tvm.topi

TVM Operator Inventory.

TOPI is the operator collection library for TVM, to provide sugars for constructing compute declaration as well as optimized schedules.

Some of the schedule function may have been specially optimized for a specific workload.

Classes:

Analyzer()	Integer arithmetic analyzer
Cast(dtype, value[, span])	Cast expression.
PrimExpr()	Base class of all primitive expressions.

Exceptions:

InvalidShapeError

Invalid shape for a topi function.

Functions:

abs(x)	Take absolute value of the input of x, element-wise.
acos(x)	Take arc cos of input x.
acosh(x)	Take arc cosh of input x.
add(lhs, rhs)	Addition with auto-broadcasting
adv_index(data, indices)	Numpy style indexing with tensors.
all(data[, axis, keepdims])	Logical AND of array elements over a given axis or a list of axes
any(data[, axis, keepdims])	Logical OR of array elements over a given axis or a list of axes
arange(start[, stop, step, dtype])	Creates a tensor with evenly spaced values within a given interval.
argmax(data[, axis, keepdims, select_last_index])	Returns the indices of the maximum values along an axis.
argmin(data[, axis, keepdims, select_last_index])	Returns the indices of the minimum values along an axis.
argsort(data[, valid_count, axis, ...])	Performs sorting along the given axis and returns an array of indices having the same shape as an input array that index data in sorted order.
argwhere(output_shape, condition)	Find the indices of elements of a tensor that are non-zero.
asin(x)	Take arc sin of input x.
asinh(x)	Take arc sinh of input x.
atan(x)	Take atan of input x.
atanh(x)	Take atanh of input x.
binary_search(ib, sequence_offset, ...)	Common IR generator for binary search used by CPU and GPU backends.
bitwise_and(lhs, rhs)	Compute element-wise bitwise and of data.
bitwise_not(data)	Compute element-wise bitwise not of data.
bitwise_or(lhs, rhs)	Compute element-wise bitwise or of data.
bitwise_xor(lhs, rhs)	Compute element-wise bitwise xor of data.
broadcast_to(data, shape)	Broadcast the src to the target shape
cast(x, dtype[, span])	Cast input to specified data type.
ceil(x)	Take ceil of input x.
ceil_log2(x)	Compute integer ceil log2 with a special code path for vulkan SPIR-V does not support log2 on fp64.
clip(x, a_min, a_max)	Clip (limit) the values in an array.
collapse_sum(data, target_shape)	Return a summation of data to the given shape.
concatenate(a_tuple[, axis])	Join a sequence of arrays along an existing axis.
const_vector(vector[, name])	convert a const numpy 1-dimensional vector to tvm tensor
cos(x)	Take cos of input x.
cosh(x)	Take cosh of input x.
cumprod(data[, axis, dtype, exclusive])	Numpy style cumprod op.
cumsum(data[, axis, dtype, exclusive])	Numpy style cumsum op.
decl_buffer(shape[, dtype, name, data, ...])	Declare a new symbolic buffer.
dft(re_data, im_data, inverse)	Computes the discrete Fourier transform of input (calculation along the last axis).
div(a, b[, span])	Compute a / b as in C/C++ semantics.
divide(lhs, rhs)	Division with auto-broadcasting
dynamic_strided_slice(a, begin, end, ...)	Slice of an array.
einsum(subscripts, *operand)	Evaluates the Einstein summation convention on the operands.
elemwise_sum(xs)	Perform element-wise sum on inputs
equal(lhs, rhs)	Compute (lhs==rhs) with auto-broadcasting
erf(x)	Take gauss error function of input x.
erf_legalize(attrs, inputs, types)	Legalizes ERF op.
exp(x)	Take exponential of input x.
expand_dims(a, axis[, num_newaxis])	Expand the shape of an array.
expand_like(a, shape_like, axis)	Expand an input array with the shape of second array.
extern(shape, inputs, fcompute[, name, ...])	Compute several tensors via an extern function.
eye(n[, m, k, dtype])	Generate an identity matrix or a matrix with ones on the k-th diagonal.
fast_erf(x)	Take gauss error function of input x using fast_erf implementation.
fast_exp(x)	Take exponential of input x using fast_exp implementation
fast_tanh(x)	Take hyperbolic tangent of input x using fast_tanh implementation
fixed_point_multiply(x, multiplier, shift)	Fixed point multiplication between data and a fixed point constant expressed as multiplier * 2^(-shift), where multiplier is a Q-number with 31 fractional bits
fixed_point_multiply_per_axis(x, y, lshift, ...)	Fixed point multiplication between data and a fixed point constant expressed as multiplier * 2^(-shift), where multiplier is a Q-number with 31 fractional bits
flip(a[, axis])	Flip/reverse elements of an array in a particular axis.
floor(x)	Take floor of input x.
floor_divide(lhs, rhs)	Floor division with auto-broadcasting
floor_mod(lhs, rhs)	Floor modulus with auto-broadcasting
floordiv(a, b[, span])	Compute the floordiv of two expressions.
floormod(a, b[, span])	Compute the floormod of two expressions.
full(shape, dtype, fill_value)	Fill tensor with fill_value
full_like(x, fill_value)	Construct a tensor with same shape as input tensor,
gather(data, axis, indices)	Gather values along given axis from given indices.
gather_nd(a, indices[, batch_dims])	Gather elements from a n-dimension array..
get_const_tuple(in_tuple)	Verifies input tuple is IntImm or Var, returns tuple of int or Var.
greater(lhs, rhs)	Compute (lhs>rhs) with auto-broadcasting
greater_equal(lhs, rhs)	Compute (lhs>=rhs) with auto-broadcasting
hybrid_argwhere_1d(output_shape, condition)	Find the indices of elements of a 1-D tensor that are non-zero.
hybrid_argwhere_2d(output_shape, condition)	Find the indices of elements of a 2-D tensor that are non-zero.
hybrid_argwhere_3d(output_shape, condition)	Find the indices of elements of a 3-D tensor that are non-zero.
hybrid_argwhere_4d(output_shape, condition)	Find the indices of elements of a 4-D tensor that are non-zero.
hybrid_argwhere_5d(output_shape, condition)	Find the indices of elements of a 5-D tensor that are non-zero.
identity(x)	Take identity of input x.
invert_permutation(data)	Computes the inverse permutation of data.
isfinite(x)	Check if value of x is finite, element-wise.
isinf(x)	Check if value of x is infinite, element-wise.
isnan(x)	Check if value of x is NaN, element-wise.
layout_transform(array, src_layout, dst_layout)	Transform the layout according to src_layout and dst_layout
left_shift(lhs, rhs)	Left shift with auto-broadcasting
less(lhs, rhs)	Compute (lhs<rhs) with auto-broadcasting
less_equal(lhs, rhs)	Compute (lhs<=rhs) with auto-broadcasting
log(x)	Take logarithm of input x.
log10(x)	Take logarithm to the base 10 of input x.
log2(x)	Take logarithm to the base 2 of input x.
logical_and(lhs, rhs)	Compute element-wise logical and of data.
logical_not(data)	Compute element-wise logical not of data.
logical_or(lhs, rhs)	Compute element-wise logical or of data.
logical_xor(lhs, rhs)	Compute element-wise logical xor of data.
make_idx(b, e, s, z, i)	Return the array position in the selection that corresponds to an array position in the full array.
matmul(a, b[, transp_a, transp_b])	Creates an operation that calculates a matrix multiplication (row-major notation): A(i, k) * B(k, j) if trans_a == trans_b, the usual transposed combinations, otherwise
matrix_set_diag(data, diagonal[, k, align])	Returns a tensor with the diagonals of input tensor replaced with the provided diagonal values.
max(data[, axis, keepdims])	Maximum of array elements over a given axis or a list of axes
maximum(lhs, rhs)	Take element-wise maximum of two tensors with auto-broadcasting
meshgrid(a_tuple, indexing)	Create coordinate matrices from coordinate vectors.
min(data[, axis, keepdims])	Minimum of array elements over a given axis or a list of axes
minimum(lhs, rhs)	Take element-wise maximum of two tensors with auto-broadcasting
mod(lhs, rhs)	Modulus with auto-broadcasting
multiply(lhs, rhs)	Multiplication with auto-broadcasting
ndarray_size(array[, dtype])	Get the number of elements of input array
negative(x)	Take negation of input x.
not_equal(lhs, rhs)	Compute (lhs!=rhs) with auto-broadcasting
one_hot(indices, on_value, off_value, depth, ...)	Returns a one-hot tensor where the locations repsented by indices take value on_value, other locations take value off_value.
power(lhs, rhs)	Power with auto-broadcasting
prod(data[, axis, keepdims])	Product of array elements over a given axis or a list of axes
reinterpret(x, dtype)	Reinterpret input to specified data type.
repeat(a, repeats, axis)	Repeats elements of an array.
reshape(a, newshape)	Reshape the array
reverse_sequence(a, seq_lengths[, seq_axis, ...])	Reverse the tensor for variable length slices.
right_shift(lhs, rhs)	Right shift with auto-broadcasting
round(x)	Round elements of x to nearest integer.
rsqrt(x)	Take inverse square root of input x.
scanop(data, binop, identity_value, op_name)	Cumulative binary operator (scan) with similar axis behavior as np.cumsum and np.cumprod.
scatter_elements(data, indices, updates[, ...])	Scatter elements from updates to corresponding indices of copied data.
scatter_nd(data, indices, updates, mode)	Scatter elements from a n-dimension array.
searchsorted(sorted_sequence, values[, ...])	Find indices where elements should be inserted to maintain order.
sequence_mask(data, valid_length[, ...])	Sets all elements outside the expected length of the sequence to a constant value.
shape(array[, dtype])	Get the shape of input array
sigmoid(x)	Take sigmoid tanh of input x.
sign(x)	Returns -1, 0, 1 based on sign of x.
sin(x)	Take sin of input x.
sinh(x)	Take sinh of input x.
sliding_window(data, axis, window_shape, strides)	Slide a window over the data tensor.
sort(data[, axis, is_ascend])	Performs sorting along the given axis and returns an array in sorted order.
sparse_reshape(sparse_indices, prev_shape, ...)	Reshape a Sparse Tensor
sparse_to_dense(sparse_indices, ...[, ...])	Converts a sparse representation into a dense tensor.
split(ary, indices_or_sections[, axis])	Split an array into multiple sub-arrays.
sqrt(x)	Take square root of input x.
squeeze(a[, axis])	Remove single-dimensional entries from the shape of an array.
stack(a, axis)	Repeats the whole array multiple times.
stft(data, n_fft, hop_length, win_length, ...)	The STFT computes the Fourier transform of short overlapping windows of the input.
strided_set(a, v, begin, end[, strides])	Set slice of an array.
strided_slice(a, begin, end[, strides, ...])	Slice of an array.
subtract(lhs, rhs)	Subtraction with auto-broadcasting
sum(data[, axis, keepdims])	Sum of array elements over a given axis or a list of axes
take(a, indices[, axis, batch_dims, mode])	Take elements from an array along an axis.
take_legalize(attrs, inputs, types)	Legalizes dyn.topk op.
tan(x)	Take tan of input x.
tanh(x)	Take hyperbolic tanh of input x.
tensordot(a, b, axes)	A generalization of matrix multiplication to tensor.
tile(a, reps)	Repeats the whole array multiple times.
topk(data[, k, axis, ret_type, is_ascend, dtype])	Get the top k elements in an input tensor along the given axis.
transpose(a[, axes])	Permute the dimensions of an array.
trilu(data, k, upper)	Given a 2-D matrix or batches of 2-D matrices, returns the upper or lower triangular part of the tensor.
trunc(x)	Take truncated value of the input of x, element-wise.
unique(data[, is_sorted, return_counts])	Find the unique elements of a 1-D tensor.
unravel_index(indices, shape)	Convert a flat index or array of flat indices into a tuple of coordinate arrays.
where(condition, x, y)	Get the elements, either from x or y, depending on the condition.
within_index(b, e, s, i)	Return a boolean value that indicates if i is within the given index.

class tvm.topi.Analyzer

Integer arithmetic analyzer

This is a stateful analyzer class that can be used to perform various symbolic integer analysis.

Methods:

bind(var, expr)	Bind a variable to the expression.
can_prove(expr[, strength])	Check whether we can prove expr to be true.
can_prove_equal(lhs, rhs)	Whether we can prove that lhs == rhs
canonical_simplify(expr)	Simplify expression via canonicalization.
const_int_bound(expr)	Find constant integer bound for expr.
constraint_scope(constraint)	Create a constraint scope.
int_set(expr, dom_map)	Compute a symbolic IntSet that covers expr for all values in dom_map.
modular_set(expr)	Find a modular set that expr belongs to.
rewrite_simplify(expr)	Simplify expression via rewriting rules.
simplify(expr[, steps])	Simplify expression via both rewrite and canonicalization.
update(var, info[, override])	Update infomation about var

Attributes:

enabled_extensions

Return the currently enabled extensions

bind(var: Var, expr: PrimExpr | Range)

Bind a variable to the expression.

参数:

• var (tvm.tir.Var) -- The variable.

• expr (Union[tir.PrimExpr, ir.Range]) -- The expression or the range to bind to.

can_prove(expr, strength=ProofStrength.DEFAULT)

Check whether we can prove expr to be true.

参数:

• expr (PrimExpr) -- The expression.

• strength (ProofStrength) -- The proof strength

返回:

result -- The result.

返回类型:

Expr

can_prove_equal(lhs: PrimExpr, rhs: PrimExpr)

Whether we can prove that lhs == rhs

参数:

• lhs (PrimExpr) -- The left-hand side of the comparison

• rhs (PrimExpr) -- The right-hand side of the comparison

返回:

result -- Whether we can prove that lhs == rhs

返回类型:

bool

canonical_simplify(expr)

Simplify expression via canonicalization.

参数:

expr (PrimExpr) -- The expression.

返回:

result -- The result.

返回类型:

Expr

const_int_bound(expr)

Find constant integer bound for expr.

参数:

expr (PrimExpr) -- The expression.

返回:

bound -- The result bound

返回类型:

ConstIntBound

constraint_scope(constraint)

Create a constraint scope.

参数:

constraint (PrimExpr) -- The constraint expression.

返回:

scope -- The constraint scope

返回类型:

ConstraintScope

示例

x = te.var("x")
analyzer = tvm.arith.Analyzer()
with analzyer.constraint_scope(x % 3 == 0):
    # constraint in effect
    assert analyzer.modular_set(x).coeff == 3
# constraint no longer in effect
assert analyzer.modular_set(x).coeff != 3

property enabled_extensions: Extension

Return the currently enabled extensions

int_set(expr, dom_map)

Compute a symbolic IntSet that covers expr for all values in dom_map.

参数:

• expr (PrimExpr) -- The expression.

• dom_map (Dict[tvm.tir.Var, tvm.arith.IntSet]) -- The domain for variables to be relaxed.

返回:

result -- The result.

返回类型:

IntSet

modular_set(expr)

Find a modular set that expr belongs to.

参数:

expr (PrimExpr) -- The expression.

返回:

result -- The result.

返回类型:

ModularSet

rewrite_simplify(expr)

Simplify expression via rewriting rules.

参数:

expr (PrimExpr) -- The expression.

返回:

result -- The result.

返回类型:

Expr

simplify(expr, steps=2)

Simplify expression via both rewrite and canonicalization.

参数:

• expr (PrimExpr) -- The expression.

• steps (The simplification runs in the order of) -- rewrite_simplify (step 1) -> canonical_simplify (step 2) -> rewrite_simplify (step 3) -> canonical_simplify (step 4) -> ... param steps controls how many steps to run. Default is 2, i.e., rewrite_simplify + canonical_simplify.

返回:

result -- The result.

返回类型:

Expr

update(var, info, override=False)

Update infomation about var

参数:

• var (tvm.tir.Var) -- The variable.

• info (tvm.Object) -- Related information.

• override (bool) -- Whether allow override.

class tvm.topi.Cast(dtype, value, span: Span | None = None)

Cast expression.

参数:

• dtype (str) -- The data type

• value (PrimExpr) -- The value of the function.

• span (Optional[Span]) -- The location of this expression in the source code.

exception tvm.topi.InvalidShapeError

Invalid shape for a topi function. i.e. call winograd template for non-3x3 kernel)

class tvm.topi.PrimExpr

Base class of all primitive expressions.

PrimExpr is used in the low-level code optimizations and integer analysis.

tvm.topi.abs(x)

Take absolute value of the input of x, element-wise.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.acos(x)

Take arc cos of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.acosh(x)

Take arc cosh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.add(lhs, rhs)

Addition with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.adv_index(data, indices)

Numpy style indexing with tensors.

参数:

• data (tvm.te.Tensor) -- Input data.

