#!/usr/bin/env python
# ******************************************************************************
# Copyright 2024 Brainchip Holdings Ltd.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ******************************************************************************
__all__ = ["QuantizedConv2DTranspose", "get_qconv_transpose", "QuantizedDepthwise2DTranspose"]
import numpy as np
from onnx import AttributeProto as AP, TensorProto as TP
from onnx.helper import make_node
from .base_layer import OnnxLayer
from .subgraph_ops import cast_tensors_to, get_scale_out_ops
from .subgraph_ops.activation import get_activation_ops
from .compute_shapes import compute_onnx_conv_output
from .layer_compatibility import check_clip_relu_compatibility, check_conv_depthwise_compatibility
from ..graph_tools import (TENSOR_SHAPE, get_field, get_node, get_variable, to_field,
check_node_attributes)
from ..quantization.weights import quantize_weights, quantize_vector, align_to
from ..quantization.outputs import downscale
def get_qconv_transpose(nodes, graph):
conv_node = nodes[0]
assert conv_node.op_type == 'ConvTranspose'
# Check supported attributes
weights = get_variable(conv_node.input[1], graph)
valid_attr = {'auto_pad': ['NOTSET'], 'dilations': [[1] * (weights.ndim - 2)],
"pads": [0, 0, 0, 0]}
check_node_attributes(conv_node, valid_attr)
if bool(get_field(conv_node, 'output_padding', False)) or bool(
get_field(conv_node, 'output_shape', False)):
raise ValueError("Unsupported attributes output_padding or output_shape")
act_node = get_node(nodes, 'Relu')
clip_node = get_node(nodes, 'Clip')
# Retrieve attributes
strides = get_field(conv_node, 'strides', (1, 1))
group = get_field(conv_node, "group", 1)
activation = bool(act_node) or bool(clip_node)
if group == 1:
qconv = QuantizedConv2DTranspose(strides=strides,
name=conv_node.name,
activation=activation)
else:
# need to check supported attributes
check_conv_depthwise_compatibility(conv_node, graph)
qconv = QuantizedDepthwise2DTranspose(strides=strides,
name=conv_node.name,
activation=activation)
# Sets the weights to configure the operation chain
qconv.set_weight("kernel", weights)
# If third attribute is there and it is not empty, then there is a bias
if len(conv_node.input) == 3 and conv_node.input[2]:
qconv.set_weight("bias", get_variable(conv_node.input[2], graph))
if clip_node:
check_clip_relu_compatibility(clip_node, graph)
qconv.set_weight("max_value", get_variable(clip_node.input[2], graph))
return qconv
[docs]class QuantizedConv2DTranspose(OnnxLayer):
"""Intermediate representation of the upsampling layer QuantizedConv2DTranspose().
Args:
strides (list of int, optional): the convolutional strides. Defaults to [1, 1].
activation (bool, optional): whether to apply relu operation. Defaults to False.
name (str, optional): the node name. Defaults to ''.
"""
def __init__(self,
strides=[1, 1],
activation=False,
name=''):
super().__init__("QuantizedConv2DTranspose",
strides=strides,
name=name)
# Save properties need to serialize operation name
self.serialize_attr["activation"] = activation
self.serialize_attr["scale"] = True
# Declare weights
self._add_weight("kernel")
self._add_weight("bias")
self._add_weight("max_value")
def __build__(self, input_ts, downscale=True):
assert input_ts.dtype == np.int8
assert downscale, f"{self.name} ({self.base_name}) does not support 32bit output"
assert self.weights["kernel"].ndim == 4
# Compute output shape
conv_output_shape = compute_onnx_conv_output(self, input_ts.shape,
apply_pool=False, transpose=True)
output_ts = TENSOR_SHAPE(conv_output_shape, np.dtype("int8"))
return output_ts
def __quantize__(self, qinput, out_tensor_range, force_fp=False):
i_scale = qinput.weights["scale"]
