# Copyright (C) 2020 Arm Limited or its affiliates. All rights reserved. # # SPDX-License-Identifier: Apache-2.0 # # 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 # # 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. # Description: # The SupportedOperators class which is a collection of all supported operators and parameter checks. from collections import defaultdict import numpy as np from .data_type import BaseType from .data_type import DataType from .numeric_util import is_integer from .operation import get_slice_offsets from .operation import Op # Custom decorator function to allow formatting docstrings containing "{}" def docstring_format_args(args): def docstring(func): func.__doc__ = func.__doc__.format(*args) return func return docstring def warn_cpu(op, msg): print("Warning: {} {}, placing on CPU".format(op.type, msg)) class SupportedOperators: # Categorised lists of supported operators npu_pre_ops = set((Op.SplitSliceRead,)) convolution_ops = set((Op.Conv2DBias, Op.Conv2D, Op.QuantizedConv2D,)) depthwise_convolution_ops = set((Op.DepthwiseConv2DBias,)) transpose_convolution_ops = set((Op.Conv2DBackpropInput,)) convolution_like_ops = convolution_ops | depthwise_convolution_ops | transpose_convolution_ops max_pooling_ops = Op.op_set(Op.is_maxpool_op) avg_pooling_ops = Op.op_set(Op.is_avgpool_op) pooling_ops = set((Op.ReduceSum,)) | max_pooling_ops | avg_pooling_ops resizing_ops = set((Op.ResizeBilinear,)) fc_vector_products = set((Op.QuantizedMatMul, Op.MatMul, Op.FullyConnected,)) mac_main_ops = ( # RNN/LSTM/GRU set((Op.BlockLSTM,)) # conv/depthwiseconv/transposeconv | convolution_like_ops # pooling | pooling_ops # resizing/upscaling | resizing_ops # FC layers | fc_vector_products ) unary_elem_wise_main_ops = Op.op_set(Op.is_unary_elementwise_op) binary_elem_wise_min_max_ops = set((Op.Minimum, Op.Maximum,)) binary_elem_wise_shift_ops = set((Op.SHL, Op.SHR,)) binary_elem_wise_add_mul_sub = set((Op.Add, Op.Mul, Op.Sub,)) binary_elem_wise_main_ops = binary_elem_wise_min_max_ops | binary_elem_wise_add_mul_sub | binary_elem_wise_shift_ops elem_wise_main_ops = binary_elem_wise_main_ops | unary_elem_wise_main_ops supported_int32_tensor_ops = ( set((Op.ReduceSum, Op.CLZ,)) | binary_elem_wise_add_mul_sub | binary_elem_wise_shift_ops ) activation_ops = set((Op.Relu, Op.Relu6, Op.ReluN1To1, Op.Sigmoid, Op.Tanh, Op.Softmax,)) npu_post_ops = ( # activation functions activation_ops # concatenation write direction | set((Op.ConcatSliceWrite,)) # Quantization | set((Op.Quantize,)) ) split_ops = set((Op.Split, Op.SplitV, Op.StridedSlice, Op.Slice, Op.UnpackReshaped, Op.Unpack,)) concat_ops = set((Op.Concat, Op.ConcatTFLite, Op.PackReshaped, Op.Pack,)) memory_only_ops = set((Op.Squeeze, Op.Reshape, Op.QuantizedReshape, Op.ExpandDims,)) | concat_ops | split_ops shapeless_input_ops = binary_elem_wise_main_ops | set((Op.Split, Op.SplitV,)) supported_fused_activations = set((Op.Relu, Op.Relu6, Op.ReluN1To1, Op.Tanh, Op.Sigmoid, Op.LUT,)) supported_operators = npu_pre_ops | mac_main_ops | elem_wise_main_ops | npu_post_ops | memory_only_ops # Supported data types supported_op_dtypes = set((DataType.uint8, DataType.int8, DataType.int16, DataType.int32)) supported_bias_dtypes = set((DataType.int32, DataType.int64)) # Defined ranges for allowed values: tens_dim_range = (1, 65535) stride_range = (1, 3) dilation_range = (1, 2) dilated_height_range = (1, 64) dilated_product_range = (1, 64 * 64) weights_limit = 127 * 65536 def __init__(self): # Setup supported operator restriction checkers self.supported_operator_restrictions = {} self.supported_operator_restrictions.update( {op: self.check_depthwise_convolution_restrictions for op in SupportedOperators.depthwise_convolution_ops} ) self.supported_operator_restrictions.update( {op: self.check_transpose_convolution_restrictions for op in SupportedOperators.transpose_convolution_ops} ) self.supported_operator_restrictions.update( {op: self.check_pooling_restrictions