• indices (A list of tvm.te.Tensor) -- Tensor index.

返回:

result -- Output tensor

返回类型:

tvm.te.Tensor

tvm.topi.all(data, axis=None, keepdims=False)

Logical AND of array elements over a given axis or a list of axes

参数:

• data (tvm.te.Tensor) -- The input tvm boolean tensor

• axis (None or int or tuple of int) -- Axis or axes along which a logical AND is performed. The default, axis=None, will perform logical AND over all elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.any(data, axis=None, keepdims=False)

Logical OR of array elements over a given axis or a list of axes

参数:

• data (tvm.te.Tensor) -- The input tvm boolean tensor

• axis (None or int or tuple of int) -- Axis or axes along which a logical OR is performed. The default, axis=None, will perform logical OR over all elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.arange(start, stop=None, step=1, dtype='float32')

Creates a tensor with evenly spaced values within a given interval.

参数:

• start (tvm.Expr, optional) -- Start of interval. The interval includes this value. The default start value is 0.

• stop (tvm.Expr) -- Stop of interval. The interval does not include this value.

• step (tvm.Expr, optional) -- Spacing between values. The default step size is 1.

• dtype (str, optional) -- The target data type.

返回:

result -- The resulting tensor.

返回类型:

tvm.te.Tensor

tvm.topi.argmax(data, axis=None, keepdims=False, select_last_index=False)

Returns the indices of the maximum values along an axis.

参数:

• data (tvm.te.Tensor) -- The input tvm tensor

• axis (None or int or tuple of int) -- Axis or axes along which a argmax operation is performed. The default, axis=None, will find the indices of the maximum element of the elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

• select_last_index (bool) -- Whether to select the last index if the maximum element appears multiple times, else select the first index.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.argmin(data, axis=None, keepdims=False, select_last_index=False)

Returns the indices of the minimum values along an axis.

参数:

• data (tvm.te.Tensor) -- The input tvm tensor

• axis (None or int or tuple of int) -- Axis or axes along which a argmin operation is performed. The default, axis=None, will find the indices of minimum element all of the elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

• select_last_index (bool) -- Whether to select the last index if the minimum element appears multiple times, else select the first index.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.argsort(data, valid_count=None, axis=-1, is_ascend=1, dtype='float32')

Performs sorting along the given axis and returns an array of indices having the same shape as an input array that index data in sorted order.

参数:

• data (tvm.te.Tensor) -- The input tensor.

• valid_count (tvm.te.Tensor, optional) -- 1-D tensor for valid number of boxes.

• axis (int, optional) -- Axis along which to sort the input tensor. By default the flattened array is used.

• is_ascend (boolean, optional) -- Whether to sort in ascending or descending order.

• dtype (string, optional) -- DType of the output indices.

返回:

out -- Sorted index tensor.

返回类型:

tvm.te.Tensor

示例

# An example to use argsort
dshape = (1, 5, 6)
data = te.placeholder(dshape, name="data")
axis = 0
is_ascend = False
out = argsort(data, axis=axis, is_ascend=is_ascend)
np_data = np.random.uniform(dshape)
s = topi.generic.schedule_argsort(out)
f = tvm.build(s, [data, out], "llvm")
dev = tvm.cpu()
tvm_data = tvm.nd.array(np_data, dev)
tvm_out = tvm.nd.array(np.zeros(dshape, dtype=data.dtype), dev)
f(tvm_data, tvm_out)

tvm.topi.argwhere(output_shape, condition)

Find the indices of elements of a tensor that are non-zero.

参数:

condition (tvm.te.Tensor) -- Tensor with boolean values.

返回:

out -- Indices of non-zero elements.

返回类型:

tvm.te.Tensor

tvm.topi.asin(x)

Take arc sin of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.asinh(x)

Take arc sinh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.atan(x)

Take atan of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.atanh(x)

Take atanh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.binary_search(ib, sequence_offset, search_range, sorted_sequence, value, right, out_dtype)

Common IR generator for binary search used by CPU and GPU backends.

sorted_sequence is a N-D Buffer whose innermost dimension we want to search for value, and search_range is the size of the innermost dimension. sequence_offset is a 1-D linearlized offset specifying which of innermost sequences to search.

So the search for value is performed over sorted_sequence[sequence_offset:(sequence_offset + search_range)]. Note that we index N-D Buffer by 1-D linearlized indices.

tvm.topi.bitwise_and(lhs, rhs)

Compute element-wise bitwise and of data.

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.bitwise_not(data)

Compute element-wise bitwise not of data.

参数:

data (tvm.te.Tensor or Expr)

返回:

ret -- Returns Expr if the operand are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.bitwise_or(lhs, rhs)

Compute element-wise bitwise or of data.

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.bitwise_xor(lhs, rhs)

Compute element-wise bitwise xor of data.

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.broadcast_to(data, shape)

Broadcast the src to the target shape

We follows the numpy broadcasting rule. See also https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html

参数:

• data (tvm.te.Tensor) -- The input data

• shape (list or tuple) -- The target shape to be broadcasted.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.cast(x, dtype, span=None)

Cast input to specified data type.

参数:

• x (tvm.te.Tensor or Expr) -- Input argument.

• dtype (str) -- Data type.

• span (Optional[Span]) -- The location of the cast in the source.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.ceil(x)

Take ceil of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.ceil_log2(x)

Compute integer ceil log2 with a special code path for vulkan SPIR-V does not support log2 on fp64. Instead, we compute integer ceil_log2 via clz intrinsic when the target is vulkan.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.clip(x, a_min, a_max)

Clip (limit) the values in an array. Given an interval, values outside the interval are clipped to the interval edges.

参数:

• x (tvm.te.Tensor) -- Input argument.

• a_min (tvm.tir.PrimExpr) -- Minimum value.

• a_max (tvm.tir.PrimExpr) -- Maximum value.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.collapse_sum(data, target_shape)

Return a summation of data to the given shape.

collapse_sum is intended as the backward operator of topi broadcast operators in the automatic differentiation process.

We expect that data is the result of broadcasting some tensor of target_shape in some broadcast operation. Thus target_shape and data.shape must follow broadcast rules.

During computation, the axes of data.shape and target_shape are checked from right to left. For every axis, if it either: - exist in data but not in target_shape, or - is larger than 1 in data and equals to 1 in target_shape, data will be summed over this axis.

参数:

• data (tvm.te.Tensor) -- The input tensor.

• shape (Tuple[int]) -- The shape to collapse to.

返回:

ret -- The result tensor after summation.

返回类型:

tvm.te.Tensor

tvm.topi.concatenate(a_tuple, axis=0)

Join a sequence of arrays along an existing axis.

参数:

• a_tuple (tuple of tvm.te.Tensor) -- The arrays to concatenate

• axis (int, optional) -- The axis along which the arrays will be joined. Default is 0.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.const_vector(vector, name='const_vector')

convert a const numpy 1-dimensional vector to tvm tensor

参数:

• vector (numpy.ndarray) -- Const input array

• name (str, optional) -- The name of output op

返回:

tensor -- The created tensor

返回类型:

Tensor

tvm.topi.cos(x)

Take cos of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.cosh(x)

Take cosh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.cumprod(data: Tensor, axis: int | None = None, dtype: int | None = None, exclusive: bool | None = None) → Tensor

Numpy style cumprod op. Return the cumulative product of the elements along a given axis.

参数:

• data (tvm.te.Tensor) -- The input data to the operator.

• axis (int, optional) -- Axis along which the cumulative product is computed. The default (None) is to compute the cumproduct over the flattened array.

• dtype (string, optional) -- Type of the returned array and of the accumulator in which the elements are multiplied. If dtype is not specified, it defaults to the dtype of data.

• exclusive (bool, optional) -- If True, will return exclusive product in which the first element is not included. In other terms, if True, the j-th output element would be the product of the first (j-1) elements. Otherwise, it would be the product of the first j elements.

返回:

result -- The result has the same size as data, and the same shape as data if axis is not None. If axis is None, the result is a 1-d array.

返回类型:

tvm.te.Tensor

tvm.topi.cumsum(data: Tensor, axis: int | None = None, dtype: str | None = None, exclusive: bool | None = None) → Tensor

Numpy style cumsum op. Return the cumulative sum of the elements along a given axis.

参数:

• data (tvm.te.Tensor) -- The input data to the operator.

• axis (int, optional) -- Axis along which the cumulative sum is computed. The default (None) is to compute the cumsum over the flattened array.

• dtype (string, optional) -- Type of the returned array and of the accumulator in which the elements are summed. If dtype is not specified, it defaults to the dtype of data.

• exclusive (bool, optional) -- If True, will return exclusive sum in which the first element is not included. In other terms, if True, the j-th output element would be the sum of the first (j-1) elements. Otherwise, it would be the sum of the first j elements.

返回:

result -- The result has the same size as data, and the same shape as data if axis is not None. If axis is None, the result is a 1-d array.

返回类型:

tvm.te.Tensor

tvm.topi.decl_buffer(shape, dtype=None, name='buffer', data=None, strides=None, elem_offset=None, scope='', data_alignment=-1, offset_factor=0, buffer_type='', axis_separators=None, span=None)

Declare a new symbolic buffer.

Normally buffer is created automatically during lower and build. This is only needed if user want to specify their own buffer layout.

See the note below for detailed discussion on usage of buffer.

参数:

• shape (tuple of Expr) -- The shape of the buffer.

• dtype (str, optional) -- The data type of the buffer.

• name (str, optional) -- The name of the buffer.

• data (tir.Var, optional) -- The data pointer in the buffer.

• strides (array of Expr) -- The stride of the buffer.

• elem_offset (Expr, optional) -- The beginning offset of the array to data. In terms of number of elements of dtype.

• scope (str, optional) -- The storage scope of the buffer, if not global. If scope equals empty string, it means it is global memory.

• data_alignment (int, optional) -- The alignment of data pointer in bytes. If -1 is passed, the alignment will be set to TVM's internal default.

• offset_factor (int, optional) -- The factor of elem_offset field, when set, elem_offset is required to be multiple of offset_factor. If 0 is pssed, the alignment will be set to 1. if non-zero is passed, we will created a Var for elem_offset if elem_offset is not None.

• buffer_type (str, optional, {"", "auto_broadcast"}) -- auto_broadcast buffer allows one to implement broadcast computation without considering whether dimension size equals to one. TVM maps buffer[i][j][k] -> buffer[i][0][k] if dimension j's shape equals 1.

• axis_separators (list of int, optional) -- If passed, a list of separators between groups of axes, each of which is flattened to an output axis. For flat memory spaces, should either be None, or an empty list.

• span (Optional[Span]) -- The location of the decl_buffer creation in the source.

返回:

buffer -- The created buffer

返回类型:

tvm.tir.Buffer

示例

Here's an example of how broadcast buffer can be used to define a symbolic broadcast operation,

m0, m1, m2 = te.var("m0"), te.var("m1"), te.var("m2")
n0, n1, n2 = te.var("n0"), te.var("n1"), te.var("n2")
o0, o1, o2 = te.var("o0"), te.var("o1"), te.var("o2")
A = te.placeholder((m0, m1, m2), name='A')
B = te.placeholder((n0, n1, n2), name='B')
C = te.compute((o0, o1, o2), lambda i, j, k: A[i, j, k] + B[i, j, k], name='C')
Ab = tvm.tir.decl_buffer(A.shape, A.dtype, name="Ab", buffer_type="auto_broadcast")
Bb = tvm.tir.decl_buffer(B.shape, B.dtype, name="Bb", buffer_type="auto_broadcast")
s = te.create_schedule(C.op)
fadd = tvm.build(s, [A, B, C], target='llvm', name='bcast_add', binds={A:Ab, B:Bb})
dev = tvm.cpu(0)
a = tvm.nd.array(np.random.uniform(size=(2, 4, 3)).astype(A.dtype), dev)
b = tvm.nd.array(np.random.uniform(size=(2, 1, 3)).astype(B.dtype), dev)
c = tvm.nd.array(np.zeros((2, 4, 3), dtype=C.dtype), dev)
fadd(a, b, c)
tvm.testing.assert_allclose(c.numpy(), a.numpy() + b.numpy())

备注

Buffer data structure reflects the DLTensor structure in dlpack. While DLTensor data structure is very general, it is usually helpful to create function that only handles specific case of data structure and make compiled function benefit from it.

If user pass strides and elem_offset is passed as None when constructing the function, then the function will be specialized for the DLTensor that is compact and aligned. If user pass a fully generic symbolic array to the strides, then the resulting function becomes fully generic.

tvm.topi.dft(re_data: Tensor, im_data: Tensor, inverse: IntImm)

Computes the discrete Fourier transform of input (calculation along the last axis). This gives frequency components of the signal as they change over time.

参数:

• re_data (relay.Expr) -- N-D tensor, real part of the input signal.

• im_data (relay.Expr) -- N-D tensor, imaginary part of the input signal. If the signal is real, then the values of this tensor are zeros.

• inverse (bool) -- Whether to perform the inverse discrete fourier transform.

返回:

• re_output (relay.Expr) -- The Fourier Transform of the input (Real part).

• im_output (relay.Expr) -- The Fourier Transform of the input (Imaginary part).

tvm.topi.div(a, b, span=None)

Compute a / b as in C/C++ semantics.

参数:

• a (PrimExpr) -- The left hand operand, known to be non-negative.

• b (PrimExpr) -- The right hand operand, known to be non-negative.

• span (Optional[Span]) -- The location of this operator in the source.

返回:

res -- The result expression.

返回类型:

PrimExpr

备注

When operands are integers, returns truncdiv(a, b, span).

tvm.topi.divide(lhs, rhs)

Division with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.dynamic_strided_slice(a, begin, end, strides, output_shape)

Slice of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be sliced.

• begin (tvm.te.Tensor) -- The indices to begin with in the slicing.

• end (tvm.te.Tensor) -- Indices indicating end of the slice.

• strides (tvm.te.Tensor) -- Specifies the stride values, it can be negative in that case, the input tensor will be reversed in that particular axis.

• output_shape (list of PrimExpr) -- Specifies the output shape

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.einsum(subscripts, *operand)

Evaluates the Einstein summation convention on the operands.

参数:

• subscripts (string) -- Specifies the subscripts for summation as comma separated list of subscript labels. An implicit (classical Einstein summation) calculation is performed unless the explicit indicator ‘->’ is included as well as subscript labels of the precise output form.

• a_tuple (tuple of tvm.te.Tensor) -- These are the Tensors for the operation. The only difference of einsum between in tvm and numpy is it needs an extra brackets for the tensors. For example, topi.einsum("ij, jk -> ik", (A, B)).

返回:

out -- The calculation based on the Einstein summation convention.

返回类型:

tvm.te.Tensor

tvm.topi.elemwise_sum(xs)

Perform element-wise sum on inputs

参数:

xs (list of tvm.te.Tensor) -- Input arguments.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.equal(lhs, rhs)

Compute (lhs==rhs) with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.erf(x)

Take gauss error function of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.erf_legalize(attrs, inputs, types)

Legalizes ERF op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr.

返回类型:

tvm.relay.Expr

tvm.topi.exp(x)

Take exponential of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.expand_dims(a, axis, num_newaxis=1)

Expand the shape of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be expanded.

• num_newaxis (int, optional) -- Number of newaxis to be inserted on axis

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.expand_like(a, shape_like, axis)

Expand an input array with the shape of second array. This operation can always be composed of unsqueezing and expanding dims on those unsqueezed axes.

示例

input = [ 12.  19.  27.]
input.shape = (3,)

new_shape_array = [[[1,2],[2,3],[1,3]],
                [[1,4],[4,3],[5,2]],
                [[7,1],[7,2],[7,3]]]
new_shape_array.shape = (3, 3, 2)

expand_like(input, [1,2], new_shape_array) =
                [[[12,12],[12,12],[12,12]],
                [[19,19],[19,19],[19,19]],
                [[27,27],[27,27],[27,27]]]

参数:

• a (tvm.te.Tensor) -- The tensor to be expanded.

• shape_like (tvm.te.Tensor) -- The tensor to with target shape.

• axis (list of int) -- axis to be expanded on

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.extern(shape, inputs, fcompute, name='extern', dtype=None, in_buffers=None, out_buffers=None, tag='', attrs=None)

Compute several tensors via an extern function.

参数:

• shape (tuple or list of tuples.) -- The shape of the outputs.

• inputs (list of Tensor) -- The inputs

• fcompute (lambda function of inputs, outputs-> stmt) --

  Specifies the IR statement to do the computation. See the following note for function signature of fcompute

  备注

  Parameters

  • ins (list of tvm.tir.Buffer) - Placeholder for each inputs

  • outs (list of tvm.tir.Buffer) - Placeholder for each outputs

  Returns

  • stmt (tvm.tir.Stmt) - The statement that carries out array computation.