# Perform cross-layer equalization, i.e.: rescale weights with input scale.
# To do that first reshape i_scale to put it into axis = 0 and be capable of broadcasting.
assert i_scale.ndim <= 1
kernel = self.weights["kernel"]
kernel = kernel / align_to(i_scale, kernel.ndim, axis=0)
# Quantize and set weights. The quantize weights function assumes the weight
# ordering is (FCHW) like convolution but ConvTranspose weights are (C,F,kH,kW)
kernel = kernel.transpose((1, 0, 2, 3))
qweights, i_scale = quantize_weights(kernel)
qweights = qweights.transpose((1, 0, 2, 3))
# Prepare tensors list with unique names
conv_name = self.name
prefix = conv_name + "_"
weights_dict = {}
bias = self.weights["bias"]
weights_dict[prefix + "Wi"] = qweights
if "Biased" in self.op_type:
qbias = quantize_vector(bias, i_scale)
weights_dict[prefix + "B"] = qbias
# Reshape i_scale to match with channel axis
i_scale = align_to(i_scale, qweights.ndim)
# Quantize max value when there is an activation
if "Clipped" in self.op_type:
qmax_value = quantize_vector(self.weights["max_value"], i_scale, signed=False)
weights_dict[prefix + "max_value"] = qmax_value
# Now consider calibrated output range
scale, s_out, ocalib_scale = downscale(out_tensor_range, i_scale, force_fp=force_fp)
weights_dict.update({prefix + "M": scale.astype("uint8"), prefix + "S_out": s_out})
# Return quantized weights and output scale
return weights_dict, ocalib_scale
@staticmethod
def build_subgraph(op_type):
# Cast input, weights (and bias) into float.
t_names = ["X", "W", ""]
if "Biased" in op_type:
t_names[-1] = "bias"
nodes, t_names = cast_tensors_to(t_names)
# Transpose convolution
nodes.append(make_node("ConvTranspose", inputs=t_names, outputs=["Yi"]))
nodes[-1].attribute.append(AP(name="strides", ref_attr_name="strides", type=AP.INTS))
# Activation (optional)
if "ReLU" in op_type:
# Replace previous output as relu input
nodes[-1].output.__setitem__(0, nodes[-1].op_type)
nodes += get_activation_ops(nodes[-1].output[0], "Yi", "ReLUClipped" in op_type)
# Scale out (with saturation) in float domain
shift_nodes, shift_t_names = cast_tensors_to(["Scale", "Shift"])
nodes += shift_nodes
nodes += get_scale_out_ops("Yi", "Yscaled", *shift_t_names)
# Cast output to expect type
nodes.append(make_node("Cast", ["Yscaled"], ["Y"], to=TP.INT8))
return nodes
[docs]class QuantizedDepthwise2DTranspose(QuantizedConv2DTranspose):
""" Intermediate representation of the upsampling layer QuantizedDepthwise2DTranspose.
Inherits from QuantizedConv2DTranspose: only different attribute is group.
Args:
strides (list of int, optional): the convolutional strides. Defaults to [1, 1].
activation (bool, optional): whether to apply relu operation. Defaults to False.
name (str, optional): the node name. Defaults to ''.
"""
def __init__(self,
strides=[1, 1],
activation=False,
name='',
):
super().__init__(activation=activation,
strides=strides,
name=name)
self.base_name = "QuantizedDepthwise2DTranspose"
def __build__(self, input_ts, downscale=True):
# ConvTranspose weights are (C,F,kH,kW)
kernel_shape = self.weights["kernel"].shape
expect_shape = (input_ts.shape[1], 1, *kernel_shape[-2:])
if expect_shape != kernel_shape:
raise ValueError("Kernel shape does not match with the following format: "
f"(input channels, 1, Kx, Ky). Receives: {kernel_shape} and "
f"expected: {expect_shape}")
# Include group in node as attribute
self.attribute.append(to_field("groups", expect_shape[0]))
return super().__build__(input_ts, downscale=downscale)
@staticmethod
def build_subgraph(op_type):
# Cast input, weights (and bias) into float.
t_names = ["X", "W", ""]
if "Biased" in op_type:
t_names[-1] = "bias"
nodes, t_names = cast_tensors_to(t_names)
# Transpose convolution
nodes.append(make_node("ConvTranspose", inputs=t_names, outputs=["Yi"]))
nodes[-1].attribute.append(AP(name="strides", ref_attr_name="strides", type=AP.INTS))
nodes[-1].attribute.append(AP(name="group", ref_attr_name="groups", type=AP.INT))
# Activation (optional)
if "ReLU" in op_type:
# Replace previous output as relu input
nodes[-1].output.__setitem__(0, nodes[-1].op_type)
nodes += get_activation_ops(nodes[-1].output[0], "Yi", "ReLUClipped" in op_type)
# Scale out (with saturation) in float domain
shift_nodes, shift_t_names = cast_tensors_to(["Scale", "Shift"])
nodes += shift_nodes
nodes += get_scale_out_ops("Yi", "Yscaled", *shift_t_names)
# Cast output to expect type
nodes.append(make_node("Cast", ["Yscaled"], ["Y"], to=TP.INT8))
return nodes