Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Relay/TOPI][TFLite] Implemented MATRIX_SET_DIAG Operator for Relay/TOPI and TFLite Frontend. #6303

Merged
merged 4 commits into from
Aug 27, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 29 additions & 0 deletions include/tvm/topi/transform.h
Original file line number Diff line number Diff line change
Expand Up @@ -1511,6 +1511,35 @@ inline Tensor sparse_to_dense(const Tensor& sparse_indices, const Array<Integer>
name, tag);
}

/*!
* \brief Returns a tensor with the diagonal of input tensor replaced with the provided diagonal.
* \param input input tensor.
* \param diagonal values to be filled in the diagonal.
* \param name output tensor name.
* \param tag output tensor tag.
* \return new tensor with given diagonal values.
*/
inline Tensor matrix_set_diag(const Tensor& input, const Tensor& diagonal,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A suggestion:- may be if we can support alignment and k(offset) similar to MatrixSetDiagV3 in tf, it will be good. we can support directly for tensorflow ops as well.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That might take some time.
Would it be fine to have that in a follow-up PR?

const std::string name = "T_matrix_set_diag",
const std::string tag = kInjective) {
size_t ndim = input->shape.size() - 1;

return compute(
input->shape,
[&](const Array<Var>& iter_vars) {
auto get_diag = [&]() {
Array<PrimExpr> diagonal_indices;
for (size_t i = 0; i < ndim; i++) {
diagonal_indices.push_back(iter_vars[i]);
}
return diagonal(diagonal_indices);
};
return if_then_else((PrimExpr)iter_vars[ndim] == iter_vars[ndim - 1], get_diag(),
input(iter_vars));
},
name, tag);
}

} // namespace topi
} // namespace tvm
#endif // TVM_TOPI_TRANSFORM_H_
28 changes: 28 additions & 0 deletions python/tvm/relay/frontend/tflite.py
Original file line number Diff line number Diff line change
Expand Up @@ -107,6 +107,7 @@ def __init__(self, model, subgraph, exp_tab):
'LOGICAL_NOT': self.convert_logical_not,
'LOGICAL_OR': self.convert_logical_or,
'LOGISTIC': self.convert_logistic,
'MATRIX_SET_DIAG': self.convert_matrix_set_diag,
'MAX_POOL_2D': self.convert_max_pool2d,
'MAXIMUM': self.convert_maximum,
'MEAN': self.convert_reduce_mean,
Expand Down Expand Up @@ -2989,6 +2990,33 @@ def convert_reverse_v2(self, op):
out = _op.reverse(input_expr, axis)
return out

def convert_matrix_set_diag(self, op):
"""Convert TFLite MATRIX_SET_DIAG"""

input_tensors = self.get_input_tensors(op)
assert len(input_tensors) == 2, "input tensor's length should be 2"

assert input_tensors[0].tensor.Type() == input_tensors[1].tensor.Type(), \
"input and diagonal should be the same type of tensors"

if input_tensors[0].qnn_params:
# Check that input and output tensor have same qnn params.
output_tensors = self.get_output_tensors(op)
assert self.has_same_qnn_params(input_tensors[0], output_tensors[0]), \
"TFLite MATRIX_SET_DIAG requires input and output tensors' \
scale and zero points to be equal"

# Check that input and diagonal tensor have same qnn params.
assert self.has_same_qnn_params(input_tensors[0], input_tensors[1]), \
"TFLite MATRIX_SET_DIAG requires input and diagonal tensors' \
scale and zero points to be equal"

input_expr = self.get_tensor_expr(input_tensors[0])
diagonal_expr = self.get_tensor_expr(input_tensors[1])

out = _op.matrix_set_diag(input_expr, diagonal_expr)
return out


def get_expr(self, input_tensor_idx):
return self.exp_tab.get_expr(get_tensor_name(self.subgraph, input_tensor_idx))
Expand Down
1 change: 1 addition & 0 deletions python/tvm/relay/op/_transform.py
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,7 @@
_reg.register_reduce_schedule("collapse_sum_to")
_reg.register_injective_schedule("unravel_index")
_reg.register_injective_schedule("sparse_to_dense")
_reg.register_injective_schedule("matrix_set_diag")