• name (str, optional) -- The name hint of the tensor

• dtype (str or list of str, optional) -- The data types of outputs, by default dtype will be same as inputs.

• in_buffers (tvm.tir.Buffer or list of tvm.tir.Buffer, optional) -- Input buffers.

• out_buffers (tvm.tir.Buffer or list of tvm.tir.Buffer, optional) -- Output buffers.

tag: str, optional

Additonal tag information about the compute.

attrs: dict, optional

The additional auxiliary attributes about the compute.

返回:

tensor -- The created tensor or tuple of tensors contains multiple outputs.

返回类型:

Tensor or list of Tensors

示例

In the code below, C is generated by calling external PackedFunc tvm.contrib.cblas.matmul

A = te.placeholder((n, l), name="A")
B = te.placeholder((l, m), name="B")
C = te.extern((n, m), [A, B],
               lambda ins, outs: tvm.tir.call_packed(
                  "tvm.contrib.cblas.matmul",
                    ins[0], ins[1], outs[0], 0, 0), name="C")

tvm.topi.eye(n: int, m: int | None = None, k: int = 0, dtype: str = 'float32') → Tensor

Generate an identity matrix or a matrix with ones on the k-th diagonal.

参数:

• n (int) -- Number of rows

• m (int, optional) -- Number of columns. If None, defaults to n.

• k (int, optional) -- Index of the diagonal. 0 (default) refers to the main diagonal. A positive value refers to an upper diagonal, and a negative value to a lower diagonal.

• dtype (str, optional) -- Data type of the returned array.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.fast_erf(x)

Take gauss error function of input x using fast_erf implementation.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.fast_exp(x)

Take exponential of input x using fast_exp implementation

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.fast_tanh(x)

Take hyperbolic tangent of input x using fast_tanh implementation

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.fixed_point_multiply(x, multiplier, shift)

Fixed point multiplication between data and a fixed point constant expressed as multiplier * 2^(-shift), where multiplier is a Q-number with 31 fractional bits

参数:

• x (tvm.te.Tensor or Expr) -- Input argument.

• multiplier (int) -- Multiplier of a fixed floating point number described as multiplier*2^(-shift).

• shift (int) -- Shift of a fixed floating point number described as multiplier*2^(-shift).

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.fixed_point_multiply_per_axis(x: Tensor, y: Tensor, lshift: Tensor, rshift: Tensor, is_lshift_required: int, is_rshift_required: int, axes)

Fixed point multiplication between data and a fixed point constant expressed as multiplier * 2^(-shift), where multiplier is a Q-number with 31 fractional bits

参数:

• x (tvm.te.Tensor) -- Input argument.

• y (tvm.te.Tensor) -- Multiplier of a fixed floating point number described as multiplier*2^(-shift).

• lshift (tvm.te.Tensor) -- Left shifts of a fixed floating point number described as multiplier*2^(-shift).

• rshift (tvm.te.Tensor) -- Right shifts of a fixed floating point number described as multiplier*2^(-shift).

• is_lshift_required (int) -- Whether we need to do left shift or not.

• is_rshift_required (int) -- Whether we need to do right shift or not.

返回:

z -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.flip(a, axis=0)

Flip/reverse elements of an array in a particular axis.

参数:

• a (tvm.te.Tensor) -- The tensor to be expanded.

• axis (int, optional) -- The axis along which the tensors will be reveresed.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.floor(x)

Take floor of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.floor_divide(lhs, rhs)

Floor division with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.floor_mod(lhs, rhs)

Floor modulus with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.floordiv(a, b, span=None)

Compute the floordiv of two expressions.

参数:

• a (PrimExpr) -- The left hand operand

• b (PrimExpr) -- The right hand operand

• span (Optional[Span]) -- The location of this operator in the source.

返回:

res -- The result expression.

返回类型:

PrimExpr

tvm.topi.floormod(a, b, span=None)

Compute the floormod of two expressions.

参数:

• a (PrimExpr) -- The left hand operand

• b (PrimExpr) -- The right hand operand

• span (Optional[Span]) -- The location of this operator in the source.

返回:

res -- The result expression.

返回类型:

PrimExpr

tvm.topi.full(shape, dtype, fill_value)

Fill tensor with fill_value

参数:

• shape (tuple) -- Input tensor shape.

• dtype (str) -- Data type

• fill_value (float) -- Value to be filled

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.full_like(x, fill_value)

Construct a tensor with same shape as input tensor,

then fill tensor with fill_value.

参数:

• x (tvm.te.Tensor) -- Input argument.

• fill_value (float) -- Value to be filled

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.gather(data, axis, indices)

Gather values along given axis from given indices.

E.g. for a 3D tensor, output is computed as:

out[i][j][k] = data[indices[i][j][k]][j][k]  # if axis == 0
out[i][j][k] = data[i][indices[i][j][k]][k]  # if axis == 1
out[i][j][k] = data[i][j][indices[i][j][k]]  # if axis == 2

indices must have same shape as data, except at dimension axis which must just be not null. Output will have same shape as indices.

参数:

• data (tvm.te.Tensor) -- The input data to the operator.

• axis (int) -- The axis along which to index.

• indices (tvm.te.Tensor) -- The indices of the values to extract.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.gather_nd(a, indices, batch_dims=0)

Gather elements from a n-dimension array..

参数:

• a (tvm.te.Tensor) -- The source array.

• indices (tvm.te.Tensor) -- The indices of the values to extract.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.get_const_tuple(in_tuple)

Verifies input tuple is IntImm or Var, returns tuple of int or Var.

参数:

in_tuple (tuple of Expr) -- The input.

返回:

out_tuple -- The output.

返回类型:

tuple of int

tvm.topi.greater(lhs, rhs)

Compute (lhs>rhs) with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.greater_equal(lhs, rhs)

Compute (lhs>=rhs) with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.hybrid_argwhere_1d(output_shape, condition)

Find the indices of elements of a 1-D tensor that are non-zero.

参数:

condition (tvm.te.Tensor) -- 1-D tensor with boolean values.

返回:

out -- Indices of non-zero elements.

返回类型:

tvm.te.Tensor

tvm.topi.hybrid_argwhere_2d(output_shape, condition)

Find the indices of elements of a 2-D tensor that are non-zero.

参数:

condition (tvm.te.Tensor) -- 2-D tensor with boolean values.

返回:

out -- Indices of non-zero elements.

返回类型:

tvm.te.Tensor

tvm.topi.hybrid_argwhere_3d(output_shape, condition)

Find the indices of elements of a 3-D tensor that are non-zero.

参数:

condition (tvm.te.Tensor) -- 3-D tensor with boolean values.

返回:

out -- Indices of non-zero elements.

返回类型:

tvm.te.Tensor

tvm.topi.hybrid_argwhere_4d(output_shape, condition)

Find the indices of elements of a 4-D tensor that are non-zero.

参数:

condition (tvm.te.Tensor) -- 4-D tensor with boolean values.

返回:

out -- Indices of non-zero elements.

返回类型:

tvm.te.Tensor

tvm.topi.hybrid_argwhere_5d(output_shape, condition)

Find the indices of elements of a 5-D tensor that are non-zero.

参数:

condition (tvm.te.Tensor) -- 5-D tensor with boolean values.

返回:

out -- Indices of non-zero elements.

返回类型:

tvm.te.Tensor

tvm.topi.identity(x)

Take identity of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.invert_permutation(data)

Computes the inverse permutation of data.

参数:

data (tvm.te.Tensor) -- Input data

返回:

result -- Output tensor

返回类型:

tvm.te.Tensor

示例

data = [3, 4, 0, 2, 1]
topi.invert_permutation(data) = [2, 4, 3, 0, 1]

tvm.topi.isfinite(x)

Check if value of x is finite, element-wise.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.isinf(x)

Check if value of x is infinite, element-wise.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.isnan(x)

Check if value of x is NaN, element-wise.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.layout_transform(array, src_layout, dst_layout, schedule_rule='None')

Transform the layout according to src_layout and dst_layout

参数:

• array (tvm.te.Tensor) -- The source array.

• src_layout (str) -- the source layout.

• dst_layout (str) -- the destination layout.

• schedule_rule (str) -- the schedule rule to apply if any

tvm.topi.left_shift(lhs, rhs)

Left shift with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.less(lhs, rhs)

Compute (lhs<rhs) with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.less_equal(lhs, rhs)

Compute (lhs<=rhs) with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.log(x)

Take logarithm of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.log10(x)

Take logarithm to the base 10 of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.log2(x)

Take logarithm to the base 2 of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.logical_and(lhs, rhs)

Compute element-wise logical and of data.

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.logical_not(data)

Compute element-wise logical not of data.

参数:

data (tvm.te.Tensor or Expr)

返回:

ret -- Returns Expr if the operand are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.logical_or(lhs, rhs)

Compute element-wise logical or of data.

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.logical_xor(lhs, rhs)

Compute element-wise logical xor of data.

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.make_idx(b, e, s, z, i)

Return the array position in the selection that corresponds to an array position in the full array.

The returned value is only meaningful if within_index() returns True for the same set of parameters.

参数:

• b (Expr) -- beginning of the index

• e (Expr) -- end of the index

• s (Expr) -- strides of index

• z (Expr) -- size of the indexed dimension

• i (Expr) -- array position

返回:

position -- int expression that corresponds to an array position in the selection.

返回类型:

Expr

tvm.topi.matmul(a, b, transp_a=False, transp_b=False)

Creates an operation that calculates a matrix multiplication (row-major notation): A(i, k) * B(k, j) if trans_a == trans_b, the usual transposed combinations, otherwise

参数:

• a (The matrix A)

• b (The matrix B)

• trans_a (Is A's layout transposed?)

• trans_b (Is B's layout transposed?)

返回类型:

A Tensor whose op member is the matmul operation

tvm.topi.matrix_set_diag(data, diagonal, k=0, align='RIGHT_LEFT')

Returns a tensor with the diagonals of input tensor replaced with the provided diagonal values.

参数:

• data (relay.Expr) -- Input Tensor.

• diagonal (relay.Expr) -- Values to be filled in the diagonal.

• k (int or tuple of int, optional) -- Diagonal Offset(s). The diagonal or range of diagonals to set. (0 by default) Positive value means superdiagonal, 0 refers to the main diagonal, and negative value means subdiagonals. k can be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band. k[0] must not be larger than k[1].

• align (string, optional) -- Some diagonals are shorter than max_diag_len and need to be padded. align is a string specifying how superdiagonals and subdiagonals should be aligned, respectively. There are four possible alignments: "RIGHT_LEFT" (default), "LEFT_RIGHT", "LEFT_LEFT", and "RIGHT_RIGHT". "RIGHT_LEFT" aligns superdiagonals to the right (left-pads the row) and subdiagonals to the left (right-pads the row). It is the packing format LAPACK uses. cuSPARSE uses "LEFT_RIGHT", which is the opposite alignment.

返回:

result -- New tensor with given diagonal values.

返回类型:

relay.Expr

示例

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]]

topi.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]]]

tvm.topi.max(data, axis=None, keepdims=False)

Maximum of array elements over a given axis or a list of axes

参数:

• data (tvm.te.Tensor) -- The input tvm tensor

• axis (None or int or tuple of int) -- Axis or axes along which the max operation is performed. The default, axis=None, will find the max element from all of the elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.maximum(lhs, rhs)

Take element-wise maximum of two tensors with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.meshgrid(a_tuple, indexing)

Create coordinate matrices from coordinate vectors.

参数:

• a_tuple (tuple of tvm.te.Tensor) -- The coordinate vectors or scalars.

• indexing (str) -- Indexing mode, either "ij" or "xy".

返回:

result -- The resulting grids for each axis.

返回类型:

tuple of tvm.te.Tensor

tvm.topi.min(data, axis=None, keepdims=False)

Minimum of array elements over a given axis or a list of axes

参数:

• data (tvm.te.Tensor) -- The input tvm tensor

• axis (None or int or tuple of int) -- Axis or axes along which a minimum operation is performed. The default, axis=None, will find the minimum element from all of the elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.minimum(lhs, rhs)

Take element-wise maximum of two tensors with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.mod(lhs, rhs)

Modulus with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.multiply(lhs, rhs)

Multiplication with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.ndarray_size(array, dtype='int32')

Get the number of elements of input array

参数:

• array (tvm.te.Tensor) -- The source tensor.

• dtype (str, optional) -- The target data type.

返回:

result -- The resulting tensor.

返回类型:

tvm.te.Tensor

tvm.topi.negative(x)

Take negation of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.not_equal(lhs, rhs)

Compute (lhs!=rhs) with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.one_hot(indices, on_value, off_value, depth, axis, dtype)

Returns a one-hot tensor where the locations repsented by indices take value on_value, other locations take value off_value. Final dimension is <indices outer dimensions> x depth x <indices inner dimensions>.

参数:

• indices (tvm.te.Tensor) -- Locations to set to on_value.

• on_value (tvm.te.Tensor) -- Value to fill at indices.

• off_value (tvm.te.Tensor) -- Value to fill at all other positions besides indices.

• depth (int) -- Depth of the one-hot dimension.

• axis (int) -- Axis to fill.

• dtype (relay.DataType) -- Data type of the output tensor.

返回:

ret -- The one-hot tensor.

返回类型:

relay.Expr

示例

indices = [0, 1, 2]

relay.one_hot(indices, 3) =
    [[1, 0, 0],
     [0, 1, 0],
     [0, 0, 1]]

tvm.topi.power(lhs, rhs)

Power with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.prod(data, axis=None, keepdims=False)

Product of array elements over a given axis or a list of axes

参数:

• data (tvm.te.Tensor) -- The input tvm tensor

• axis (None or int or tuple of int) -- Axis or axes along which a prod operation is performed. The default, axis=None, will get the prod element over all of the elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.reinterpret(x, dtype)

Reinterpret input to specified data type.

参数:

• x (tvm.te.Tensor) -- Input argument.

• dtype (str) -- Data type.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.repeat(a, repeats, axis)

Repeats elements of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be repeated.

• repeats (int, required) -- Number of repetitions for each element

• axis (int, optional) -- The axis along which to repeat values

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.reshape(a, newshape)

Reshape the array

参数:

• a (tvm.te.Tensor) -- The tensor to be reshaped

• newshape (tuple of ints) -- The new shape

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.reverse_sequence(a, seq_lengths, seq_axis=1, batch_axis=0)

Reverse the tensor for variable length slices. Input is first sliced along batch axis and then elements are reversed along seq axis.

参数:

• a (tvm.te.Tensor) -- The tensor to be reversed.

• seq_lengths (tvm.te.Tensor) -- A 1D Tensor with length a.dims[batch_axis] Must be one of the following types: int32, int64 if seq_lengths[i] > a.dims[seq_axis], it is rounded to a.dims[seq_axis] if seq_lengths[i] < 1, it is rounded to 1

• seq_axis (int, optional) -- The axis along which the elements will be reversed. Default is 1.

• batch_axis (int, optional) -- The axis along which the tensor will be sliced. Default is 0.

返回:

ret -- The computed result of same shape and type as of input.

返回类型:

tvm.te.Tensor

tvm.topi.right_shift(lhs, rhs)

Right shift with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.round(x)

Round elements of x to nearest integer.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.rsqrt(x)

Take inverse square root of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.scanop(data: Tensor, binop: Callable[[tvm.Expr, tvm.Expr], tvm.Expr], identity_value: tvm.Expr, op_name: str, axis: int | None = None, dtype: str | None = None, exclusive: bool | None = None) → Tensor

Cumulative binary operator (scan) with similar axis behavior as np.cumsum and np.cumprod.

See cumprod and cumsum for an example of use.

E.g. if * is your binary operator and the input tensor is [1, 2, 3, 4] the output may be [1, 1 * 2, 1 * 2 * 3, 1 * 2 * 3 * 4]

参数:

• data (tvm.te.Tensor) -- The input data to the operator.

• binop (Callable (tvm.Expr, tvm.Expr) -> tvm.Expr) -- A binary operator which should be associative and commutative. E.g. if * is your operator then a * (b * c) = (a * b) * c and a * b = b * a

• identity_value (tvm.Expr) -- A value for the binary operation which provides the identity property. E.g. if * is your operator and i is the identity_value then a * i = a for all a in the domain of your operation.

• axis (int, optional) -- Axis along which the operation is computed. The default (None) is to compute the cumulative operation over the flattened array.

• dtype (string, optional) -- Type of the returned array and of the accumulator in which the elements are computed. If dtype is not specified, it defaults to the dtype of data.