for op in SupportedOperators.pooling_ops} ) self.supported_operator_restrictions.update( {op: self.check_resize_restrictions for op in SupportedOperators.resizing_ops} ) self.supported_operator_restrictions.update( {op: self.check_vector_product_restrictions for op in SupportedOperators.fc_vector_products} ) self.supported_operator_restrictions.update( {op: self.check_element_wise_restrictions for op in SupportedOperators.elem_wise_main_ops} ) self.supported_operator_restrictions.update( {op: self.check_memory_only_restrictions for op in SupportedOperators.memory_only_ops} ) self.supported_operator_restrictions.update( {op: self.check_activation_ops for op in SupportedOperators.activation_ops} ) # Setup the generic constraints. Note: the order matters self.generic_constraints = [] self.generic_constraints.append(SupportedOperators.constraint_tens_defined_shape) self.generic_constraints.append(SupportedOperators.constraint_tens_output_shapeless) self.generic_constraints.append(SupportedOperators.constraint_tens_input_shapeless) self.generic_constraints.append(SupportedOperators.constraint_tens_shape_size) self.generic_constraints.append(SupportedOperators.constraint_tens_dtype) self.generic_constraints.append(SupportedOperators.constraint_tens_int32_ops) self.generic_constraints.append(SupportedOperators.constraint_tens_dimension) self.generic_constraints.append(SupportedOperators.constraint_tens_quant_none_check) self.generic_constraints.append(SupportedOperators.constraint_tens_quant_scale) self.generic_constraints.append(SupportedOperators.constraint_faf) # Setup specific constraints. The key in the dictionary must be a tuple of op types the constraints apply to self.specific_constraints = defaultdict(list) # Conv-like ops have the same checks applied to them: conv_like_ops = tuple(SupportedOperators.convolution_like_ops) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_stride_type) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_stride_range) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_dilation_type) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_dilation_range) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_dilated_height_range) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_dilated_product_range) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_weights_type) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_weights_nonconst) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_weights_limit) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_bias_type) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_bias_40bit) self.specific_constraints[conv_like_ops].append(SupportedOperators.constraint_batch_size) def get_constraints_list(self, op_type): constraint_list = list(self.generic_constraints) for ops in self.specific_constraints: if op_type in ops: constraint_list.extend(self.specific_constraints[ops]) return constraint_list def is_operator_supported(self, op): if op.type not in SupportedOperators.supported_operators: if op.type not in (Op.Placeholder, Op.SubgraphInput, Op.Const): print("Info: {} '{}' is not supported on the NPU. Placing on CPU instead".format(op.type, op.name)) return False for constraint in self.get_constraints_list(op.type): valid, extra = constraint(op) if not valid: print("Warning: {} '{}' is not supported on the NPU. Placing on CPU instead".format(op.type, op.name)) print(" - {}".format(constraint.