# concatenate
_reg.register_schedule("concatenate", strategy.schedule_concatenate)
Expand Down
41 changes: 41 additions & 0 deletions python/tvm/relay/op/transform.py
Original file line number Diff line number Diff line change
Expand Up @@ -1167,3 +1167,44 @@ def sparse_to_dense(sparse_indices, output_shape, sparse_values, default_value=0
if default_value == 0:
default_value = const(0)
return _make.sparse_to_dense(sparse_indices, output_shape, sparse_values, default_value)


def matrix_set_diag(data, diagonal):
"""
Returns a tensor with the diagonal of input tensor replaced with the provided diagonal values.

Parameters
----------
data : relay.Expr
Input Tensor.
diagonal : relay.Expr
Values to be filled in the diagonal.

Returns
-------
result : relay.Expr
New tensor with given diagonal values.

Examples
--------
.. code-block:: python

data = [[[7, 7, 7, 7],
[7, 7, 7, 7],
[7, 7, 7, 7]],
[[7, 7, 7, 7],
[7, 7, 7, 7],
[7, 7, 7, 7]]]

diagonal = [[1, 2, 3],
[4, 5, 6]]

relay.matrix_set_diag(input, diagonal) =
[[[1, 7, 7, 7],
[7, 2, 7, 7],
[7, 7, 3, 7]],
[[4, 7, 7, 7],
[7, 5, 7, 7],
[7, 7, 6, 7]]]
"""
return _make.matrix_set_diag(data, diagonal)
1 change: 1 addition & 0 deletions python/tvm/topi/testing/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -60,3 +60,4 @@
get_elemwise_schedule, get_conv2d_nchw_implement, dispatch
from .adaptive_pool_python import adaptive_pool
from .grid_sample_python import affine_grid_python, grid_sample_nchw_python
from .matrix_set_diag import matrix_set_diag
47 changes: 47 additions & 0 deletions python/tvm/topi/testing/matrix_set_diag.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,47 @@
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you 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.
# pylint: disable=invalid-name
"""MatrixSetDiag in Python"""
import numpy as np

def matrix_set_diag(input_np, diagonal):
"""matrix_set_diag operator implemented in numpy.

Returns a numpy array with the diagonal of input array
replaced with the provided diagonal values.

Parameters
----------
input : numpy.ndarray
Input Array.
Shape = [D1, D2, D3, ... , Dn-1 , Dn]
diagonal : numpy.ndarray
Values to be filled in the diagonal.
Shape = [D1, D2, D3, ... , Dn-1]

Returns
-------
result : numpy.ndarray
New Array with given diagonal values.
Shape = [D1, D2, D3, ... , Dn-1 , Dn]
"""
out = np.array(input_np, copy=True)
n = min(input_np.shape[-1], input_np.shape[-2])
for i in range(n):
out[..., i, i] = diagonal[..., i]

return out
40 changes: 40 additions & 0 deletions python/tvm/topi/transform.py
Original file line number Diff line number Diff line change
Expand Up @@ -798,3 +798,43 @@ def sparse_to_dense(sparse_indices, output_shape, sparse_values, default_value=0
"""

return cpp.sparse_to_dense(sparse_indices, output_shape, sparse_values, default_value)

def matrix_set_diag(data, diagonal):
"""
Returns a tensor with the diagonal of input tensor replaced with the provided diagonal values.

Parameters
----------
data : relay.Expr
Input Tensor.
diagonal : relay.Expr
Values to be filled in the diagonal.

Returns
-------
result : relay.Expr
New tensor with given diagonal values.