• exclusive (bool, optional) -- If True will return exclusive cumulative operation in which the first element is not included. In other terms, if True, the j-th output element would be the cumulative operation of the first (j-1) elements. Otherwise, it would be the cumulative operation of the first j elements. The cumulative operation of zero elements is assumed to be the identity_value.

返回:

result -- The result has the same size as data, and the same shape as data if axis is not None. If axis is None, the result is a 1-d array.

返回类型:

tvm.te.Tensor

tvm.topi.scatter_elements(data, indices, updates, axis=0, reduction='update')

Scatter elements from updates to corresponding indices of copied data.

Data, indices, updates and output have the same shape. Indices can not have duplicates (if idx1 != idx2, then indices[idx1] != indices[idx2]) if reduction == "update".

output[indices[i][j]][j] = f(output[indices[i][j]][j], updates[i][j]) if axis = 0
output[i][indices[i][j]] = f(output[i][indices[i][j]], updates[i][j]) if axis = 1

where the update function f is determined by the reduction. Five types of the function are supported: "update", "add", "mul", "min" and "max" (see below)

参数:

• data (tvm.te.Tensor) -- The source array.

• indices (tvm.te.Tensor) -- The indices of the values to extract.

• updates (tvm.te.Tensor) -- The updates to apply at the Indices

• axis (optional, int) -- The axis to scatter on. It is zero by default.

• reduction (optional, string) -- The update mode for the algorithm, either "update", "add", "mul", "min" or "max" If update, the update values will replace the input data If add, the update values will be added to the input data If mul, the input data will be multiplied on the update values If mean, the input data will be mean between the update values and the input data If min, there is choice of minimal between the update values and the input data If max, there is choice of maximal between the update values and the input data It is "update" by default

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.scatter_nd(data, indices, updates, mode)

Scatter elements from a n-dimension array.

Given updates with shape (Y_0, ..., Y_{K-1}, X_M, ..., X_{N-1}), indices with shape (M, Y_0, ..., Y_{K-1}), and output copied from data with shape (X_0, X_1, ..., X_{N-1}), scatter_nd computes

output[indices[0, y_0, ..., y_{K-1}],
       ...,
       indices[M-1, y_0, ..., y_{K-1}],
       x_M,
       ...,
       x_{N-1}
      ] = f(output[...], updates[y_0, ..., y_{K-1}, x_M, ..., x_{N-1}])

where the update function f is determinted by the mode.

参数:

• data (tvm.te.Tensor) -- The source array.

• indices (tvm.te.Tensor) -- The indices of the values to extract.

• updates (tvm.te.Tensor) -- The updates to apply at the Indices

• mode (string) -- The update mode for the algorithm, either "update" or "add" If update, the update values will replace the input data If add, the update values will be added to the input data

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.searchsorted(sorted_sequence, values, right=False, out_dtype='int64')

Find indices where elements should be inserted to maintain order.

If sorted_sequence is N-dimensional, the innermost dimension of values are searched in the corresponding dimension of sorted_sequence.

参数:

• sorted_sequence (te.Tensor) -- N-D or 1-D Tensor, containing monotonically increasing sequence on the innermost dimension.

• values (te.Tensor) -- N-D Tensor containing the search values. When sorted_sequence is 1-D, the shape of values can be arbitrary. Otherwise, ranks of sorted_sequence and values must be the same, and outer N-1 axes must have the same size.

• right (bool, optional) -- Controls which index is returned if a value lands exactly on one of sorted values. If False, the index of the first suitable location found is given. If true, return the last such index. If there is no suitable index, return either 0 or N (where N is the size of the innermost dimension).

• dtype (string, optional) -- The data type of the output indices.

返回:

indices -- Tensor with same shape as values, representing the indices of elements of values if they are inserted in sorted_sequence.

返回类型:

te.Tensor

tvm.topi.sequence_mask(data, valid_length, mask_value=0, axis=0)

Sets all elements outside the expected length of the sequence to a constant value.

This function takes an n-dimensional input array of the form [MAX_LENGTH, batch_size, ...] or [batch_size, MAX_LENGTH, ...] and returns an array of the same shape.

axis means the axis of the length dimension and can only be 0 or 1. If axis is 0, the data must have shape [MAX_LENGTH, batch_size, ...]. Otherwise (axis=1), the data must have shape [batch_size, MAX_LENGTH, ...].

valid_length gives the length of each sequence. valid_length should be a 1D int array with positive ints and has dimension [batch_size,].

参数:

• data (tvm.te.Tensor) -- N-D with shape [MAX_LENGTH, batch_size, ...] or [batch_size, MAX_LENGTH, ...] depending on the value of axis.

• valid_length (tvm.te.Tensor) -- 1-D with shape [batch_size,]

• mask_value (float, optional) -- The masking value, default 0

• axis (int, optional) -- axis of the length dimension, must be 0 or 1, default 0

返回:

output -- N-D with shape [MAX_LENGTH, batch_size, ...] or [batch_size, MAX_LENGTH, ...] depending on the value of axis.

返回类型:

tvm.te.Tensor

tvm.topi.shape(array, dtype='int32')

Get the shape of input array

参数:

• array (tvm.te.Tensor) -- The source tensor.

• dtype (str, optional) -- The target data type.

返回:

result -- The resulting tensor.

返回类型:

tvm.te.Tensor

tvm.topi.sigmoid(x)

Take sigmoid tanh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.sign(x)

Returns -1, 0, 1 based on sign of x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.sin(x)

Take sin of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.sinh(x)

Take sinh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.sliding_window(data, axis, window_shape, strides)

Slide a window over the data tensor.

参数:

• data (relay.Expr) -- The input data to the operator.

• axis (int) -- What axis the window begins sliding over. Window will be slid over this axis and all following axes. The axis value determines the window shape (and thus, the number of strides): window shape and strides must both be of length data.ndim-axis.

• window_shape (List[int]) -- The window shape to form over the input. Window shape must be of length data.ndim-axis.

• strides (List[int]) -- How to stride the window along each dimension. Strides must be of length data.ndim-axis.

返回:

result -- The resulting tensor.

返回类型:

relay.Expr

tvm.topi.sort(data, axis=-1, is_ascend=1)

Performs sorting along the given axis and returns an array in sorted order.

参数:

• data (tvm.te.Tensor) -- The input tensor.

• axis (int, optional) -- Axis along which to sort the input tensor. By default the flattened array is used.

• is_ascend (boolean, optional) -- Whether to sort in ascending or descending order.

• dtype (string, optional) -- DType of the output indices.

返回:

out -- Sorted index tensor.

返回类型:

tvm.te.Tensor

tvm.topi.sparse_reshape(sparse_indices, prev_shape, new_shape, new_sparse_indices_shape, new_shape_shape)

Reshape a Sparse Tensor

参数:

• sparse_indices (relay.Expr) -- A 2-D tensor[N, n_dim] of integers containing location of sparse values, where N is the number of sparse values and n_dim is the number of dimensions of the dense_shape

• prev_shape (relay.Expr) -- A 1-D tensor containing the previous shape of the dense tensor

• new_shape (relay.Expr) -- A 1-D tensor containing the new shape of the dense tensor

返回:

result -- Output tensor.

返回类型:

relay.Expr

示例

sparse_indices = [[0, 0, 0],
                    [0, 0, 1],
                    [0, 1, 0],
                    [1, 0, 0],
                    [1, 2, 3]]
prev_shape = [2, 3, 4]
new_shape = [9, -1]
new_sparse_indices, new_shape = relay.sparse_reshape(
    sparse_indices, prev_shape, new_shape)
new_sparse_indices = [[0, 0],
                      [0, 1],
                      [1, 2],
                      [4, 2],
                      [8, 1]]
new_shape = [9, 4]

tvm.topi.sparse_to_dense(sparse_indices, output_shape, sparse_values, default_value=0)

Converts a sparse representation into a dense tensor.

Example:: - sparse_to_dense([[0, 0], [1, 1]], [2, 2], [3, 3], 0) = [[3, 0], [0, 3]]

参数:

• sparse_indices (tvm.te.Tensor) -- A 0-D, 1-D, or 2-D tensor of integers containing location of sparse values.

• output_shape (A list of integers) -- Shape of the dense output tensor.

• sparse_values (tvm.te.Tensor) -- A 0-D or 1-D tensor containing the sparse values for the sparse indices.

• default_value (tvm.te.Tensor) -- A 0-D tensor containing the default value for the remaining locations. Defaults to 0.

返回:

result -- Dense tensor of shape output_shape. Has the same type as sparse_values.

返回类型:

tvm.te.Tensor

tvm.topi.split(ary, indices_or_sections, axis=0)

Split an array into multiple sub-arrays.

参数:

• ary (tvm.te.Tensor)

• indices_or_sections (int or 1-D array)

• axis (int)

返回:

ret

返回类型:

tuple of tvm.te.Tensor

tvm.topi.sqrt(x)

Take square root of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.squeeze(a, axis=None)

Remove single-dimensional entries from the shape of an array.

参数:

• a (tvm.te.Tensor)

• axis (None or int or tuple of ints, optional) -- Selects a subset of the single-dimensional entries in the shape. If an axis is selected with shape entry greater than one, an error is raised.

返回:

squeezed

返回类型:

tvm.te.Tensor

tvm.topi.stack(a, axis)

Repeats the whole array multiple times.

参数:

• a (tvm.te.Tensor) -- The tensor to be stacked.

• axis (int, optional) -- The axis in the result array along which the input arrays are stacked.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.stft(data, n_fft, hop_length, win_length, window, normalized, onesided, output_shape)

The STFT computes the Fourier transform of short overlapping windows of the input. This gives frequency components of the signal as they change over time. :param data: Either a 1-D tensor or a 2-D batch tensor. :type data: relay.Expr :param n_fft: The size of Fourier transform :type n_fft: int :param hop_length: The distance between neighboring sliding window frames :type hop_length: int :param win_length: The size of window frame and STFT filter :type win_length: int :param window: A 1-D tensor window frame :type window: relay.Expr :param normalized: Whether to return the normalized STFT results :type normalized: bool :param onesided: Whether to return onesided result or fill with conjugate symmetry :type onesided: bool

返回:

output -- Tensor containing the STFT result

返回类型:

relay.Expr

示例

data = [1, 2, 3, 4, 5, 6]
window = [4, 3, 2]
[n_fft, hop_length, win_length, normalized, onesided] = [3, 3, 3, False, True]
relay.stft(data, n_fft, hop_length, win_length, window, normalized, onesided)
-> [[[15.0000,  0.0000], [34.0000,  0.0000]], [[ 4.5000,  0.8660], [ 1.0000, -1.7321]]]

tvm.topi.strided_set(a, v, begin, end, strides=None)

Set slice of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be sliced.

• v (tvm.te.Tensor) -- The values to set

• begin (tvm.te.Tensor) -- The indices to begin with in the slicing.

• end (tvm.te.Tensor) -- Indices indicating end of the slice.

• strides (tvm.te.Tensor, optional) -- Specifies the stride values, it can be negative in that case, the input tensor will be reversed in that particular axis.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.strided_slice(a, begin, end, strides=None, axes=None, slice_mode='end', assume_inbound=True)

Slice of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be sliced.

• begin (list of int) -- The indices to begin with in the slicing.

• end (list of int) -- Indices indicating end of the slice.

• strides (list of int, optional) -- Specifies the stride values, it can be negative in that case, the input tensor will be reversed in that particular axis.

• axes (list of int, optional) -- Axes along which slicing is applied. When it is specified, begin, end strides, and axes need to a list of integers of the same length.

• slice_mode (str, optional) -- The slice mode [end, size]. end - The ending indices for the slice [default]. size - The input strides will be ignored, input end in this mode indicates the sizeof a slice starting at the location specified by begin. If end[i] is -1, all remaining elements in that dimension are included in the slice.

• assume_inbound (bool, optional) -- A flag to indicate if all indices are assumed to be inbound

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.subtract(lhs, rhs)

Subtraction with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.sum(data, axis=None, keepdims=False)

Sum of array elements over a given axis or a list of axes

参数:

• data (tvm.te.Tensor) -- The input tvm tensor

• axis (None or int or tuple of int) -- Axis or axes along which a sum is performed. The default, axis=None, will sum all of the elements of the input array. If axis is negative it counts from the last to the first axis.

• keepdims (bool) -- If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.take(a, indices, axis=None, batch_dims=0, mode='clip')

Take elements from an array along an axis.

参数:

• a (tvm.te.Tensor) -- The source array.

• indices (tvm.te.Tensor) -- The indices of the values to extract.

• axis (int, optional) -- The axis over which to select values. By default, the flattened input array is used.

• batch_dims (int) -- The number of batch dimensions. By default is 0.

• mode (str, optional) -- Specifies how out-of-bound indices will behave. clip - clip to the range (default) wrap - wrap around the indices fast - no clip or wrap around (user must make sure indices are in-bound)

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.take_legalize(attrs, inputs, types)

Legalizes dyn.topk op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current op

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.tan(x)

Take tan of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.tanh(x)

Take hyperbolic tanh of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.tensordot(a, b, axes)

A generalization of matrix multiplication to tensor.

参数:

• a (The tensor A)

• b (The tensor B)

• axes (The number of dimensions to reduce over)

返回类型:

A Tensor computing the result

tvm.topi.tile(a, reps)

Repeats the whole array multiple times.

参数:

• a (tvm.te.Tensor) -- The tensor to be tiled.

• reps (tuple of ints, required) -- The number of times for repeating the tensor

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.topk(data, k=1, axis=-1, ret_type='both', is_ascend=False, dtype='int64')

Get the top k elements in an input tensor along the given axis.

参数:

• data (tvm.te.Tensor) -- The input tensor.

• k (int or tvm.te.Tensor, optional) -- Number of top elements to select. Return all elements if k < 1.

• axis (int, optional) -- Axis long which to sort the input tensor.

• ret_type (str, optional) -- The return type [both, values, indices]. "both": return both top k data and indices. "values": return top k data only. "indices": return top k indices only.

• is_ascend (boolean, optional) -- Whether to sort in ascending or descending order.

• dtype (string, optional) -- The data type of the indices output.

返回:

out -- The computed result.

返回类型:

tvm.te.Tensor or List[tvm.te.Tensor]

tvm.topi.transpose(a, axes=None)

Permute the dimensions of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be expanded.

• axes (tuple of ints, optional) -- By default, reverse the dimensions.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.trilu(data, k, upper)

Given a 2-D matrix or batches of 2-D matrices, returns the upper or lower triangular part of the tensor.

参数:

• data (tvm.te.Tensor) -- The tensor that trilu will be applied to. Must be either a 2D matrix or a tensor of batches of 2D matrices.

• k (tvm.te.Tensor) -- The number of diagonals above or below the main diagonal to exclude or include.

• upper (bool) -- If True, only upper triangular values of input are kept, if False, the lower triangular values are kept.

返回:

ret -- The new tensor with appropriate diagonals set to zero.

返回类型:

relay.Expr

示例

x = [[0, 1, 2],
     [3, 4, 5],
     [6, 7, 8]]

relay.trilu(x, True, 0) =
    [[0, 1, 2],
     [0, 4, 5],
     [0, 0, 8]]

tvm.topi.trunc(x)

Take truncated value of the input of x, element-wise.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.unique(data, is_sorted=True, return_counts=False)

Find the unique elements of a 1-D tensor. Please note output and counts are all padded to have the same length of data and element with index >= num_unique[0] has undefined value.

参数:

• data (tvm.te.Tensor) -- A 1-D tensor of integers.

• sorted (bool) -- Whether to sort the unique elements in ascending order before returning as output.

• return_counts (bool) -- Whether to return the count of each unique element.

返回:

• unique (tvm.te.Tensor) -- A 1-D tensor containing the unique elements of the input data tensor. The same size as the input data. If there are less unique elements than input data, the end of the tensor is padded with zeros.

• indices (tvm.te.Tensor) -- A 1-D tensor. The same size as output. For each entry in output, it contains the index of its first occurence in the input data. The end of the tensor is padded with the length of the input data.

• inverse_indices (tvm.te.Tensor) -- A 1-D tensor. For each entry in data, it contains the index of that data element in the unique array. (Note that inverse_indices is very similar to indices if output is not sorted.)

• num_unique (tvm.te.Tensor) -- A 1-D tensor with size=1 containing the number of unique elements in the input data tensor.

• counts (optional) (tvm.te.Tensor) -- A 1-D tensor containing the count of each unique element in the output.

