| /external/llvm-project/mlir/test/Dialect/Shape/ |
| D | ops.mlir | 7 func @shape_num_elements(%shape : !shape.shape) -> !shape.size {
8 %init = shape.const_size 1
9 %num_elements = shape.reduce(%shape, %init) : !shape.shape -> !shape.size {
10 ^bb0(%index : index, %extent : !shape.size, %acc : !shape.size):
11 %acc_next = shape.mul %acc, %extent
12 : !shape.size, !shape.size -> !shape.size
13 shape.yield %acc_next : !shape.size
15 return %num_elements : !shape.size
19 func @extent_tensor_num_elements(%shape : tensor<?xindex>) -> index {
21 %num_elements = shape.reduce(%shape, %init) : tensor<?xindex> -> index {
[all …]
|
| D | canonicalize.mlir | 5 // CHECK: shape.const_shape [2, 3, 4] : tensor<?xindex>
6 %0 = shape.shape_of %arg0 : tensor<2x3x4xf32> -> tensor<?xindex>
14 func @f() -> (!shape.shape, !shape.shape) {
15 // CHECK: shape.const_shape [2, 3] : !shape.shape
16 // CHECK: shape.const_shape [4, 5] : !shape.shape
18 %0 = shape.const_shape [2, 3, 4, 5] : !shape.shape
19 %head, %tail = "shape.split_at"(%0, %c2) : (!shape.shape, i32) -> (!shape.shape, !shape.shape)
20 return %head, %tail : !shape.shape, !shape.shape
28 func @f() -> (!shape.shape, !shape.shape) {
29 // CHECK: shape.const_shape [2, 3, 4] : !shape.shape
[all …]
|
| D | invalid.mlir | 3 func @reduce_op_args_num_mismatch(%shape : !shape.shape, %init : !shape.size) {
5 %num_elements = shape.reduce(%shape, %init) : !shape.shape -> !shape.size {
6 ^bb0(%index: index, %dim: !shape.size):
7 shape.yield %dim : !shape.size
14 func @reduce_op_arg0_wrong_type(%shape : !shape.shape, %init : !shape.size) {
16 %num_elements = shape.reduce(%shape, %init) : !shape.shape -> !shape.size {
17 ^bb0(%index: f32, %dim: !shape.size, %acc: !shape.size):
18 %new_acc = "shape.add"(%acc, %dim)
19 : (!shape.size, !shape.size) -> !shape.size
20 shape.yield %new_acc : !shape.size
[all …]
|
| D | remove-shape-constraints.mlir | 1 // RUN: mlir-opt -allow-unregistered-dialect -split-input-file -remove-shape-constraints -canonical…
2 // RUN: mlir-opt -allow-unregistered-dialect -split-input-file -remove-shape-constraints <%s | File…
8 func @f(%arg0 : !shape.shape, %arg1 : !shape.shape) -> index {
9 // REPLACE-NEXT: %[[WITNESS:.+]] = shape.const_witness true
10 // REPLACE-NOT: shape.cstr_eq
11 // REPLACE: shape.assuming %[[WITNESS]]
14 %0 = shape.cstr_broadcastable %arg0, %arg1 : !shape.shape, !shape.shape
15 %1 = shape.assuming %0 -> index {
17 shape.assuming_yield %2 : index
26 func @f(%arg0 : !shape.shape, %arg1 : !shape.shape) -> index {
[all …]
|
| D | shape-to-shape.mlir | 1 // RUN: mlir-opt -shape-to-shape-lowering -split-input-file %s | FileCheck %s
4 // CHECK-SAME: ([[ARG:%.*]]: !shape.shape) -> !shape.size
5 func @num_elements_to_reduce(%shape : !shape.shape) -> !shape.size {
6 %num_elements = shape.num_elements %shape : !shape.shape -> !shape.size
7 return %num_elements : !shape.size
9 // CHECK: [[C1:%.*]] = shape.const_size 1
10 // CHECK: [[NUM_ELEMENTS:%.*]] = shape.reduce([[ARG]], [[C1]]) : !shape.shape -> !shape.size
11 // CHECK: ^bb0({{.*}}: index, [[DIM:%.*]]: !shape.size, [[ACC:%.*]]: !shape.size
12 // CHECK: [[NEW_ACC:%.*]] = shape.mul [[DIM]], [[ACC]]
13 // CHECK: shape.yield [[NEW_ACC]] : !shape.size
[all …]
|
| /external/tensorflow/tensorflow/compiler/xla/service/ |
| D | shape_inference.h | 16 // Shape inference is used by the XLA service as the user builds up
35 // For a given operation and input shapes, infers what the resulting shape is
38 // the shape that results from an operation is inferred. Some methods have
39 // overloads for inferring shape at the HLO level.