Examples
--------
.. code-block:: python

data = [[[7, 7, 7, 7],
[7, 7, 7, 7],
[7, 7, 7, 7]],
[[7, 7, 7, 7],
[7, 7, 7, 7],
[7, 7, 7, 7]]]

diagonal = [[1, 2, 3],
[4, 5, 6]]

relay.matrix_set_diag(input, diagonal) =
[[[1, 7, 7, 7],
[7, 2, 7, 7],
[7, 7, 3, 7]],
[[4, 7, 7, 7],
[7, 5, 7, 7],
[7, 7, 6, 7]]]
"""
return cpp.matrix_set_diag(data, diagonal)
50 changes: 50 additions & 0 deletions src/relay/op/tensor/transform.cc
Original file line number Diff line number Diff line change
Expand Up @@ -3093,5 +3093,55 @@ RELAY_REGISTER_OP("sparse_to_dense")
.set_attr<FInferCorrectLayout>("FInferCorrectLayout", ElemwiseArbitraryLayout)
.set_attr<FTVMCompute>("FTVMCompute", SparseToDenseCompute);

// relay.matrix_set_diag
bool MatrixSetDiagRel(const Array<Type>& types, int num_inputs, const Attrs& attrs,
const TypeReporter& reporter) {
// `types` contains: [input, diagonal, result]
CHECK_EQ(types.size(), 3);

const auto* input = types[0].as<TensorTypeNode>();
CHECK(input);

const auto* diagonal = types[1].as<TensorTypeNode>();
CHECK(diagonal);

int d_ndims = diagonal->shape.size();
for (int i = 0; i < d_ndims - 1; i++) {
reporter->AssertEQ(input->shape[i], diagonal->shape[i]);
}
auto min_dim = if_then_else(input->shape[d_ndims - 1] >= input->shape[d_ndims],
input->shape[d_ndims], input->shape[d_ndims - 1]);
Comment on lines +3112 to +3113
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is if_then_else appropriate here? Could a ? x : y not be used?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for reviewing.
if_then_else is needed here because input->shape[i] is a PrimExpr, and so a ? x : y can't be used.

reporter->Assert(diagonal->shape[d_ndims - 1] >= min_dim);

reporter->Assign(types[2], TensorType(input->shape, input->dtype));
return true;
}

Array<te::Tensor> MatrixSetDiagCompute(const Attrs& attrs, const Array<te::Tensor>& inputs,
const Type& out_type) {
return Array<te::Tensor>{topi::matrix_set_diag(inputs[0], inputs[1])};
}

Expr MakeMatrixSetDiag(Expr input, Expr diagonal) {
static const Op& op = Op::Get("matrix_set_diag");
return Call(op, {input, diagonal}, Attrs(), {});
}

TVM_REGISTER_GLOBAL("relay.op._make.matrix_set_diag").set_body_typed(MakeMatrixSetDiag);

RELAY_REGISTER_OP("matrix_set_diag")
.describe(
R"code(Returns a tensor with the diagonal of input tensor replaced with the provided diagonal values.
**input** Input tensor.
**diagonal** Values to be filled in the diagonal.
)code" TVM_ADD_FILELINE)
.set_num_inputs(2)
.add_argument("input", "Tensor", "Input Tensor.")
.add_argument("diagonal", "Tensor", "Values to be filled in the diagonal.")
.set_support_level(10)
.add_type_rel("MatrixSetDiag", MatrixSetDiagRel)
.set_attr<FTVMCompute>("FTVMCompute", MatrixSetDiagCompute)
.set_attr<TOpPattern>("TOpPattern", kInjective);

} // namespace relay
} // namespace tvm
4 changes: 4 additions & 0 deletions src/topi/transform.cc
Original file line number Diff line number Diff line change
Expand Up @@ -176,5 +176,9 @@ TVM_REGISTER_GLOBAL("topi.one_hot").set_body([](TVMArgs args, TVMRetValue* rv) {
*rv = one_hot(args[0], args[1], args[2], depth, axis, dtype);
});