示例

[output, indices, num_unique] = unique([4, 5, 1, 2, 3, 3, 4, 5], False, False)
output          =  [4, 5, 1, 2, 3, _, _, _]
indices         =  [0, 1, 2, 3, 4, _, _, _]
inverse_indices =  [0, 1, 2, 3, 4, 4, 0, 1]
num_unique      =  [5]

[output, indices, num_unique, counts] = unique([4, 5, 1, 2, 3, 3, 4, 5], False, True)
output          =  [4, 5, 1, 2, 3, _, _, _]
indices         =  [0, 1, 2, 3, 4, _, _, _]
inverse_indices =  [0, 1, 2, 3, 4, 4, 0, 1]
num_unique      =  [5]
counts          =  [2, 2, 1, 1, 2, _, _, _]

[output, indices, num_unique] = unique([4, 5, 1, 2, 3, 3, 4, 5], True)
output          =  [1, 2, 3, 4, 5, _, _, _]
indices         =  [2, 3, 4, 0, 1, _, _, _]
inverse_indices =  [3, 4, 0, 1, 2, 2, 3, 4]
num_unique      =  [5]

tvm.topi.unravel_index(indices, shape)

Convert a flat index or array of flat indices into a tuple of coordinate arrays.

Example:: - unravel_index([22, 41, 37], [7, 6]) = [[3, 6, 6], [4, 5, 1]]

参数:

• indices (relay.Expr) -- An integer array containing indices.

• shape (relay.Expr) -- The shape of the array.

返回:

result -- The tuple of coordinate arrays.

返回类型:

relay.Expr

tvm.topi.where(condition, x, y)

Get the elements, either from x or y, depending on the condition.

参数:

• condition (tvm.te.Tensor) -- The condition array.

• x (tvm.te.Tensor) -- First array to be selected.

• y (tvm.te.Tensor) -- Second array to be selected.

返回:

result -- A Tensor selected from x or y depending on condition.

返回类型:

tvm.te.Tensor

tvm.topi.within_index(b, e, s, i)

Return a boolean value that indicates if i is within the given index.

参数:

• b (Expr) -- beginning of the index

• e (Expr) -- end of the index

• s (Expr) -- strides of index

• i (Expr) -- array position

返回:

selected -- bool expression that is True is the array position would be selected by the index and False otherwise

返回类型:

Expr

tvm.topi.nn

Neural network operators

Classes:

Workload(in_dtype, out_dtype, height, width, ...)

Functions:

adaptive_pool(data, output_size, pool_type)	Perform pooling on height and width dimension of data.
adaptive_pool1d(data, output_size, pool_type)	Perform pooling on three dimensional data.
adaptive_pool3d(data, output_size, pool_type)	Perform pooling on three dimensional data.
add(lhs, rhs)	Addition with auto-broadcasting
add_alter_layout(_attrs, _inputs, _tinfos, ...)	Change add layout.
batch_matmul(tensor_a, tensor_b[, oshape, ...])	Compute batch matrix multiplication of tensor_a and tensor_b.
batch_matmul_legalize(attrs, inputs, types)	Legalizes batch_matmul op.
batch_norm(data, gamma, beta, moving_mean, ...)	Batch normalization layer (Ioffe and Szegedy, 2014).
batch_to_space_nd(data, block_shape, ...)	Perform space to batch transformation on the data
bias_add_legalize(_attrs, _inputs, _tinfos)	Legalize bias_add layout.
binarize_pack(data[, axis, name])	Binarization and bit-packing along a certain axis.
binary_dense(data, weight)	Binary matrix multiplication using xor and bit-count.
bitpack(data, bits, pack_axis, bit_axis, ...)	Packs data into format necessary for bitserial computation
bitserial_conv2d_legalize(attrs, inputs, types)	Legalizes Bitserial Conv2D op.
bitserial_conv2d_nchw(data, kernel, stride, ...)	Bitserial Conv2D operator.
bitserial_conv2d_nhwc(data, kernel, stride, ...)	Bitserial Conv2D operator.
bitserial_dense(data, weight, data_bits, ...)	The default implementation of bitserial dense in topi.
concatenate(a_tuple[, axis])	Join a sequence of arrays along an existing axis.
conv(inp, filt, stride, padding, dilation, ...)	Convolution operator in NCHW or NHWC layout.
conv1d(data, kernel[, strides, padding, ...])	1D convolution forward operator.
conv1d_ncw(data, kernel[, strides, padding, ...])	1D convolution in NCW layout.
conv1d_nwc(data, kernel[, strides, padding, ...])	1D convolution in NWC layout.
conv1d_transpose_ncw(data, kernel, stride, ...)	Transposed 1D convolution ncw forward operator.
conv2d(input, filter, strides, padding, dilation)	Conv2D operator.
conv2d_NCHWc(data, kernel, stride, padding, ...)	Conv2D operator for nChw[x]c layout.
conv2d_NCHWc_int8(data, kernel, stride, ...)	Conv2D operator for nChw[x]c layout.
conv2d_alter_layout(attrs, inputs, tinfos, ...)	Change Conv2D layout.
conv2d_gemm_weight_transform(kernel, tile_N, ...)	Weight transformation for winograd
conv2d_hwcn(Input, Filter, stride, padding, ...)	Convolution operator in HWCN layout.
conv2d_infer_layout(workload, cfg)	Infer input/output shapes and layouts from a workload and cfg.
conv2d_legalize(attrs, inputs, types)	Legalizes Conv2D op.
conv2d_nchw(Input, Filter, stride, padding, ...)	Convolution operator in NCHW layout.
conv2d_nhwc(Input, Filter, stride, padding, ...)	Convolution operator in NHWC layout.
conv2d_transpose_alter_layout(attrs, inputs, ...)	Change Conv2D_Transpose layout.
conv2d_transpose_legalize(attrs, inputs, types)	Legalizes Transposed 2D convolution op.
conv2d_transpose_nchw(Input, Filter, ...)	Transposed 2D convolution nchw forward operator.
conv2d_transpose_nchw_preprocess(data, ...)	Preprocess data and kernel to make the compute pattern of conv2d_transpose the same as conv2d
conv2d_winograd_nchw(data, weight, strides, ...)	Conv2D Winograd in NCHW layout.
conv2d_winograd_nchw_without_weight_transform(...)	Conv2D Winograd without layout transform in NCHW layout.
conv2d_winograd_nhwc(data, weight, strides, ...)	Conv2D Winograd in NHWC layout.
conv2d_winograd_nhwc_without_weight_transform(...)	Conv2D Winograd without layout transform in NHWC layout.
conv2d_winograd_nnpack_weight_transform(...)	Weight transformation for winograd
conv2d_winograd_weight_transform(kernel, ...)	Weight transformation for winograd
conv3d_alter_layout(attrs, inputs, tinfos, ...)	Change Conv3D layout.
conv3d_ncdhw(Input, Filter, stride, padding, ...)	Conv3D operator in NCDHW layout.
conv3d_ndhwc(Input, Filter, stride, padding, ...)	Convolution operator in NDHWC layout.
conv3d_transpose_legalize(attrs, inputs, types)	Legalizes Transposed 3D convolution op.
conv3d_transpose_ncdhw(Input, Filter, ...)	Transposed 3D convolution ncdhw forward operator.
conv3d_transpose_ncdhw_preprocess(data, ...)	Preprocess data and kernel to make the compute pattern of conv3d_transpose the same as conv3d
conv3d_winograd_weight_transform(kernel, ...)	Weight transformation for 3D winograd
correlation_nchw(data1, data2, kernel_size, ...)	Correlation operator in NCHW layout.
declaration_conv2d_transpose_impl(data, ...)	Implementation of conv2d transpose
declaration_conv3d_transpose_impl(data, ...)	Implementation of conv3d transpose
deformable_conv2d_nchw(data, offset, kernel, ...)	Deformable conv2D operator in NCHW layout.
deformable_conv2d_nhwc(data, offset, kernel, ...)	Deformable conv2D operator in NHWC layout.
dense(data, weight[, bias, out_dtype, ...])	The default implementation of dense in topi.
dense_alter_layout(attrs, inputs, tinfos, ...)	Change dense layout.
dense_legalize(attrs, inputs, types)	Legalizes dense op.
dense_pack(data, weight[, bias, out_dtype])	The default implementation of dense_pack in topi.
depth_to_space(data, block_size[, layout, mode])	Perform depth to space transformation on the data
depthwise_conv2d_NCHWc(Input, Filter, ...[, ...])	Depthwise convolution NCHW[x]c forward operator.
depthwise_conv2d_backward_input_nhwc(Filter, ...)	Depthwise convolution nhwc backward wrt input operator.
depthwise_conv2d_backward_weight_nhwc(Input, ...)	Depthwise convolution nhwc backward wrt weight operator.
depthwise_conv2d_infer_layout(workload, cfg)	Infer input/output shapes and layouts from a workload and cfg.
depthwise_conv2d_nchw(Input, Filter, stride, ...)	Depthwise convolution nchw forward operator.
depthwise_conv2d_nhwc(Input, Filter, stride, ...)	Depthwise convolution nhwc forward operator.
dilate(data, strides[, dilation_value, name])	Dilate data with given dilation value (0 by default).
equal_const_int(expr, value)	Returns if expr equals value.
fast_softmax(x[, axis])	Perform softmax activation on the data.
fifo_buffer(data, buffer, axis)	FIFO buffer to enable computation reuse in CNNs with sliding indow input
flatten(data)	Flattens the input array into a 2-D array by collapsing the higher dimensions.
get_const_int(expr)	Verifies expr is integer and get the constant value.
get_const_tuple(in_tuple)	Verifies input tuple is IntImm or Var, returns tuple of int or Var.
get_pad_tuple(padding, kernel)	Common code to get the pad option
get_pad_tuple1d(padding, kernel)	Common code to get the pad option
get_pad_tuple3d(padding, kernel)	Common code to get the pad option
get_pad_tuple_generic(padding, kernel)	Common code to get the pad option
global_pool(data, pool_type[, layout])	Perform global pooling on height and width dimension of data.
group_conv1d_ncw(data, kernel[, strides, ...])	1D convolution forward operator for NCW layout.
group_conv1d_nwc(data, kernel[, strides, ...])	1D convolution forward operator for NWC layout.
group_conv1d_transpose_ncw(data, kernel, ...)	Transposed 1D group convolution ncw forward operator.
group_conv2d_nchw(Input, Filter, stride, ...)	Group convolution operator in NCHW layout.
group_conv2d_nhwc(Input, Filter, stride, ...)	Group convolution operator in NHWC layout.
group_conv2d_transpose_nchw(data, kernel, ...)	Group convolution operator in NCHW layout.
group_conv3d_transpose_ncdhw(data, kernel, ...)	Transposed group 3D convolution ncdhw forward operator.
group_norm(data, gamma, beta, num_groups, ...)	Group normalization operator.
instance_norm(data, gamma, beta, axis[, epsilon])	Instance normalization operator.
layer_norm(data, gamma, beta, axis[, epsilon])	Layer normalization operator.
layout_transform(tensor, current_layout, ...)	Transform a tensor with the current layout to the desired layout.
leaky_relu(x, alpha)	Take leaky relu of input x.
log_softmax(x[, axis])	Perform log softmax activation on the data
lrn(data, size[, axis, alpha, beta, bias])	Perform the across channels local response normalisation on the input data.
lstm(Xs, Wi, Wh[, Bi, Bh, h_init, c_init, ...])	General LSTM implemented using TE scan.
matmul(tensor_a, tensor_b[, bias, ...])	The default implementation of matmul in topi.
matmul_legalize(attrs, inputs, types)	Legalizes matmul op.
mirror_pad(data, pad_before[, pad_after, ...])	Pad Input with mirroring either symmetric or reflected.
namedtuple(typename, field_names, *[, ...])	Returns a new subclass of tuple with named fields.
nll_loss(predictions, targets, weights, ...)	Negative log likelihood loss on the input data.
pad(data, pad_before[, pad_after, ...])	Pad Input with zeros.
pool1d(data, kernel, stride, dilation, ...)	Perform pooling on width dimension of data.
pool2d(data, kernel, stride, dilation, ...)	Perform pooling on height and width dimension of data.
pool3d(data, kernel, stride, dilation, ...)	Perform pooling on depth, height and width dimension of data.
pool_grad(grads, data, kernel, stride, ...)	Gradient of pooling on height and width dimension of data.
prelu(x, slope[, axis])	PReLU.
qnn_conv2d_alter_layout(_attrs, _inputs, ...)	Change qnn.conv2d layout.
qnn_dense_alter_layout(_attrs, _inputs, ...)	Change qnn.dense layout.
qnn_requantize_alter_layout(_attrs, _inputs, ...)	Change requantize layout.
reduce(function, iterable[, initial])	Apply a function of two arguments cumulatively to the items of an iterable, from left to right.
relu(x)	Take relu of input x.
rms_norm(data, weight, axis[, epsilon])	Root mean square normalization operator.
scale_shift_nchw(Input, Scale, Shift)	Batch normalization operator in inference.
scale_shift_nchwc(Input, Scale, Shift)	Batch normalization operator in inference.
scale_shift_nhwc(Input, Scale, Shift)	Batch normalization operator in inference.
simplify(expr)	Simplify the expression if it is Expr, directly return if it is int.
simulated_dequantize(data, in_dtype[, ...])	Simulated QNN dequantize operator that mimics QNN outputs without changing datatype.
simulated_quantize(data, out_dtype[, ...])	Simulated QNN quantize operator that mimics QNN outputs without changing datatype.
softmax(x[, axis])	Perform softmax activation on the data.
softmax_common(x, axis, use_fast_exp)	The common part of softmax and fast_softmax
space_to_batch_nd(data, block_shape, ...[, ...])	Perform batch to space transformation on the data
space_to_depth(data, block_size[, layout])	Perform space to depth transformation on the data
sparse_add(dense_data, sparse_data, ...)	Computes sparse-dense addition
sparse_conv2d(dense_data, sparse_data, ...)	Computes sparse-conv2d(1*1) of data and (weight_data, weight_indices, weight_indptr)
sparse_dense(dense_data, sparse_data, ...[, ...])	Computes sparse-dense matrix multiplication of data and (weight_data, weight_indices, weight_indptr).T, if sparse_lhs=False or Computes sparse-dense matrix multiplication of (data_data, data_indices, data_indptr) and weight.T, if sparse_lhs=True
sparse_dense_alter_layout(_attrs, _inputs, ...)	Change Sparse Dense layout.
sparse_dense_sp_lhs(data_data, data_indices, ...)	Computes sparse-dense matrix multiplication of (data_data, data_indices, data_indptr) and weight.T
sparse_dense_sp_rhs(data, weight_data, ...)	Computes sparse-dense matrix multiplication of data and (weight_data, weight_indices, weight_indptr).T
sparse_transpose(sparse_data, ...)	Transpose a square sparse matrix, A is an n-by-n sparse matrix in the CSR format.
strided_slice(a, begin, end[, strides, ...])	Slice of an array.
try_get_conv2d_sparse_input(args)	Analyze the input data from the given args.
try_get_sparse_input(args)	Analyze the input data from the given args.
unpack_NCHWc_to_nchw(packed_out, out_dtype)	Unpack conv2d_NCHWc output from layout NCHWc to NCHW
upsampling(data, scale_h, scale_w[, layout, ...])	Perform upsampling on the data.
upsampling3d(data, scale_d, scale_h, scale_w)	Perform upsampling on the data.
winograd_transform_matrices(tile_size, ...)	Compute the A, B, and G transform matrices for tile_size as a tvm.Expr.

class tvm.topi.nn.Workload(in_dtype, out_dtype, height, width, in_filter, out_filter, kernel_h, kernel_w, padt, padl, padb, padr, dilation_h, dilation_w, stride_h, stride_w)

Attributes:

dilation_h	Alias for field number 12
dilation_w	Alias for field number 13
height	Alias for field number 2
in_dtype	Alias for field number 0
in_filter	Alias for field number 4
kernel_h	Alias for field number 6
kernel_w	Alias for field number 7
out_dtype	Alias for field number 1
out_filter	Alias for field number 5
padb	Alias for field number 10
padl	Alias for field number 9
padr	Alias for field number 11
padt	Alias for field number 8
stride_h	Alias for field number 14
stride_w	Alias for field number 15
width	Alias for field number 3

dilation_h

Alias for field number 12

dilation_w

Alias for field number 13

height

Alias for field number 2

in_dtype

Alias for field number 0

in_filter

Alias for field number 4

kernel_h

Alias for field number 6

kernel_w

Alias for field number 7

out_dtype

Alias for field number 1

out_filter

Alias for field number 5

padb

Alias for field number 10

padl

Alias for field number 9

padr

Alias for field number 11

padt

Alias for field number 8

stride_h

Alias for field number 14

stride_w

Alias for field number 15

width

Alias for field number 3

tvm.topi.nn.adaptive_pool(data, output_size, pool_type, layout='NCHW')

Perform pooling on height and width dimension of data.