41 // TODO(b/73352135): Shape inference does not issue very good error messages, in
42 // part because HloInstruction::ToString() is not available since shape
47 // Infers the shape produced by applying the given unary operation to the
48 // given input shape.
49 static StatusOr<Shape> InferUnaryOpShape(HloOpcode opcode,
50 const Shape& shape);
[all …]
|
| D | hlo_creation_utils.cc | 40 TF_ASSIGN_OR_RETURN(Shape unary_op_shape, in MakeUnaryHlo()
50 TF_ASSIGN_OR_RETURN(Shape binary_op_shape, in MakeBinaryHlo()
62 Shape binary_op_shape, in MakeCompareHlo()
74 Shape pad_shape, in MakePadHlo()
75 ShapeInference::InferPadShape(operand->shape(), padding_value->shape(), in MakePadHlo()
86 TF_ASSIGN_OR_RETURN(Shape slice_shape, ShapeInference::InferSliceShape( in MakeSliceHlo()
87 operand->shape(), start_indices, in MakeSliceHlo()
102 Shape convolve_shape, in MakeConvolveHlo()
104 lhs->shape(), rhs->shape(), feature_group_count, batch_group_count, in MakeConvolveHlo()
115 Shape transpose_shape, in MakeTransposeHlo()
[all …]
|
| /external/tensorflow/tensorflow/compiler/xla/ |
| D | shape_util.h | 32 #include "tensorflow/compiler/xla/shape.h"
49 // An index for specifying a particular nested subshape within a shape. Used in
53 // shape. For a non-nested tuple, an index has a single element. For example,
166 // Returns true if this shape index starts with 'prefix'.
179 // Namespaced collection of (static) shape utilities.
185 // Data structure which describes the coordinates and the shape, of a tuple
186 // shaped sub-shape.
189 IndexedShape(ShapeIndex index, Shape shape) in IndexedShape()
190 : index(std::move(index)), shape(std::move(shape)) {} in IndexedShape()
192 Shape shape; member
[all …]
|
| D | shape_util.cc | 118 // Constructs and returns the new shape with the given minor_to_major order in
120 StatusOr<Shape> MakeShapeWithLayoutInternal( in MakeShapeWithLayoutInternal()
132 TF_ASSIGN_OR_RETURN(Shape shape, in MakeShapeWithLayoutInternal()
139 *shape.mutable_layout() = LayoutUtil::MakeLayout( in MakeShapeWithLayoutInternal()
141 if (!shape.has_layout()) { in MakeShapeWithLayoutInternal()
142 return InvalidArgument("Shape has no layout."); in MakeShapeWithLayoutInternal()
144 TF_RETURN_IF_ERROR(ShapeUtil::ValidateShape(shape)); in MakeShapeWithLayoutInternal()
145 return shape; in MakeShapeWithLayoutInternal()
149 /* static */ bool ShapeUtil::Equal(const Shape& lhs, const Shape& rhs) { in Equal()
150 bool equal = Shape::Equal()(lhs, rhs); in Equal()
[all …]
|
| D | layout_util_test.cc | 29 Shape MakeShapeWithLayout(PrimitiveType element_type, in MakeShapeWithLayout()
32 Shape shape = ShapeUtil::MakeShape(element_type, dimensions); in MakeShapeWithLayout() local
33 *shape.mutable_layout() = LayoutUtil::MakeLayout(minor_to_major); in MakeShapeWithLayout()
34 return shape; in MakeShapeWithLayout()
39 Shape shape = in TEST_F() local
41 Shape other_shape = in TEST_F()
44 Shape tuple0 = ShapeUtil::MakeTupleShape({}); in TEST_F()
45 Shape tuple1 = ShapeUtil::MakeTupleShape({shape}); in TEST_F()
46 Shape tuple2 = ShapeUtil::MakeTupleShape({shape, shape}); in TEST_F()
58 Shape other_tuple2 = ShapeUtil::MakeTupleShape({shape, other_shape}); in TEST_F()