__doc__)) if extra: print(" {}".format(extra)) return False if op.type in self.supported_operator_restrictions: return self.supported_operator_restrictions[op.type](op) return True @staticmethod def constraint_tens_defined_shape(op): "Input(s) and Output Tensors must have a defined shape" valid = True extra = [] tensors = [tens for tens in op.inputs + op.outputs if tens] for tens in tensors: if not tens.has_fully_defined_shape(): valid = False extra.append("Tensor '{}' has shape: {}".format(tens.name, tens.shape)) return valid, ", ".join(extra) @staticmethod def constraint_tens_output_shapeless(op): "Scalar or Broadcasting Tensors are only valid for Input Tensors" valid = True extra = [] for tens in op.outputs: if tens.shape == []: valid = False extra.append("Output Tensor '{}' is shapeless".format(tens.name)) return valid, ", ".join(extra) @classmethod @docstring_format_args([shapeless_input_ops]) def constraint_tens_input_shapeless(cls, op): "Scalar or Broadcasting Input Tensors are only valid for op type: {}" valid = True extra = [] tensors = [tens for tens in op.inputs if tens] for tens in tensors: if (tens.shape == []) and (op.type not in cls.shapeless_input_ops): valid = False extra.append(tens.name) extra = "Op has shapeless input tensor(s): {}".format(", ".join(extra)) return valid, extra @staticmethod def constraint_tens_shape_size(op): "Input(s) and Output Tensors must not be greater than 4D" valid = True extra = [] tensors = [tens for tens in op.inputs + op.outputs if tens] for tens in tensors: if len(tens.shape) > 4: valid = False extra.append("Tensor '{}' has shape: {}".format(tens.name, tens.shape)) return valid, ", ".join(extra) @classmethod @docstring_format_args([supported_op_dtypes]) def constraint_tens_dtype(cls, op): "Input(s), Output and Weight Tensors must be of type: {}" valid = True extra = [] tensors = [tens for tens in op.get_ifm_ifm2_weights_ofm() if tens] tensors = tensors if tensors else op.inputs for tens in tensors: if tens.dtype not in cls.supported_op_dtypes: valid = False extra.append("Tensor '{}' has data type: {}".format(tens.name, tens.dtype)) return valid, ", ".join(extra) @classmethod @docstring_format_args([supported_int32_tensor_ops]) def constraint_tens_int32_ops(cls, op): "Tensors which are int32 are only valid when op type is: {}" valid = True extra = [] tensors = [tens for tens in op.get_ifm_ifm2_weights_ofm() if tens] tensors = tensors if tensors else op.inputs for tens in tensors: if (tens.dtype == DataType.int32) and (op.type not in cls.supported_int32_tensor_ops): valid = False extra.append(tens.name) extra = "Op has int32 tensor(s): {}".format(", ".join(extra)) return valid, extra @classmethod @docstring_format_args(tens_dim_range) def constraint_tens_dimension(cls, op): "Tensor dimensions must be in the range [{}, {}]" tens_min, tens_max = cls.tens_dim_range valid = True extra = [] tensors = [tens for tens in op.get_ifm_ifm2_weights_ofm() if tens] tensors = tensors if tensors else op.inputs for tens in tensors: if not all(tens_min <= dim <= tens_max for dim in tens.shape): valid = False extra.append("Tensor '{}' has shape: {}".format(tens.name, tens.shape)) return valid, ", ".join(extra) @staticmethod def constraint_tens_quant_none_check(op): "Tensors must have quantization parameters" valid = True extra = [] tensors = [tens for tens in op.get_ifm_ifm2_weights_ofm() if tens] for tens in tensors: if tens.quantization is None: valid = False extra.append("Tensor '{}' has no quantization parameters".format(tens.name)) return valid, ", ".join(extra) @staticmethod def constraint_tens_quant_scale(op): "Tensors with quantization scales must be finite" valid = True extra = [] tensors = [tens for tens in op.get_ifm_ifm2_weights_ofm() if tens] for tens in tensors: if (tens.quantization.scale_f32 is not None) and np.isinf(tens.quantization.scale_f32).any(): valid = False extra.append("Tensor '{}' has quantization scale: {}".format(tens.name, tens.quantization.scale_f32)) return valid, ", ".join(extra) @classmethod @docstring_format_args([supported_fused_activations]) def constraint_faf(cls, op): "The fused activation function (if present) must be one of type: {}" faf = op.activation valid = (faf is None) or (faf in cls.supported_fused_activations) extra = "Op has its fused activation function as: {}".format(faf) return valid, extra @staticmethod def constraint_stride_type(op): "Stride values for both width and height must be integer types" w = op.attrs["stride_w"] h = op.attrs["stride_h"] valid = is_integer(w) and is_integer(h) extra = "Op has stride WxH as: {}x{}".format(repr(w), repr(h)) return valid, extra @classmethod @docstring_format_args(stride_range) def constraint_stride_range(cls, op): "Stride values for both width and height must be in the range [{}, {}]" w = op.attrs["stride_w"] h = op.attrs["stride_h"] stride_min, stride_max = cls.stride_range valid = (stride_min <= w <= stride_max) and (stride_min <= h <= stride_max) extra = "Op has stride WxH as: {}x{}".format(w, h) return valid, extra @staticmethod def constraint_dilation_type(op): "Dilation factor values for both width and height must be integer types" w = op.attrs.get("dilation_w_factor", 1) h = op.attrs.get("dilation_h_factor", 1) valid = is_integer(w) and is_integer(h) extra = "Op has dilation factor WxH as: {}x{}".format(repr(w), repr(h)) return valid, extra @classmethod @docstring_format_args(dilation_range) def constraint_dilation_range(cls, op): "Dilation factor values for both width and height must be in the range [{}, {}]" w = op.attrs.get("dilation_w_factor", 1) h = op.attrs.get("dilation_h_factor", 1) dilation_min, dilation_max = cls.dilation_range valid = (dilation_min <= w <= dilation_max) and (dilation_min <= h <= dilation_max) extra = "Op has dilation factor WxH as: {}x{}".format(w, h) return valid, extra @classmethod @docstring_format_args(dilated_height_range) def constraint_dilated_height_range(cls, op): "Dilated kernel height must be in the range [{}, {}]" h = (op.weights.shape[0] - 1) * op.attrs.get("dilation_h_factor", 1) + 1 dilated_height_min, dilated_height_max = cls.dilated_height_range valid = dilated_height_min <= h <= dilated_height_max extra = "Op has dilated kernel height as: {}".format(h) return valid, extra @classmethod @docstring_format_args(dilated_product_range) def constraint_dilated_product_range(cls, op): "Product of dilated kernel width and height must be in the range [{}, {}]" weights = op.weights w = (weights.shape[1] - 1) * op.attrs.get("dilation_w_factor", 1) + 1 h = (weights.shape[0] - 1) * op.attrs.get("dilation_h_factor", 1) + 1 product = w * h dilated_product_min, dilated_product_max = cls.dilated_product_range valid = dilated_product_min <= product <= dilated_product_max extra = "Op has product of dilated kernel width and height as: {}".format(product) return valid, extra @staticmethod def constraint_weights_type(op): "Weight Tensor must be 8-bit" weights = op.weights valid = weights.element_size() == 1 extra = "Tensor '{}' is {}-bit".format(weights.name, int(weights.element_size() * 8)) return valid, extra @staticmethod def constraint_weights_nonconst(op): "Weight tensor cannot be non-constant" weights = op.weights valid = weights.values is not None extra = "Tensor '{}' has non-constant values".format(weights.name) return valid, extra @classmethod @docstring_format_args([weights_limit]) def constraint_weights_limit(cls, op): "The sum of the weights cannot exceed {}" weights = op.weights values = weights.quant_values.astype(np.int64) - weights.quantization.zero_point limit = np.amax(np.sum(np.absolute(values), axis=(0, 1, 2))) valid = limit <= cls.weights_limit extra = "Tensor '{}' has the sum of weights: {}".format(weights.name, limit) return valid, extra @classmethod @docstring_format_args([supported_bias_dtypes]) def constraint_bias_type(cls, op): "Optional Bias Tensor must be of type: {}" valid = True extra = "" bias = op.bias if bias: valid = bias.dtype in cls.supported_bias_dtypes extra = "Tensor '{}' has data type: {}".format(bias.name, bias.dtype) return valid, extra @staticmethod def constraint_bias_40bit(op): "Optional Bias Tensor values must fit within 40-bits" valid = True extra = "" bias = op.bias if bias and bias.dtype == DataType.int64: valid = all(len(bin(quant_value)[2:]) <= 40 for quant_value in bias.quant_values) extra = "Tensor '{}' has values larger than 40-bits".format(bias.name) return valid, extra @staticmethod def