The pooling kernel and stride sizes are automatically chosen for desired output sizes. It decides the height and width dimension according to the layout string, in which 'W' and 'H' means width and height respectively. Width and height dimension cannot be split. For example, NCHW, NCHW16c, etc. are valid for pool, while NCHW16w, NCHW16h are not. See parameter layout for more information of the layout string convention.

参数:

• data (tvm.te.Tensor) -- n-D with shape of layout

• output_size (tuple of int) -- output height and width.

• pool_type (str) -- Pool type, 'max' or 'avg'

• layout (string) -- Layout of the input data. The layout is supposed to be composed of upper cases, lower cases and numbers, where upper case indicates a dimension and the corresponding lower case with factor size indicates the split dimension. For example, NCHW16c can describe a 5-D tensor of [batch_size, channel, height, width, channel_block], in which channel_block=16 is a split of dimension channel.

返回:

output -- n-D in the same layout

返回类型:

tvm.te.Tensor

tvm.topi.nn.adaptive_pool1d(data, output_size, pool_type, layout='NCW')

Perform pooling on three dimensional data. See the two dimensional version above for details.

tvm.topi.nn.adaptive_pool3d(data, output_size, pool_type, layout='NCDHW')

Perform pooling on three dimensional data. See the two dimensional version above for details.

tvm.topi.nn.add(lhs, rhs)

Addition with auto-broadcasting

参数:

• lhs (tvm.te.Tensor or Expr) -- The left operand

• rhs (tvm.te.Tensor or Expr) -- The right operand

返回:

ret -- Returns Expr if both operands are Expr. Otherwise returns Tensor.

返回类型:

tvm.te.Tensor or Expr

tvm.topi.nn.add_alter_layout(_attrs, _inputs, _tinfos, _out_type)

Change add layout.

Add is not a QNN-specific function, but this generic exists so that bias add operations can be fused with input zero point add optimizations, which only happens if the previous operator is quantized.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.batch_matmul(tensor_a, tensor_b, oshape=None, out_dtype=None, transpose_a=False, transpose_b=True, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

Compute batch matrix multiplication of tensor_a and tensor_b.

Both tensor_a and tensor_b can be transposed. For legacy reason, we use NT format (transpose_a=False, transpose_b=True) by default.

参数:

• tensor_a (tvm.te.Tensor) -- 3-D with shape [batch, M, K] or [batch, K, M].

• tensor_b (tvm.te.Tensor) -- 3-D with shape [batch, K, N] or [batch, N, K].

• oshape (List[Optional]) -- Explicit intended output shape of the computation. Can be useful in cases with dynamic input shapes.

• out_dtype (Optional[str]) -- Specifies the output data type for mixed precision batch matmul.

• transpose_a (Optional[bool] = False) -- Whether the first tensor is in transposed format.

• transpose_b (Optional[bool] = True) -- Whether the second tensor is in transposed format.

• auto_scheduler_rewritten_layout (Optional[str] = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the tensor

返回:

output -- 3-D with shape [batch, M, N]

返回类型:

tvm.te.Tensor

tvm.topi.nn.batch_matmul_legalize(attrs, inputs, types)

Legalizes batch_matmul op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current batch_matmul

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.batch_norm(data: Tensor, gamma: Tensor, beta: Tensor, moving_mean: Tensor, moving_var: Tensor, axis: int | None = None, epsilon: float | None = None, center: bool | None = None, scale: bool | None = None, training: bool | None = None, momentum: float | None = None) → List[Tensor]

Batch normalization layer (Ioffe and Szegedy, 2014).

Normalizes the input at each batch, i.e. applies a transformation that maintains the mean activation close to 0 and the activation standard deviation close to 1.

参数:

• data (tvm.te.Tensor) -- Input to be batch-normalized.

• gamma (tvm.te.Tensor) -- Scale factor to be applied to the normalized tensor.

• beta (tvm.te.Tensor) -- Offset to be applied to the normalized tensor.

• moving_mean (tvm.te.Tensor) -- Running mean of input.

• moving_var (tvm.te.Tensor) -- Running variance of input.

• axis (int, optional, default=1) -- Specify along which shape axis the normalization should occur.

• epsilon (float, optional, default=1e-5) -- Small float added to variance to avoid dividing by zero.

• center (bool, optional, default=True) -- If True, add offset of beta to normalized tensor, If False, beta is ignored.

• scale (bool, optional, defualt=True) -- If True, scale normalized tensor by gamma. If False, gamma is ignored.

• training (bool, optional, defualt=False) -- Indicating whether it is in training mode. If True, update moving_mean and moving_var.

• momentum (float, optional, default=0.1) -- The value used for the moving_mean and moving_var update.

返回:

• output (list of tvm.te.Tensor) -- Normalized data with same shape as input

• moving_mean (tvm.te.Tensor) -- Running mean of input.

• moving_var (tvm.te.Tensor) -- Running variance of input.

tvm.topi.nn.batch_to_space_nd(data, block_shape, crop_begin_list, crop_end_list)

Perform space to batch transformation on the data

参数:

• data (tvm.te.Tensor) -- N-D Tensor with shape [batch, spatial_shape, remaining_shapes], where spatial_shape has M dimensions.

• block_size (list of ints) -- list of size [M] where M is number of spatial dims, specifies block size for each spatial dimension.

• crop_begin_list (list of ints) -- list of shape [M] where M is number of spatial dims, specifies begin crop size for each spatial dimension.

• crop_end_list (list of ints) -- list of shape [M] where M is number of spatial dims, specifies end crop size for each spatial dimension.

返回:

output

返回类型:

tvm.te.Tensor

tvm.topi.nn.bias_add_legalize(_attrs, _inputs, _tinfos)

Legalize bias_add layout.

Bias add is not a QNN-specific function, but this generic exists so that empty channels can be excised from quantized conv2d operators and folded into bias adds.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

tvm.topi.nn.binarize_pack(data, axis=None, name='PackedInput')

Binarization and bit-packing along a certain axis.

参数:

• data (tvm.te.Tensor) -- n-D input, can be any layout.

• axis (None or int) -- The axis along which to do binarization and bit-packing, default is the last axis.

• name (str, optional) -- The name prefix operators generate.

返回:

output -- n-D, the same layout as input, dtype is uint32.

返回类型:

tvm.te.Tensor

tvm.topi.nn.binary_dense(data, weight)

Binary matrix multiplication using xor and bit-count.

参数:

• data (tvm.te.Tensor) -- 2-D with shape [batch, in_dim], dtype is uint32.

• weight (tvm.te.Tensor) -- 2-D with shape [out_dim, in_dim], dtype is uint32.

返回:

output -- 2-D with shape [batch, out_dim], dtype is float32.

返回类型:

tvm.te.Tensor

tvm.topi.nn.bitpack(data, bits, pack_axis, bit_axis, pack_type, name='QuantizeInput')

Packs data into format necessary for bitserial computation

参数:

• pack_axis (int) -- index of the axis to pack in data

• bit_axis (int) -- index of axis to place bit axis in resulting packed data

tvm.topi.nn.bitserial_conv2d_legalize(attrs, inputs, types)

Legalizes Bitserial Conv2D op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.bitserial_conv2d_nchw(data, kernel, stride, padding, activation_bits, weight_bits, pack_dtype='uint32', out_dtype='int16', unipolar=True)

Bitserial Conv2D operator.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• kernel (tvm.te.Tensor) -- 4-D with shape [num_filter, in_channel, filter_height, filter_width]

• stride (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two or four ints) -- padding size, [pad_height, pad_width], [pad_top, pad_left, pad_down, pad_right]

• activation_bits (int) -- number of bits used for activations/input elements

• weight_bits (int) -- number of bits used for weight elements

• out_dtype (str) -- return type of convolution

• pack_dtype (str) -- bit packing type

• unipolar (bool) -- if binarization style is in unipolar 1/0 format, instead of bipolar -1/+1 format

返回:

output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.bitserial_conv2d_nhwc(data, kernel, stride, padding, activation_bits, weight_bits, pack_dtype='uint32', out_dtype='int16', unipolar=True)

Bitserial Conv2D operator.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• kernel (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• stride (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two or four ints) -- padding size, [pad_height, pad_width], [pad_top, pad_left, pad_down, pad_right]

• activation_bits (int) -- number of bits used for activations/input elements

• weight_bits (int) -- number of bits used for weight elements

• out_dtype (str) -- return type of convolution

• pack_dtype (str) -- bit packing type

• unipolar (bool) -- if binarization style is in unipolar 1/0 format, instead of bipolar -1/+1 format

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.bitserial_dense(data, weight, data_bits, weight_bits, pack_dtype='uint32', out_dtype='int16', unipolar=True)

The default implementation of bitserial dense in topi.

参数:

• data (tvm.te.Tensor) -- 2-D with shape [batch, in_dim]

• weight (tvm.te.Tensor) -- 2-D with shape [out_dim, in_dim] or 3-D with shape [out_dim, weight_bits, in_dim]

返回:

output -- 2-D with shape [batch, out_dim]

返回类型:

tvm.te.Tensor

tvm.topi.nn.concatenate(a_tuple, axis=0)

Join a sequence of arrays along an existing axis.

参数:

• a_tuple (tuple of tvm.te.Tensor) -- The arrays to concatenate

• axis (int, optional) -- The axis along which the arrays will be joined. Default is 0.

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv(inp: Tensor, filt: Tensor, stride: int | Sequence[int], padding: int | Sequence[int], dilation: int | Sequence[int], groups: int, data_layout: str, kernel_layout: str = '', out_dtype: str | None = None, auto_scheduler_rewritten_layout: str | None = None, meta_schedule_original_shape=None, auto_scheduler_should_rewrite_layout: bool = False)

Convolution operator in NCHW or NHWC layout.

Supports 1D, 2D, 3D, ... and grouping.

参数:

• inp (tvm.te.Tensor) -- N-D with shape [batch, in_channel, in_height, in_width, ...] in data_layout

• filt (tvm.te.Tensor) -- N-D with shape [num_filter, in_channel // groups, filter_height, filter_width, ...] in kernel_layout

• stride (int or a list/tuple of dim ints) -- (where dim=2 for NCHW, dim=1 for NCH, etc.) Stride size, or [stride_height, stride_width, ...]

• padding (int or a list/tuple of dim or 2*dim ints) -- (where dim=2 for NCHW, dim=1 for NCH, etc.) padding size, or [pad_height, pad_width, ...] for dim ints, or [pad_top, pad_left, pad_bottom, pad_right] for 2*dim ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• groups (int) -- number of groups

• data_layout (str) -- Layout of the input. N indicates batch dimension, C indicates channels, any other character indicates HW (or H or HWD for 1D and 3D).

• kernel_layout (Optional[str]) -- Layout of the filter. I indicates input channels, O indicates output channels, any other character indicates HW dimension of the filter (or H or HWD for 1D and 3D). If kernel_layout is empty, use data_layout to infer the default kernel_layout. Default kernel_layout is OIHW for NCHW data layout, HWIO for NHWC data layout.

• out_dtype (str) -- Elements are converted to this type before elementwise multiplication and summation.

• auto_scheduler_rewritten_layout (str) -- Layout from autoscheduler's layout rewritting.

• meta_schedule_original_shape (Optional[List[PrimExpr]]) -- The original shape of the input tensor.

• auto_scheduler_should_rewrite_layout (bool) -- Should auto scheduler be allowed to rewrite the layout of the filter tensor. Defaults to false. This can cause errors if used with grouped convs.

返回:

Output -- N-D with shape [batch, out_channel, out_height, out_width, ...] in data_layout

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv1d(data, kernel, strides=1, padding='VALID', dilation=1, groups=1, data_layout='NCW', kernel_layout='', out_dtype=None)

1D convolution forward operator.

参数:

• data (tvm.te.Tensor) -- 3-D input shape [batch, in_channel, in_width] for data_layout == 'NCW' and [batch, in_width, in_channel] for data_layout == 'NWC'

• kernel (tvm.te.Tensor) -- 3-D kernel with shape [num_filter, in_channel, filter_size] for kernel_layout == 'OIW' and [filter_size, in_channel, num_filter] for kernel_layout == 'WIO'

• strides (int or tuple) -- The spatial stride along width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• dilation (int or tuple) -- Dilation rate if convolution should be dilated.

• data_layout (str) -- How input data is laid out, must be one of ['NCW', 'NWC']

• kernel_layout (Optiona[str]) -- The layout of the kernel. If unspecified, use default layout. "OIW" if data_layout == "NCW", "WIO" if data_layout == "NWC".

• out_dtype (str) -- The output data type. If None then output is same type as input.

tvm.topi.nn.conv1d_ncw(data, kernel, strides=1, padding='VALID', dilation=1, out_dtype=None)

1D convolution in NCW layout. See conv() for details on parameters

tvm.topi.nn.conv1d_nwc(data, kernel, strides=1, padding='VALID', dilation=1, out_dtype=None)

1D convolution in NWC layout. See conv() for details on parameters

tvm.topi.nn.conv1d_transpose_ncw(data, kernel, stride, padding, out_dtype, output_padding)

Transposed 1D convolution ncw forward operator.

参数:

• data (tvm.te.Tensor) -- 3-D with shape [batch, in_channel, in_width]

• kernel (tvm.te.Tensor) -- 3-D with shape [in_channel, num_filter, filter_width]

• stride (ints) -- The spatial stride along width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• out_dtype (str) -- The output data type. This is used for mixed precision.

• output_padding (ints) -- Used to recover the actual output shape in case there are more than one possible shape. Must be smaller than stride.

返回:

output -- 3-D with shape [batch, out_channel, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d(input, filter, strides, padding, dilation, data_layout='NCHW', kernel_layout='', out_dtype=None)

Conv2D operator.

参数:

• input (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width] in data_layout

• filter (tvm.te.Tensor) -- 4-D with shape [num_filter, in_channel, filter_height, filter_width] in kernel_layout

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• data_layout (str) -- layout of data

• kernel_layout (Optional[str]) -- layout of kernel. If unspecified, use default layout inferred from data_layout. "OIHW" if data_layout == "NCHW", "HWIO" if data_layout == "NHWC".

返回:

output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_NCHWc(data, kernel, stride, padding, dilation, layout, out_layout, out_dtype='float32')

Conv2D operator for nChw[x]c layout.

参数:

• data (tvm.te.Tensor) -- 5-D with shape [batch, in_channel_chunk, in_height, in_width, in_channel_block]

• kernel (tvm.te.Tensor) -- 6-D with shape [num_filter_chunk, in_channel_chunk, filter_height, filter_width, in_channel_block, num_filter_block]

• stride (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• layout (str) -- Input data layout

• out_layout (str) -- Output data layout

• out_dtype (str) -- output data type

返回:

output -- 5-D with shape [batch, out_channel_chunk, out_height, out_width, out_channel_block]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_NCHWc_int8(data, kernel, stride, padding, dilation, layout, out_layout, out_dtype='int32', n_elems=4)

Conv2D operator for nChw[x]c layout.

参数:

• data (tvm.te.Tensor) -- 5-D with shape [batch, in_channel_chunk, in_height, in_width, in_channel_block]

• kernel (tvm.te.Tensor) -- 7-D with shape [num_filter_chunk, in_channel_chunk, filter_height, filter_width, in_channel_block/4, num_filter_block, 4]

• stride (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• layout (str) -- Input data layout

• out_layout (str) -- Output data layout

• out_dtype (str) -- output data type

• n_elems (int) -- numer of int8 elements accumulated

返回:

output -- 5-D with shape [batch, out_channel_chunk, out_height, out_width, out_channel_block]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_alter_layout(attrs, inputs, tinfos, out_type)

Change Conv2D layout.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.conv2d_gemm_weight_transform(kernel, tile_N, tile_K, use_scalable_vectors=False, use_sme=False)

Weight transformation for winograd

参数:

• kernel (Tensor) -- The raw kernel tensor with layout "NHWC".

• tile_N (int) -- Tile size across N axis of the weight transformation for ConvGemm. (N = OC)

• tile_K (int) -- Tile size across K axis of the weight transformation for ConvGemm. (K = KW * KH * IC)

• use_scalable_vectors (bool) -- determines if operations on scalable vectors are expected

• use_sme (bool) -- determines if SME operations on scalable vectors are expected

返回:

output -- 2-D with shape [CI*KH*KW,CO]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_hwcn(Input, Filter, stride, padding, dilation, out_dtype=None)

Convolution operator in HWCN layout.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [in_height, in_width, in_channel, batch]

• Filter (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

返回:

output -- 4-D with shape [out_height, out_width, out_channel, batch]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_infer_layout(workload, cfg)

Infer input/output shapes and layouts from a workload and cfg.