[all …]
|
| /external/gemmlowp/test/ |
| D | benchmark_meta_gemm.cc | 64 struct Shape { struct
73 Shape(std::int32_t n, std::int32_t m, std::int32_t k) in Shape() argument
104 double run_gemms(std::vector<Shape>* shapes) { in run_gemms() argument
106 for (auto& shape : *shapes) { in run_gemms()
107 ops += run_gemm(shape.n, shape.m, shape.k, shape.working_set().lhs, in run_gemms()
108 shape.working_set().rhs, shape.working_set().result); in run_gemms()
159 void time_all(std::vector<Shape>* shapes, std::int32_t repetitions, in time_all()
179 void time_one(Shape* shape, double max_time) { in time_one() argument
184 std::cout << std::setprecision(6) << std::fixed << shape->n << ", " in time_one()
185 << shape->m << ", " << shape->k << ", " << std::flush; in time_one()
[all …]
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/common/ |
| D | convert.cc | 28 #include "tensorflow/lite/delegates/gpu/common/shape.h"
43 absl::Status ConvertToPHWO4I4(absl::Span<const float> in, const OHWI& shape, in ConvertToPHWO4I4() argument
45 if (in.size() != shape.DimensionsProduct()) { in ConvertToPHWO4I4()
48 in.size(), " != ", shape.DimensionsProduct())); in ConvertToPHWO4I4()
50 if (out.size() != GetElementsSizeForPHWO4I4(shape)) { in ConvertToPHWO4I4()
53 out.size(), " != ", GetElementsSizeForPHWO4I4(shape))); in ConvertToPHWO4I4()
57 for (int p = 0; p < DivideRoundUp(shape.o, kPhwo4i4ChannelsInPlane); ++p) { in ConvertToPHWO4I4()
58 for (int h = 0; h < shape.h; ++h) { in ConvertToPHWO4I4()
59 for (int w = 0; w < shape.w; ++w) { in ConvertToPHWO4I4()
60 for (int c = 0; c < DivideRoundUp(shape.i, kPhwo4i4ChannelsInPlane); in ConvertToPHWO4I4()
[all …]
|
| /external/llvm-project/mlir/docs/ |
| D | ShapeInference.md | 1 # Shape Inference
3 Shape inference as discussed here is considered a specific instance of type
6 dimensions. While some operations have no compile time fixed shape (e.g., output
7 shape is dictated by data) we could still have some knowledge of
11 shape.
15 `InferShapedTypeOpInterface` is used to implement the shape and element type
16 inference. The return type can often be deduced from the deduced return shape
20 ## Shape functions
22 The C++ interfaces are the base mechanism whereby shape inference is queried and
23 executed, but not the intended way to specify shape constraints in general.
[all …]
|
| /external/llvm-project/mlir/include/mlir/Dialect/Shape/IR/ |
| D | ShapeOps.td | 1 //===- Shape.td - Shape operations definition --------------*- tablegen -*-===//
9 // This is the operation definition file for Shape dialect operations.
16 include "mlir/Dialect/Shape/IR/ShapeBase.td"
24 // Shape op definitions
51 let summary = "Returns the broadcasted output shape of two inputs";
53 Returns the broadcasted shape for two input shapes or extent tensors. Both
54 operands can be of type `shape.shape` or `tensor<?xindex>`. The result is of
55 type `shape.shape` and, if both operands are tensors, may be of type
59 with 1's from the left. The resulting broadcasted shape is then defined as
65 In case the resulting shape is undefined, i.e. if corresponding extents are
[all …]
|
| D | ShapeBase.td | 9 // Base definitions for the `shape` dialect.
