict[str, int]]

get_initializer_for_gen_spec(input_name: str) → List[int]

get_inputs_for_gen_spec(node: onnx.onnx_ml_pb2.NodeProto) → Tuple[List[Tuple[int]], List[Tuple[int]], Dict]

static get_tensor_shapes(model: onnx.onnx_ml_pb2.ModelProto) → Dict[str, Tuple[int]]

layer_norm(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

lp_norm(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

matmul(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

maxpool2d(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

mul(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

multi_node_lp_norm(node: onnx.onnx_ml_pb2.NodeProto) → Optional[Tuple[furiosa.quantizer.ir.spec.Spec, List[str]]]

Starts from ‘Div’, traverse up to find the form of l2norm. Returns all inputs of l2norm, consist of multi node

LpNormalization is not defined in ONNX Operator spec, so that we should traverse the graph:

Input –> ReduceL2 –> Clip –> Expand –> D: ———————————–> iv –> Output

property multi_node_spec: Dict[str, List[Callable[[onnx.onnx_ml_pb2.NodeProto], Optional[Tuple[furiosa.quantizer.ir.spec.Spec, List[str]]]]]]

pad(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

reduce_l2(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

reduce_mean(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

reduce_sum(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

reshape(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

resize(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

sigmoid(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

property single_node_spec: Dict[str, Callable[[onnx.onnx_ml_pb2.NodeProto], furiosa.quantizer.ir.spec.Spec]]

property skip_node: Set[str]

slice(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

softmax(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

softplus(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

split(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

squeeze(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

sub(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

transpose(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

traverse_multi_node_spec(node: onnx.onnx_ml_pb2.NodeProto) → Optional[Tuple[furiosa.quantizer.ir.spec.Spec, List[str]]]: Returns (Spec, list of inputs of the node)

traverse_single_node_spec(node: onnx.onnx_ml_pb2.NodeProto) → Optional[Tuple[furiosa.quantizer.ir.spec.Spec, List[str]]]: Returns (Spec, list of inputs of the node)

unsqueeze(node: onnx.onnx_ml_pb2.NodeProto) → furiosa.quantizer.ir.spec.Spec

furiosa.quantizer.frontend.onnx.spec.spec_utils module

furiosa.quantizer.frontend.onnx.spec.spec_utils.gemm_shapes(input_shapes, transA, transB)

furiosa.quantizer.frontend.onnx.spec.spec_utils.horizontal_pads(f1, f2, f3, f4, s1, s2, s3, s4)

furiosa.quantizer.frontend.onnx.spec.spec_utils.implicit_axis_to_explicit(axes, input_shape)

furiosa.quantizer.frontend.onnx.spec.spec_utils.node_identifier(node: onnx.onnx_ml_pb2.NodeProto) → str: In the case of onnx, FuriosaAI uses the first output node’s name as a node identifier.

furiosa.quantizer.frontend.onnx.spec.spec_utils.slice_offset_dict(starts, axes, input_shape)

Module contents