参数:

• workload (tuple) -- conv2d workload

• cfg (tuple) -- tvm.autotvm config

返回:

Output -- Input shapes and layouts, and output shapes and layouts

返回类型:

[tuple of tuple and str, tuple of tuple and str]

tvm.topi.nn.conv2d_legalize(attrs, inputs, types)

Legalizes Conv2D op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.conv2d_nchw(Input, Filter, stride, padding, dilation, out_dtype=None)

Convolution operator in NCHW layout.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• Filter (tvm.te.Tensor) -- 4-D with shape [num_filter, in_channel, filter_height, filter_width]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

返回:

Output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_nhwc(Input, Filter, stride, padding, dilation, out_dtype='float32', auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

Convolution operator in NHWC layout.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• Filter (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str = "float32",) -- The type of output tensor

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_transpose_alter_layout(attrs, inputs, tinfos, out_type)

Change Conv2D_Transpose layout.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.conv2d_transpose_legalize(attrs, inputs, types)

Legalizes Transposed 2D convolution op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current Transposed 2D convolution

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.conv2d_transpose_nchw(Input, Filter, strides, padding, out_dtype, output_padding)

Transposed 2D convolution nchw forward operator.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• Filter (tvm.te.Tensor) -- 4-D with shape [in_channel, num_filter, filter_height, filter_width]

• strides (tuple of two ints) -- The spatial stride along height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• out_dtype (str) -- The output data type. This is used for mixed precision.

• output_padding (tuple of ints) -- Used to get the right output shape for gradients

返回:

Output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_transpose_nchw_preprocess(data, kernel, strides, padding, out_dtype, output_padding)

Preprocess data and kernel to make the compute pattern of conv2d_transpose the same as conv2d

tvm.topi.nn.conv2d_winograd_nchw(data, weight, strides, padding, dilation, out_dtype, pre_computed=False, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

Conv2D Winograd in NCHW layout. This is a clean version to be used by the auto-scheduler for both CPU and GPU.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• weight (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two ints) -- padding size, or [pad_height, pad_width]

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str, optional) -- Specifies the output data type.

• pre_computed (bool) -- Whether the kernel is precomputed

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_winograd_nchw_without_weight_transform(data, weight, strides, padding, dilation, out_dtype, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

Conv2D Winograd without layout transform in NCHW layout. This is a clean version to be used by meta-schedule for both CPU and GPU.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• weight (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two ints) -- padding size, or [pad_height, pad_width]

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str, optional) -- Specifies the output data type.

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_winograd_nhwc(data, weight, strides, padding, dilation, out_dtype, pre_computed=False, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

Conv2D Winograd in NHWC layout. This is a clean version to be used by the auto-scheduler for both CPU and GPU.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• weight (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two ints) -- padding size, or [pad_height, pad_width]

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str, optional) -- Specifies the output data type.

• pre_computed (bool) -- Whether the kernel is precomputed

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_winograd_nhwc_without_weight_transform(data, weight, strides, padding, dilation, out_dtype, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

Conv2D Winograd without layout transform in NHWC layout. This is a clean version to be used by the auto-scheduler for both CPU and GPU.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• weight (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two ints) -- padding size, or [pad_height, pad_width]

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str, optional) -- Specifies the output data type.

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_winograd_nnpack_weight_transform(kernel, convolution_algorithm, out_dtype)

Weight transformation for winograd

参数:

• kernel (Tensor) -- The raw kernel tensor with layout "NCHW". Only 3x3 kernel is supported for now.

• convolution_algorithm (int) -- The convolution algorithm for Winograd NNPACK.

返回:

output -- 4-D with shape [alpha, alpha, CO, CI]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv2d_winograd_weight_transform(kernel, tile_size)

Weight transformation for winograd

参数:

• kernel (Tensor) -- The raw kernel tensor with layout "NCHW".

• tile_size (int) -- Tile size of winograd transform. e.g. 2 for F(2x2, 3x3) and 4 for F(4x4, 3x3)

返回:

output -- 4-D with shape [alpha, alpha, CO, CI]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv3d_alter_layout(attrs, inputs, tinfos, out_type)

Change Conv3D layout.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.conv3d_ncdhw(Input, Filter, stride, padding, dilation, groups, out_dtype=None)

Conv3D operator in NCDHW layout.

参数:

• Input (tvm.te.Tensor) -- 5-D with shape [batch, in_channel, in_depth, in_height, in_width]

• Filter (tvm.te.Tensor) -- 5-D with shape [num_filter, in_channel, filter_depth, filter_height, filter_width]

• stride (int or a list/tuple of three ints) -- Stride size, or [strid_depth, stride_height, stride_width]

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• dilation (int or a list/tuple of three ints) -- dilation size, or [dilation_depth, dilation_height, dilation_width]

• groups (int) -- Number of groups.

返回:

Output -- 5-D with shape [batch, out_channel, out_depth, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv3d_ndhwc(Input, Filter, stride, padding, dilation, groups, out_dtype='float32', auto_scheduler_rewritten_layout='', meta_schedule_origin_shape=None)

Convolution operator in NDHWC layout.

参数:

• Input (tvm.te.Tensor) -- 5-D with shape [batch, in_depth, in_height, in_width, in_channel]

• Filter (tvm.te.Tensor) -- 5-D with shape [filter_depth, filter_height, filter_width, in_channel, num_filter]

• stride (int or a list/tuple of three ints) -- Stride size, or [stride_depth, stride_height, stride_width]

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• dilation (int or a list/tuple of three ints) -- dilation size, or [dilation_depth, dilation_height, dilation_width]

• groups (int) -- Number of groups.

• out_dtype (str = "float32",) -- The type of output tensor

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_origin_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

Output -- 5-D with shape [batch, out_depth, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv3d_transpose_legalize(attrs, inputs, types)

Legalizes Transposed 3D convolution op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current Transposed 3D convolution

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.conv3d_transpose_ncdhw(Input, Filter, strides, padding, out_dtype, output_padding)

Transposed 3D convolution ncdhw forward operator.

参数:

• Input (tvm.te.Tensor) -- 5-D with shape [batch, in_channel, in_depth, in_height, in_width]

• Filter (tvm.te.Tensor) -- 5-D with shape [in_channel, num_filter, filter_depth, filter_height, filter_width]

• strides (int or a list/tuple of three ints) -- The spatial stride along depth,height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• out_dtype (str) -- The output data type. This is used for mixed precision.

• output_padding (tuple of ints) -- Used to get the right output shape for gradients

返回:

Output -- 5-D with shape [batch, out_channel, out_depth, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.conv3d_transpose_ncdhw_preprocess(data, kernel, strides, padding, out_dtype, output_padding)

Preprocess data and kernel to make the compute pattern of conv3d_transpose the same as conv3d

tvm.topi.nn.conv3d_winograd_weight_transform(kernel, tile_size)

Weight transformation for 3D winograd

参数:

• kernel (Tensor) -- The raw kernel tensor with layout "NCDHW".

• tile_size (int) -- Tile size of winograd transform. e.g. 2 for F(2x2, 3x3) and 4 for F(4x4, 3x3)

返回:

output -- 5-D with shape [alpha, alpha, alpha, CO, CI]

返回类型:

tvm.te.Tensor

tvm.topi.nn.correlation_nchw(data1, data2, kernel_size, max_displacement, stride1, stride2, padding, is_multiply)

Correlation operator in NCHW layout.

参数:

• data1 (tvm.te.Tensor) -- 4-D with shape [batch, channel, height, width]

• data2 (tvm.te.Tensor) -- 4-D with shape [batch, channel, height, width]

• kernel_size (int) -- Kernel size for correlation, must be an odd number

• max_displacement (int) -- Max displacement of Correlation

• stride1 (int) -- Stride for data1

• stride2 (int) -- Stride for data2 within the neightborhood centered around data1

• padding (int or a list/tuple of 2 or 4 ints) -- Padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• is_multiply (bool) -- operation type is either multiplication or substraction

返回:

Output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.declaration_conv2d_transpose_impl(data, kernel, strides, padding, out_dtype, output_padding)

Implementation of conv2d transpose

tvm.topi.nn.declaration_conv3d_transpose_impl(data, kernel, strides, padding, out_dtype, output_padding)

Implementation of conv3d transpose

tvm.topi.nn.deformable_conv2d_nchw(data, offset, kernel, strides, padding, dilation, deformable_groups, groups, out_dtype)

Deformable conv2D operator in NCHW layout.

The deformable convolution operation is described in https://arxiv.org/abs/1703.06211

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• offset (tvm.te.Tensor) -- 4-D with shape [batch, deformable_groups * filter_height * filter_width * 2, out_height, out_width].

• kernel (tvm.te.Tensor) -- 4-D with shape [num_filter, in_channel, filter_height, filter_width]

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two ints) -- padding size, or [pad_height, pad_width]

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• deformable_groups (int) -- number of deformable groups

• groups (int) -- number of groups

返回:

output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.deformable_conv2d_nhwc(data, offset, kernel, strides, padding, dilation, deformable_groups, groups, out_dtype)

Deformable conv2D operator in NHWC layout.

The deformable convolution operation is described in https://arxiv.org/abs/1703.06211

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• offset (tvm.te.Tensor) --

  4-D with shape [batch, out_height, out_width,

  deformable_groups * filter_height * filter_width * 2].

• kernel (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, num_filter]

• strides (int or a list/tuple of two ints) -- stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of two ints) -- padding size, or [pad_height, pad_width]

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• deformable_groups (int) -- number of deformable groups

• groups (int) -- number of groups

返回:

output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.dense(data, weight, bias=None, out_dtype=None, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

The default implementation of dense in topi. This is an alias of matmul_nt operator for data tensor in non-transposed format and weight tensor in transposed format.

参数:

• data (tvm.te.Tensor) -- 2-D with shape [batch, in_dim]

• weight (tvm.te.Tensor) -- 2-D with shape [out_dim, in_dim]

• bias (Optional[tvm.te.Tensor]) -- 1-D with shape [out_dim]

• out_dtype (Optional[str]) -- The output type. This is used for mixed precision.

• auto_scheduler_rewritten_layout (str = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 2-D with shape [batch, out_dim]

返回类型:

tvm.te.Tensor

tvm.topi.nn.dense_alter_layout(attrs, inputs, tinfos, out_type)

Change dense layout.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.dense_legalize(attrs, inputs, types)

Legalizes dense op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current dense

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.dense_pack(data, weight, bias=None, out_dtype=None)

The default implementation of dense_pack in topi.

参数:

• data (tvm.te.Tensor) -- 2-D with shape [batch, in_dim]

• weight (tvm.te.Tensor) -- 2-D with shape [out_dim, in_dim]

• bias (Optional[tvm.te.Tensor]) -- 1-D with shape [out_dim]

• out_dtype (Optional[str]) -- The output type. This is used for mixed precision.

返回:

output -- 2-D with shape [batch, out_dim]

返回类型:

tvm.te.Tensor

tvm.topi.nn.depth_to_space(data, block_size, layout='NCHW', mode='DCR')

Perform depth to space transformation on the data

参数:

• data (tvm.te.Tensor) -- 4-D tensor in either NCHW or NHWC layout.

• block_size (int) -- Size of blocks to compose from channel dimension.

• layout (string) -- Either NCHW or NHWC, indicating data layout.

• mode (string) -- Either DCR or CDR, indicates how channels should be accessed. In DCR, channels are interwoven in the Tensorflow style while in CDR channels are accessed sequentially as in Pytorch.

返回:

output -- Output of shape [N, C / block_size**2, H * block_size, W * block_size]

返回类型:

tvm.te.Tensor

tvm.topi.nn.depthwise_conv2d_NCHWc(Input, Filter, stride, padding, dilation, layout, out_layout, out_dtype=None)

Depthwise convolution NCHW[x]c forward operator.

参数:

• Input (tvm.te.Tensor) -- 5-D with shape [batch, in_channel_chunk, in_height, in_width, in_channel_block]

• Filter (tvm.te.Tensor) -- 6-D with shape [out_channel_chunk, 1, filter_height, filter_width, 1, out_channel_block] In NCHWc depthwise convolution, we group kernel's in_channel and channel_multiplier together then do the tiling.

• stride (tuple of two ints) -- The spatial stride along height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• layout (str) -- Input data layout

• out_layout (str) -- Output data layout

• out_dtype (str, optional) -- Output data type

返回:

Output -- 5-D with shape [batch, out_channel_chunk, out_height, out_width, out_channel_block]

返回类型:

tvm.te.Tensor

tvm.topi.nn.depthwise_conv2d_backward_input_nhwc(Filter, Out_grad, oshape, ishape, stride, padding)

Depthwise convolution nhwc backward wrt input operator.

参数:

• Filter (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, channel_multiplier]

• Out_grad (tvm.te.Tensor) -- 4-D with shape [batch, out_height, out_width, out_channel]

• stride (tuple of two ints) -- The spatial stride along height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

返回:

Output -- 4-D with shape [batch, in_height, in_width, in_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.depthwise_conv2d_backward_weight_nhwc(Input, Out_grad, oshape, fshape, stride, padding)

Depthwise convolution nhwc backward wrt weight operator.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• Out_grad (tvm.te.Tensor) -- 4-D with shape [batch, out_height, out_width, out_channel]

• stride (tuple of two ints) -- The spatial stride along height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

返回:

Output -- 4-D with shape [filter_height, filter_width, in_channel, channel_multiplier]

返回类型:

tvm.te.Tensor

tvm.topi.nn.depthwise_conv2d_infer_layout(workload, cfg)

Infer input/output shapes and layouts from a workload and cfg.

参数:

• workload (tuple) -- conv2d workload

• cfg (tuple) -- tvm.autotvm config

返回:

Output -- Input shapes and layouts, and output shapes and layouts

返回类型:

[tuple of tuple and str, tuple of tuple and str]

tvm.topi.nn.depthwise_conv2d_nchw(Input, Filter, stride, padding, dilation, out_dtype=None)

Depthwise convolution nchw forward operator.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• Filter (tvm.te.Tensor) -- 4-D with shape [in_channel, channel_multiplier, filter_height, filter_width]

• stride (int or a list/tuple of two ints) -- The spatial stride, or (stride_height, stride_width).

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str, optional) -- Output data type

返回:

Output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.depthwise_conv2d_nhwc(Input, Filter, stride, padding, dilation, kernel_layout='HWOI', out_dtype=None)

Depthwise convolution nhwc forward operator.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• Filter (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel, channel_multiplier]

• stride (tuple of two ints) -- The spatial stride along height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (str, optional) -- Output data type

返回:

Output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.dilate(data, strides, dilation_value=0.0, name='DilatedInput')

Dilate data with given dilation value (0 by default).

参数:

• data (tvm.te.Tensor) -- n-D, can be any layout.

• strides (list / tuple of n ints) -- Dilation stride on each dimension, 1 means no dilation.

• dilation_value (int/float, optional) -- Value used to dilate the input.

• name (str, optional) -- The name prefix operators generated

返回:

Output -- n-D, the same layout as data.

返回类型:

tvm.te.Tensor

tvm.topi.nn.equal_const_int(expr, value)

Returns if expr equals value.

参数:

expr (tvm.Expr) -- The input expression.

返回:

equal -- Whether they equals.

返回类型:

bool

tvm.topi.nn.fast_softmax(x, axis=-1)

Perform softmax activation on the data. Use approximation to compute exponent for faster speed.

参数:

• data (tvm.te.Tensor) -- can be any dimension

• axis (int) -- channel axis

返回:

output -- output shape is the same as input

返回类型:

tvm.te.Tensor

tvm.topi.nn.fifo_buffer(data, buffer, axis)

FIFO buffer to enable computation reuse in CNNs with sliding indow input

Compute equivalent of

concat(buffer, data, axis=axis)
.slice_axis(axis=axis,
            begin=data.shape[axis],
            end=data.shape[axis]+buffer.shape[axis])

Useful for

• Encoding explicit re-use of computation in convolution ops operated on a sliding window input

• Implementing a FIFO queue to cache intermediate results, e.g. as in Fast WaveNet.

参数:

• data (tvm.te.Tensor) -- The input data

• buffer (tvm.te.Tensor) -- Previous value of the FIFO buffer

• axis (int) -- Specify which axis should be used for buffering

返回:

result -- Updated value for the buffer

返回类型:

tvm.te.Tensor

tvm.topi.nn.flatten(data)

Flattens the input array into a 2-D array by collapsing the higher dimensions.

参数:

data (tvm.te.Tensor) -- Input array.

返回:

output -- 2-D array with collapsed higher dimensions.

返回类型:

tvm.te.Tensor

tvm.topi.nn.get_const_int(expr)

Verifies expr is integer and get the constant value.

参数:

expr (tvm.Expr or int) -- The input expression.

返回:

out_value -- The output.