constraint_batch_size(op): "IFM Tensor batch size must be 1" ifm = op.ifm valid = ifm.shape[0] == 1 extra = "Tensor '{}' has batch size: {}".format(ifm.name, ifm.shape[0]) return valid, extra @classmethod def check_depthwise_convolution_restrictions(cls, op): # check depth ifm_tensor, ofm_tensor = op.get_ifm_ofm() if op.attrs["depth_multiplier"] > 1 and not ( (ifm_tensor.shape[3] == 1) and (ofm_tensor.shape[3] == op.attrs["depth_multiplier"]) ): print( "Warning: for depth multipliers > 1,", "number of input channels must be 1 and number of output channels must be equal to depth multiplier.", "Placing on CPU", ) return False return True @classmethod def check_transpose_convolution_restrictions(cls, op): # check stride stride_h, stride_w = op.attrs["stride_h"], op.attrs["stride_w"] if stride_h != 2 or stride_w != 2: print("Warning: stride must be equal to 2, placing on CPU") return False # check output dimensions ifm_tensor, weight_tensor, _, ofm_tensor = op.get_ifm_weights_biases_ofm() ifm_h, ifm_w = ifm_tensor.shape[1], ifm_tensor.shape[2] ofm_h, ofm_w = ofm_tensor.shape[1], ofm_tensor.shape[2] if op.attrs["padding"] == b"SAME": if (ofm_h != ifm_h * stride_h) or (ofm_w != ifm_w * stride_w): print( "Warning: for", op.type, "using SAME padding, output dimensions must equal input dimensions multiplied by stride.", "Placing on CPU", ) return False elif op.attrs["padding"] == b"VALID": kernel_h, kernel_w = weight_tensor.shape[0], weight_tensor.shape[1] if (ofm_h != (ifm_h) * stride_h + max(kernel_h - stride_h, 0)) or ( ofm_w != (ifm_w) * stride_w + max(kernel_w - stride_w, 0) ): print( "Warning: for", op.type, "using VALID padding, output dimensions must equal input dimensions multiplied by stride,", "minus difference between kernel size and stride. Placing on CPU", ) return False return True @classmethod def check_pooling_restrictions(cls, op): # check stride stride_w, stride_h = op.attrs["stride_w"], op.attrs["stride_h"] if not is_integer(stride_w) or not is_integer(stride_h): print("Warning:", op.type, "has non-integer stride, placing on CPU") return False if not 1 <= stride_w <= 3 or not 1 <= stride_h <= 3: print( "Warning: {} has stride ({}, {}), only strides in range [1, 3] are allowed. Placing on CPU".format( op.type, stride_w, stride_h ) ) return False # check data type ifm_tensor, ofm_tensor = op.get_ifm_ofm() if ifm_tensor.dtype != ofm_tensor.dtype: if op.type != Op.ReduceSum: print("Warning: input data type doesn't match output data type, placing on CPU") return False # TODO: else check ReduceSum restrictions. # check batch size if ifm_tensor.shape[0] != 1: print("Warning: input batch size must be 1, placing on CPU") return False # check kernel size kernel_w, kernel_h = op.attrs["filter_width"], op.attrs["filter_height"] if op.type in cls.avg_pooling_ops and op.attrs["padding"] == b"SAME": if not 1 <= kernel_w <= 8 or not 1 <= kernel_h <= 8: print( "Warning:", op.type, "has kernel size ({}, {}), only kernel sizes in range [1, 8] are allowed. Placing on CPU".format( kernel_w, kernel_h ), ) return False if op.type in cls.avg_pooling_ops and op.attrs["padding"] == b"VALID" or op.type in cls.max_pooling_ops: if not 1 <= kernel_w * kernel_h <= 256 * 256: print( "Warning: product of kernel width and height must be >= 1 and not exceed 256 * 256 ({}),".format( 256 * 256 ), "placing on CPU", ) return False if not 1 <= kernel_h <= 256: print("Warning:", op.type, "has kernel height outside of range [1, 256], placing on CPU") return False return True @classmethod def check_resize_restrictions(cls, op): # check unsupported upscaling factor if op.type == Op.ResizeBilinear: if op.inputs[0].shape[1] == 1 and op.inputs[0].shape[2] == 1: return True if op.inputs[0].shape == op.outputs[0].shape: return True upscaled_shape = np.array(op.inputs[0].shape[1:3]) out_shape = np.array(op.outputs[0].shape[1:3]) while (upscaled_shape < out_shape).all(): upscaled_shape *= 2 if op.attrs["align_corners"]: upscaled_shape -= 1 if np.array_equal(out_shape, upscaled_shape): return True return False @classmethod def check_vector_product_restrictions(cls, op): # check data type ifm_tensor, _, weight_tensor, bias_tensor, _ = op.get_ifm_ifm2_weights_biases_ofm() if weight_tensor.element_size() > 1: print("Warning: only 8-bit datatypes supported for {}, placing on CPU".format(op.type)) return False if not cls.check_bias_restrictions(bias_tensor): return False # check non const weights if weight_tensor.values is None: print("Warning:", op.type, "has non-const weights, placing on CPU") return False return True @classmethod def check_element_wise_restrictions(cls, op): # check data type ifm_tensor, ifm2_tensor, _, ofm_tensor = op.get_ifm_ifm2_weights_ofm() # input and output datatype must match for these operators if ( op.type in cls.binary_elem_wise_min_max_ops | cls.unary_elem_wise_main_ops and ifm_tensor.dtype != ofm_tensor.dtype ): print("Warning:", op.type, "must have same input and output datatype, placing on CPU") return False if op.type in cls.binary_elem_wise_add_mul_sub: # both inputs must have same type if ifm_tensor.dtype != ifm2_tensor.dtype: print("Warning:", op.type, "must have same datatype on both inputs, placing on CPU") return False # signed input check if ifm_tensor.dtype.type & BaseType.Signed: # output must be signed if ofm_tensor.dtype.type & BaseType.Unsigned: print("Warning: only signed output types supported for {}, placing on CPU".format(op.type)) return False # and 8, 16 or 32-bit bit_lengths = {8, 16, 32} if ofm_tensor.element_size() * 8 not in bit_lengths: print( "Warning:", op.type, "is only supported for bit lengths {}, placing on CPU".format(bit_lengths) ) return False # unsigned input check, output must be same type or int32 if ifm_tensor.dtype.type & BaseType.Unsigned and not ( ifm_tensor.dtype == ofm_tensor.dtype or ofm_tensor.dtype == DataType.int32 ): print("Warning:", op.type, "has unsigned input but output is not unsigned or int32, placing on CPU") return False elif op.type in cls.binary_elem_wise_shift_ops: if ifm_tensor.dtype != DataType.int32 or ifm2_tensor.dtype != DataType.int32: print("Warning:", op.type, "input datatypes are not int32, placing on CPU") return False if op.type in (Op.CLZ, Op.SHL) and ofm_tensor.dtype != DataType.int32: print("Warning:", op.type, "output datatype is not int32, placing on CPU") return False # check batch size if len(ifm_tensor.shape) > 2 and ifm_tensor.shape[0] != 1: print( "Warning:", op.type, "only supports batch size 1 for tensors with more than 2 dimensions, placing on CPU", ) return False if op.type in cls.binary_elem_wise_main_ops: # if op type is unary, ifm2_tensor is None if len(ifm2_tensor.shape) > 2 and ifm2_tensor.shape[0] != 1: print( "Warning:", op.type, "only supports batch size 1 for tensors with more than 2 dimensions, placing on CPU", ) return False # negative alpha values are not supported if op.type == Op.LeakyRelu and op.attrs["alpha"] < 0: print("Warning:", op.type, "has negative alpha, placing on CPU") return False # check if ifm or ifm2 has ofm shape if ifm_tensor.shape != ofm_tensor.shape and ifm2_tensor.shape != ofm_tensor.shape: print("Warning:", op.type, "input shape(s) differ from output shape, placing on CPU") return False if op.type in cls.binary_elem_wise_min_max_ops and not cls.check_quantization_restrictions_binary_elem_wise(op): return False return True @classmethod def check_memory_only_restrictions(cls, op): if op.type == Op.StridedSlice: if len(op.inputs) != 4: warn_cpu(op, "has {} input tensors, only 4 inputs are supported".format(len(op.inputs))) return False input_tens, begin_tens, end_tens, strides_tens = op.inputs if begin_tens.values is None or end_tens.values is None or strides_tens.values is None: warn_cpu(op, "has a non-constant begin, end, or stride input tensor, which is not supported") return False if not ( len(input_tens.shape) == len(op.outputs[0].shape) == len(begin_tens.values) == len(end_tens.values) == len(strides_tens.values) ): warn_cpu(op, "has