TVM_REGISTER_GLOBAL("topi.matrix_set_diag").set_body([](TVMArgs args, TVMRetValue* rv) {
*rv = matrix_set_diag(args[0], args[1]);
});

} // namespace topi
} // namespace tvm
72 changes: 72 additions & 0 deletions tests/python/frontend/tflite/test_forward.py
Original file line number Diff line number Diff line change
Expand Up @@ -2652,6 +2652,77 @@ def test_forward_reverse_v2():
_test_reverse_v2((5, 6, 4, 2), np.array([2], dtype='int32'), dtype)


#######################################################################
# MATRIX_SET_DIAG
# ---------------

def _test_matrix_set_diag(input_shape, input_type, quantized=False):
""" One iteration of MATRIX_SET_DIAG """
with tf.Graph().as_default():
diagonal_shape = list(input_shape[:-2])
diagonal_shape.append(min(input_shape[-2], input_shape[-1]))
Comment on lines +2662 to +2663
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should the broadcasting case be tested here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TFLite MATRIX_SET_DIAG doesn't seem to be a broadcast operator. So, I'll change the registration to be injective.


if quantized:
# ignoring input_type as quantized requires uint8
input = np.random.uniform(0, 256, input_shape).astype('uint8')
in_input = tf.placeholder(dtype='float32', shape=input.shape, name="input")
inq_input = tf.quantization.fake_quant_with_min_max_args(
in_input,
min=-100,
max=100,
name="q_input")

diagonal = np.random.uniform(0, 256, diagonal_shape).astype('uint8')
in_diagonal = tf.placeholder(dtype='float32', shape=diagonal.shape, name="diagonal")
inq_diagonal = tf.quantization.fake_quant_with_min_max_args(
in_diagonal,
min=-100,
max=100,
name="q_diagonal")

input_range = {'q_input': (-100, 100), 'q_diagonal': (-100, 100)}

out = array_ops.matrix_set_diag(inq_input, inq_diagonal)
out = tf.quantization.fake_quant_with_min_max_args(
out,
min=-100,
max=100,
name="out")

compare_tflite_with_tvm(
[input, diagonal],
["q_input", "q_diagonal"],
[inq_input, inq_diagonal],
[out],
quantized=True,
input_range=input_range)
else:
input = np.random.uniform(0, 100, input_shape).astype(input_type)
diagonal = np.random.uniform(0, 100, diagonal_shape).astype(input_type)

in_input = tf.placeholder(dtype=input.dtype, shape=input.shape, name="input")
in_diagonal = tf.placeholder(dtype=diagonal.dtype, shape=diagonal.shape, name="diagonal")

out = array_ops.matrix_set_diag(in_input, in_diagonal)

compare_tflite_with_tvm(
[input, diagonal],
["input", "diagonal"],
[in_input, in_diagonal],
[out])

def test_forward_matrix_set_diag():
""" MATRIX_SET_DIAG """
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

add a pkg version check > '1.14.0'

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The API docs seem to suggest that matrix_set_diag is present even in version '1.0'.
So, is there some other reason to add this check?

for dtype in [np.float32, np.int32]:
_test_matrix_set_diag((4, 4), dtype)
_test_matrix_set_diag((5, 4, 3, 4), dtype)
_test_matrix_set_diag((4, 4, 2), dtype)

_test_matrix_set_diag((4, 4), np.uint8, quantized=True)
_test_matrix_set_diag((5, 4, 3, 4), np.uint8, quantized=True)
_test_matrix_set_diag((4, 4, 2), np.uint8, quantized=True)


#######################################################################
# Custom Operators
# ----------------
Expand Down Expand Up @@ -3131,6 +3202,7 @@ def test_forward_mediapipe_hand_landmark():
test_forward_arg_min_max()
test_forward_expand_dims()
test_forward_reverse_v2()
test_forward_matrix_set_diag()

# NN
test_forward_convolution()
Expand Down
Loading