19 // Shape Inference dialect definitions
23 let name = "shape";
25 let summary = "Types and operations for shape dialect";
27 This dialect contains operations for shape inference.
29 Note: Unless explicitly stated, all functions that return a shape and take
30 shapes as input, return the invalid shape if one of its operands is an
31 invalid shape. This avoids flagging multiple errors for one verification
37 let cppNamespace = "::mlir::shape";
43 CPred<"$_self.isa<::mlir::shape::ComponentType>()">, "component type">,
[all …]
|
| /external/tensorflow/tensorflow/java/src/main/java/org/tensorflow/ |
| D | Shape.java | 20 /** The possibly partially known shape of a tensor produced by an operation. */
21 public final class Shape { class
23 /** Create a Shape representing an unknown number of dimensions. */
24 public static Shape unknown() { in unknown()
25 return new Shape(null); in unknown()
28 /** Create a Shape representing a scalar value. */
29 public static Shape scalar() { in scalar()
30 return new Shape(new long[0]); in scalar()
34 * Create a Shape representing an N-dimensional value.
36 * <p>Creates a Shape representing an N-dimensional value (N being at least 1), with the provided
[all …]
|
| /external/tensorflow/tensorflow/compiler/mlir/xla/ir/ |
| D | mlir_hlo_builder.h | 30 #include "tensorflow/compiler/xla/shape.h"
101 // Returns the shape of the given op.
102 StatusOr<const Shape*> GetShapePtr(XlaOp op) const override;
114 const Shape& shape, XlaOp lhs, XlaOp rhs, const Window& window,
123 StatusOr<XlaOp> FftInternal(const Shape& shape, XlaOp operand,
128 const Shape& shape, XlaOp a, XlaOp b,
131 StatusOr<XlaOp> CholeskyInternal(const Shape& shape, XlaOp a,
136 const Shape& shape, const string& opaque,
137 absl::optional<absl::Span<const Shape>> operand_shapes_with_layout,
144 const Shape& shape, absl::Span<const XlaOp> all_operands,
[all …]
|
| /external/tensorflow/tensorflow/python/ops/ |
| D | init_ops_v2_test.py | 40 shape=None, argument
42 if shape is None:
43 shape = [100]
44 t1 = self.evaluate(init1(shape, dtype))
45 t2 = self.evaluate(init2(shape, dtype))
46 self.assertEqual(tensor_shape.as_shape(shape), t1.shape)
47 self.assertEqual(tensor_shape.as_shape(shape), t2.shape)
50 def _duplicated_test(self, init, shape=None, dtype=dtypes.float32): argument
51 if shape is None:
52 shape = [100]
[all …]
|
| D | init_ops_v2.py | 46 def __call__(self, shape, dtype=None, **kwargs):
47 # returns a tensor of shape `shape` and dtype `dtype`
52 def __call__(self, shape, dtype=None, **kwargs): argument
56 shape: Shape of the tensor.
61 partition in a partitioned variable. `partition_shape` is the shape of
62 the partition (i.e. the shape of the returned tensor) and
64 partition w.r.t each axis. For example, a tensor of shape `(30, 100)`
65 can be partitioned into two partitions: `p0` of shape `(10, 100)` and
66 `p1` of shape `(20, 100)`; if the initializer is called with
116 the Initializer object, without knowing the shape and dtype of the variable
[all …]
|
| D | random_ops.py | 47 def random_normal(shape, argument
59 <tf.Tensor: shape=(4,), dtype=float32, numpy=..., dtype=float32)>
65 <tf.Tensor: shape=(2, 2), dtype=float32, numpy=
74 shape: A 1-D integer Tensor or Python array. The shape of the output tensor.
87 A tensor of the specified shape filled with random normal values.
89 with ops.name_scope(name, "random_normal", [shape, mean, stddev]) as name:
90 shape_tensor = tensor_util.shape_tensor(shape)
98 tensor_util.maybe_set_static_shape(value, shape)
105 def parameterized_truncated_normal(shape, argument
120 shape: A 1-D integer Tensor or Python array. The shape of the output tensor.