input tensors with shapes that are not supported") return False # check stride size if any(stride != 1 for stride in strides_tens.values): warn_cpu(op, "has stride values {}, only stride 1 values are supported".format(strides_tens.values)) return False # check ellipsis_mask if op.attrs["ellipsis_mask"] != 0: warn_cpu(op, "ellipsis_mask is {}, only 0 is supported".format(op.attrs["ellipsis_mask"])) return False # check if both new_axis_mask and shrink_axis_mask have bit set if op.attrs["new_axis_mask"] != 0 and op.attrs["shrink_axis_mask"] != 0: warn_cpu(op, "new_axis_mask and shrink_axis_mask are both non-zero, which is not supported") return False # Calculate offset start/end offset_start = get_slice_offsets(input_tens.shape, begin_tens, op.attrs["begin_mask"], is_begin=True) offset_end = get_slice_offsets(input_tens.shape, end_tens, op.attrs["end_mask"], is_begin=False) # check "end - begin" doesn't result in any zero or negative elements if any((end - begin) <= 0 for begin, end in zip(offset_start, offset_end)): warn_cpu( op, "has slice begin values {}, some of which are >= end values {}, which is illegal".format( begin_tens.values, end_tens.values ), ) return False if op.type == Op.SplitV: # check that maximum one size is set to -1, indicating that size should be inferred sizes = op.inputs[1].values num_to_be_inferred = 0 for size in sizes: if size == -1: num_to_be_inferred += 1 if num_to_be_inferred > 1: print("Warning:", op.type, "has more than one size to be inferred, which is illegal, placing on CPU") return False if op.type in set((Op.Concat, Op.ConcatTFLite,)): axis = op.attrs.get("axis", None) if axis is None: print("Warning:", op.type, "invalid or missing axis, placing on CPU") return False if axis < 0: axis += len(op.inputs[0].shape) if not 0 <= axis < len(op.inputs[0].shape): print("Warning:", op.type, "invalid axis", axis, ", placing on CPU") return False ofm = op.outputs[0] ofm_dims = len(ofm.shape) for ifm in op.inputs: if len(ifm.shape) != ofm_dims: return False for i in range(ofm_dims): if i != axis and ifm.shape[i] != ofm.shape[i]: print( "Warning:", op.type, "invalid ifm:", ifm.name, ifm.shape, "mismatch in dimension", i, ", placing on CPU", ) return False return True @classmethod def check_quantization_restrictions_binary_elem_wise(cls, op): # makes sure IFM1, IFM2 and OFM quantization are equal for binary ops assert len(op.inputs) >= 2 and len(op.outputs) == 1 if ( op.inputs[0].quantization is None or not op.inputs[0].is_scaling_equal(op.inputs[1]) or not op.inputs[0].is_scaling_equal(op.outputs[0]) ): print( "Warning: Input/output tensors with different quantization is unsupported for the", op.type, "operator" ) return False return True @classmethod def check_activation_ops(cls, op): if op.type == Op.Softmax: ifm_tensor = op.inputs[0] ofm_tensor = op.outputs[0] # check data type if ifm_tensor.dtype != ofm_tensor.dtype: print("Warning:", op.type, "input type differs from output type, placing on CPU") return False if ifm_tensor.dtype not in (DataType.uint8, DataType.int8, DataType.int16): print( "Warning: only datatypes supported for {} are uint8, int8 and int16; placing on CPU".format(op.type) ) return False # check shape if ifm_tensor.shape != ofm_tensor.shape: print("Warning:", op.type, "input shape differs from output shape, placing on CPU") return False return True @classmethod def check_bias_restrictions(cls, bias_tensor): # check data type if bias_tensor is not None and bias_tensor.dtype not in (DataType.int32, DataType.int64): print("Warning: bias tensor datatype must be int32 or int64, placing on CPU") return False # check if values fits in 40-bit if bias_tensor is not None and bias_tensor.dtype == DataType.int64: for quant_value in bias_tensor.quant_values: if not (-(1 << 39) <= quant_value < (1 << 39)): print("Warning: bias tensor values are larger than 40 bits, placing on CPU") return False return True