[all …]
|
| D | stateless_random_ops.py | 66 [ 10670227 -246211131]], shape=(3, 2), dtype=int32)
67 >>> tf.random.stateless_normal(shape=[3], seed=new_seeds[0, :])
68 <tf.Tensor: shape=(3,), dtype=float32, numpy=array([-0.59835213, -0.9578608 ,
72 seed: an RNG seed (a tensor with shape [2] and dtype `int32` or
78 A tensor with shape [num, 2] representing `num` new seeds. It will have the
83 return stateless_random_uniform(shape=[num, 2], seed=seed, dtype=seed.dtype,
103 tf.Tensor([1105988140 3], shape=(2,), dtype=int32)
104 >>> tf.random.stateless_normal(shape=[3], seed=replica_seed)
105 <tf.Tensor: shape=(3,), dtype=float32, numpy=array([0.03197195, 0.8979765 ,
109 seed: an RNG seed (a tensor with shape [2] and dtype `int32` or
[all …]
|
| /external/tensorflow/tensorflow/python/grappler/ |
| D | datasets_test.py | 15 """Tests for the datasets shape inference."""
39 'shape': tensor_shape.TensorShape([])
42 'shape': tensor_shape.TensorShape([3])
45 'shape': tensor_shape.TensorShape([1, 3])
58 self.assertEqual(test_case['shape'],
59 op_properties['IteratorGetNext'][0].shape)
64 'shape': tensor_shape.TensorShape([])
67 'shape': tensor_shape.TensorShape([3])
70 'shape': tensor_shape.TensorShape([1, 3])
83 self.assertEqual(test_case['shape'],
[all …]
|
| /external/tensorflow/tensorflow/python/keras/engine/ |
| D | input_spec.py | 35 """Specifies the rank, dtype and shape of every input to a layer.
40 compatibility checks for input structure, input rank, input shape, and
43 A None entry in a shape is compatible with any dimension,
44 a None shape is compatible with any shape.
48 shape: Shape tuple, expected shape of the input
67 # The layer will accept inputs with shape (?, 28, 28) & (?, 28, 28, 1)
70 shape=(None, 28, 28, 1),
77 shape=None, argument
85 shape = tensor_shape.TensorShape(shape)
86 if shape.rank is None:
[all …]
|
| /external/tflite-support/tensorflow_lite_support/java/src/java/org/tensorflow/lite/support/tensorbuffer/ |
| D | TensorBuffer.java | 33 /** Shape of the tensor stored in this buffer. */
34 protected int[] shape; field in TensorBuffer
46 * Creates a {@link TensorBuffer} with specified {@code shape} and {@link DataType}. Here are some
50 * Creating a float TensorBuffer with shape {2, 3}:
51 * int[] shape = new int[] {2, 3};
52 * TensorBuffer tensorBuffer = TensorBuffer.createFixedSize(shape, DataType.FLOAT32);
57 * int[] shape = new int[] {};
58 * TensorBuffer tensorBuffer = TensorBuffer.createFixedSize(shape, DataType.UINT8);
63 * int[] shape = new int[] {0};
64 * TensorBuffer tensorBuffer = TensorBuffer.createFixedSize(shape, DataType.UINT8);
[all …]
|
| /external/tensorflow/tensorflow/python/tpu/ |
| D | tpu_sharding.py | 164 def get_unpartitioned_shape(self, shape): argument
165 """Returns the shape of an unpartitioned Tensor.
167 When given the shape of a 'sharded-size' Tensor, returns the shape
168 of the full shape of its unpartitioned Tensor.
171 shape: The shape of the sharded Tensor.
174 The shape of the unpartitioned version of the Tensor.
177 ValueError: if shape has unknown sharded dimension
179 shape = tensor_shape.as_shape(shape)
180 dims = shape.as_list()
185 raise ValueError("shape %s must have a fixed size for dimension %d "
[all …]
|