返回类型:

int

tvm.topi.nn.get_const_tuple(in_tuple)

Verifies input tuple is IntImm or Var, returns tuple of int or Var.

参数:

in_tuple (tuple of Expr) -- The input.

返回:

out_tuple -- The output.

返回类型:

tuple of int

tvm.topi.nn.get_pad_tuple(padding, kernel)

Common code to get the pad option

参数:

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• kernel (tuple of int) -- Conv kernel size

返回:

• pad_top (int) -- Padding size on top

• pad_left (int) -- Padding size on left

• pad_down (int) -- Padding size on down.

• pad_right (int) -- Padding size on right.

tvm.topi.nn.get_pad_tuple1d(padding, kernel)

Common code to get the pad option

参数:

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• kernel (tuple of int) -- Conv kernel size

返回:

• pad_left (int) -- Padding size on left

• pad_right (int) -- Padding size on right.

tvm.topi.nn.get_pad_tuple3d(padding, kernel)

Common code to get the pad option

参数:

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• kernel (tuple of int) -- Conv kernel size

返回:

• pad_front (int) -- Padding size on front.

• pad_top (int) -- Padding size on top

• pad_left (int) -- Padding size on left

• pad_back (int) -- Padding size on back.

• pad_down (int) -- Padding size on down.

• pad_right (int) -- Padding size on right.

tvm.topi.nn.get_pad_tuple_generic(padding, kernel)

Common code to get the pad option

参数:

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• kernel (tuple of int) -- Conv kernel size

返回:

• pad_top (int) -- Padding size on top

• pad_down (int) -- Padding size on down.

• pad_left (int) -- Padding size on left

• pad_right (int) -- Padding size on right.

tvm.topi.nn.global_pool(data, pool_type, layout='NCHW')

Perform global pooling on height and width dimension of data.

It decides the height and width dimension according to the layout string, in which 'W' and 'H' means width and height respectively. Width and height dimension cannot be split. For example, NCHW, NCHW16c, etc. are valid for pool, while NCHW16w, NCHW16h are not. See parameter layout for more information of the layout string convention.

参数:

• data (tvm.te.Tensor) -- n-D with shape of layout

• pool_type (str) -- Pool type, 'max' or 'avg'

• layout (str) -- Layout of the input data. The layout is supposed to be composed of upper cases, lower cases and numbers, where upper case indicates a dimension and the corresponding lower case with factor size indicates the split dimension. For example, NCHW16c can describe a 5-D tensor of [batch_size, channel, height, width, channel_block], in which channel_block=16 is a split of dimension channel.

返回:

output -- n-D in same layout with height and width dimension size of 1. e.g., for NCHW, the output shape will be [batch, channel, 1, 1]

返回类型:

tvm.te.Tensor

tvm.topi.nn.group_conv1d_ncw(data, kernel, strides=1, padding='VALID', dilation=1, groups=1, out_dtype=None)

1D convolution forward operator for NCW layout.

参数:

• data (tvm.te.Tensor) -- 3-D with shape [batch, in_channel, in_width]

• kernel (tvm.te.Tensor) -- 3-D with shape [num_filter, in_channel, filter_size]

• strides (int or tuple) -- The spatial stride along width

• padding (int, tuple, or str) -- Padding size can be an integer for equal padding, a tuple of (left, right) or a string in ['VALID', 'SAME'].

• dilation (int or tuple) -- Dilation rate if convolution should be dilated.

• groups (int) -- Number of groups

• out_dtype (str) -- The output data type. If None then output is same type as input.

tvm.topi.nn.group_conv1d_nwc(data, kernel, strides=1, padding='VALID', dilation=1, groups=1, out_dtype=None)

1D convolution forward operator for NWC layout.

参数:

• data (tvm.te.Tensor) -- 3-D with shape [batch, in_width, in_channel]

• kernel (tvm.te.Tensor) -- 3-D with shape [filter_size, in_channel, num_filter]

• strides (int or tuple) -- The spatial stride along width

• padding (int, tuple, or str) -- Padding size can be an integer for equal padding, a tuple of (left, right) or a string in ['VALID', 'SAME'].

• dilation (int or tuple) -- Dilation rate if convolution should be dilated.

• groups (int) -- Number of groups

• out_dtype (str) -- The output data type. If None then output is same type as input.

tvm.topi.nn.group_conv1d_transpose_ncw(data, kernel, stride, padding, out_dtype, output_padding, groups)

Transposed 1D group convolution ncw forward operator.

参数:

• data (tvm.te.Tensor) -- 3-D with shape [batch, in_channel, in_width]

• kernel (tvm.te.Tensor) -- 3-D with shape [in_channel, num_filter, filter_width]

• stride (ints) -- The spatial stride along width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• out_dtype (str) -- The output data type. This is used for mixed precision.

• output_padding (ints) --

  Used to recover the actual output shape in case there are more

  than one possible shape. Must be smaller than stride.

  groupsint

  number of groups

返回:

output -- 3-D with shape [batch, out_channel, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.group_conv2d_nchw(Input, Filter, stride, padding, dilation, groups, out_dtype=None)

Group convolution operator in NCHW layout.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• Filter (tvm.te.Tensor) -- 4-D with shape [num_filter, in_channel // groups, filter_height, filter_width]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• groups (int) -- number of groups

• out_dtype (str) -- The output type. This is used for mixed precision.

返回:

Output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.group_conv2d_nhwc(Input, Filter, stride, padding, dilation, groups, out_dtype=None)

Group convolution operator in NHWC layout.

参数:

• Input (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel, ...]

• Filter (tvm.te.Tensor) -- 4-D with shape [filter_height, filter_width, in_channel // groups, num_filter]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• dilation (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• groups (int) -- number of groups

• out_dtype (str) -- The output type. This is used for mixed precision.

返回:

Output -- 4-D with shape [batch, out_height, out_width, out_channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.group_conv2d_transpose_nchw(data, kernel, stride, padding, out_dtype, output_padding, groups)

Group convolution operator in NCHW layout.

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• kernel (tvm.te.Tensor) -- 4-D with shape [in_channel, out_channel // groups, filter_height, filter_width]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or a list/tuple of 2 or 4 ints) -- padding size, or [pad_height, pad_width] for 2 ints, or [pad_top, pad_left, pad_bottom, pad_right] for 4 ints

• out_dtype (str) -- The output data type. This is used for mixed precision.

• output_padding (tuple of ints) -- Used to get the right output shape for gradients

• groups (int) -- number of groups

• out_dtype -- The output type. This is used for mixed precision.

返回:

Output -- 4-D with shape [batch, out_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.group_conv3d_transpose_ncdhw(data, kernel, strides, padding, out_dtype, output_padding, groups)

Transposed group 3D convolution ncdhw forward operator.

参数:

• data (tvm.te.Tensor) -- 5-D with shape [batch, in_channel, in_depth, in_height, in_width]

• kernel (tvm.te.Tensor) -- 5-D with shape [in_channel, num_filter, filter_depth, filter_height, filter_width]

• strides (int or a list/tuple of three ints) -- The spatial stride along depth,height and width

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• out_dtype (str) -- The output data type. This is used for mixed precision.

• output_padding (tuple of ints) -- Used to get the right output shape for gradients

• groups (int) -- number of groups

返回:

Output -- 5-D with shape [batch, out_channel, out_depth, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.nn.group_norm(data, gamma, beta, num_groups, channel_axis, axes, epsilon=1e-05)

Group normalization operator. It accepts fp16 and fp32 as input data type. It will cast the input to fp32 to perform the computation. The output will have the same data type as input.

参数:

• data (tvm.te.Tensor) -- N-D with shape (d_0, d_1, ..., d_{N-1})

• gamma (tvm.te.Tensor) -- 1-D with shape (r_0) where r_0 == d_{channel_axis}

• beta (tvm.te.Tensor) -- Optional, 1-D with shape (r_0) where r_0 == d_{channel_axis}

• num_groups (int) -- The number of groups

• channel_axis (int) -- The channel axis

• axes (list of int) -- Axis over the normalization applied, excluding the channel axis

• epsilon (float) -- The epsilon value to avoid division by zero.

返回:

result -- N-D with shape (d_0, d_1, ..., d_{N-1})

返回类型:

tvm.te.Tensor

tvm.topi.nn.instance_norm(data, gamma, beta, axis, epsilon=1e-05)

Instance normalization operator.

参数:

• data (tvm.te.Tensor) -- N-D with shape (d_0, d_1, ..., d_{N-1})

• gamma (tvm.te.Tensor) -- K-D with shape (r_0, r_1, ..., r_{K-1}) where K == len(axis) and d_{axis_k} == r_k

• beta (tvm.te.Tensor) -- Optional, K-D with shape (r_0, r_1, ..., r_{K-1}) where K == len(axis) and d_{axis_k} == r_k

• axis (list of int) -- Axis over the normalization applied (the axis along which the mean and variance are computed)

• epsilon (float) -- The epsilon value to avoid division by zero.

返回:

result -- N-D with shape (d_0, d_1, ..., d_{N-1})

返回类型:

tvm.te.Tensor

tvm.topi.nn.layer_norm(data, gamma, beta, axis, epsilon=1e-05)

Layer normalization operator. It accepts fp16 and fp32 as input data type. It will cast the input to fp32 to perform the computation. The output will have the same data type as input.

参数:

• data (tvm.te.Tensor) -- N-D with shape (d_0, d_1, ..., d_{N-1})

• gamma (tvm.te.Tensor) -- K-D with shape (r_0, r_1, ..., r_{K-1}) where K == len(axis) and d_{axis_k} == r_k

• beta (tvm.te.Tensor) -- Optional, K-D with shape (r_0, r_1, ..., r_{K-1}) where K == len(axis) and d_{axis_k} == r_k

• axis (list of int) -- Axis over the normalization applied

• epsilon (float) -- The epsilon value to avoid division by zero.

返回:

result -- N-D with shape (d_0, d_1, ..., d_{N-1})

返回类型:

tvm.te.Tensor

tvm.topi.nn.layout_transform(tensor: RelayExpr, current_layout: str, desired_layout: str)

Transform a tensor with the current layout to the desired layout.

E.g. layout_transform(t, "NCHW", "CNHW") --> relay.transpose(t, [1, 0, 2, 3])

参数:

• tensor (relay.Expr) -- The Tensor to transpose

• current_layout (str) -- The current layout e.g. NCHW or OIHW

• desired_layout (str) -- The desired layout, must be compatible with current_layout

返回类型:

The layout_transformed tensor.

tvm.topi.nn.leaky_relu(x, alpha)

Take leaky relu of input x.

参数:

• x (tvm.te.Tensor) -- Input argument.

• alpha (float) -- The slope for the small gradient when x < 0

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.nn.log_softmax(x, axis=-1)

Perform log softmax activation on the data

参数:

data (tvm.te.Tensor) -- N-D input data

返回:

output -- N-D output with same shape

返回类型:

tvm.te.Tensor

tvm.topi.nn.lrn(data, size, axis=1, alpha=0.0001, beta=0.75, bias=2)

Perform the across channels local response normalisation on the input data.

sum_sqr_up^i{x, y} = (bias+((alpha/size)* {sum_{j=max(0, i-size/2)}^{min(N-1,i+size/2)} (data^j{x,y})^2}))^beta output^i{x, y} = data^i{x, y}/sum_sqr_up^i{x, y} N is the number for input channels

参数:

• data (tvm.te.Tensor) -- 4-D with shape [batch, channel, height, width]

• size (int) -- normalisation window size

• axis (int) -- input data layout channel axis default value is 1 for NCHW format

• bias (float) -- offset to avoid dividing by 0

• alpha (float) -- to be divided

• beta (float) -- exponent

返回:

output -- 4-D output with same shape

返回类型:

tvm.te.Tensor

tvm.topi.nn.lstm(Xs, Wi, Wh, Bi=None, Bh=None, h_init=None, c_init=None, proj=None, p_i=None, p_f=None, p_o=None, f_act=<function sigmoid>, g_act=<function tanh>, h_act=<function tanh>, reverse=False, weight_layout: str = 'IFGO')

General LSTM implemented using TE scan.

参数:

• Xs (te.Tensor) -- Input sequence with shape (seq_len, batch_size, in_dim)

• Wi (te.Tensor) -- Input weight matrix with shape (4 * hidden_dim, in_dim). The weights are packed according to weight_layout.

• Wh (te.Tensor) -- Hidden weight matrix with shape (4 * hidden_dim, hidden_dim or proj_dim). Packed as Wh.

• Bi (te.Tensor, optional) -- Input bias with shape (4 * hidden_dim,), by default None. Packed as Wh.

• Bh (te.Tensor, optional) -- Hidden bias with shape as Bi, by default None. Packed as Wh.

• h_init (te.Tensor, optional) -- Initial hidden state with shape (batch_size, hidden_dim or proj_dim), zero if None

• c_init (te.Tensor, optional) -- Initial cell state with same shape as h_init, zero if None

• proj (te.Tensor, optional) -- Projection matrix with shape (proj_dim, hidden_dim), by default None

• p_i (te.Tensor, optional) -- Peephole LSTM matrices with shape (batch_size, hidden_dim), by default None

• p_f (te.Tensor, optional) -- Peephole LSTM matrices with shape (batch_size, hidden_dim), by default None

• p_o (te.Tensor, optional) -- Peephole LSTM matrices with shape (batch_size, hidden_dim), by default None

• f_act (F, optional) -- Gate activation functions

• g_act (F, optional) -- Gate activation functions

• h_act (F, optional) -- Gate activation functions

• reverse (bool, optional) -- Whether to process Xs in reverse, by default False

• weight_layout (str, optional) -- The packed weight layout for gates, by default "IFGO". Note: I = input, F = forget, G = cell, O = output.

返回:

result -- Tuple of hidden states (with shape (seq_len, batch_size, hidden_dim or proj_dim)), and cell states (with shape (seq_len, batch_size, hidden_dim)).

返回类型:

te.Tensor, te.Tensor

tvm.topi.nn.matmul(tensor_a, tensor_b, bias=None, out_dtype=None, transpose_a=False, transpose_b=False, auto_scheduler_rewritten_layout='', meta_schedule_original_shape=None)

The default implementation of matmul in topi.

参数:

• tensor_a (tvm.te.Tensor) -- 2-D with shape [batch, in_dim]

• tensor_b (tvm.te.Tensor) -- 2-D with shape [out_dim, in_dim]

• bias (Optional[tvm.te.Tensor]) -- 1-D with shape [out_dim]

• out_dtype (Optional[str]) -- The output type. This is used for mixed precision.

• transpose_a (Optional[bool] = False) -- Whether the tensor_a is in transposed format.

• transpose_b (Optional[bool] = False) -- Whether the tensor_b is in transposed format.

• auto_scheduler_rewritten_layout (Optional[str] = "") -- The layout after auto-scheduler's layout rewrite pass.

• meta_schedule_original_shape (Optional[List[PrimExpr]] = None) -- The original shape of the input tensor.

返回:

output -- 2-D with shape [batch, out_dim]

返回类型:

tvm.te.Tensor

tvm.topi.nn.matmul_legalize(attrs, inputs, types)

Legalizes matmul op.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current matmul

• inputs (list of tvm.relay.Expr) -- The args of the Relay expr to be legalized

• types (list of types) -- List of input and output types

返回:

result -- The legalized expr

返回类型:

tvm.relay.Expr

tvm.topi.nn.mirror_pad(data, pad_before, pad_after=None, mode='SYMMETRIC', name='MirrorPadInput')

Pad Input with mirroring either symmetric or reflected.

参数:

• data (tvm.te.Tensor) -- n-D input, can be any layout.

• pad_before (list / tuple of n ints) -- Pad width on each dimension to pad the before the axis begin.

• pad_after (list / tuple of n ints, optional) -- Pad width each dimension to pad the after the axis end.

• mode (str, optional) -- Type of mirror padding to apply. Must be SYMMETRIC or REFLECT

• name (str, optional) -- The name prefix operators generated

返回:

Output -- n-D, the same layout as Input.

返回类型:

tvm.te.Tensor

tvm.topi.nn.namedtuple(typename, field_names, *, rename=False, defaults=None, module=None)

Returns a new subclass of tuple with named fields.

>>> Point = namedtuple('Point', ['x', 'y'])
>>> Point.__doc__                   # docstring for the new class
'Point(x, y)'
>>> p = Point(11, y=22)             # instantiate with positional args or keywords
>>> p[0] + p[1]                     # indexable like a plain tuple
33
>>> x, y = p                        # unpack like a regular tuple
>>> x, y
(11, 22)
>>> p.x + p.y                       # fields also accessible by name
33
>>> d = p._asdict()                 # convert to a dictionary
>>> d['x']
11
>>> Point(**d)                      # convert from a dictionary
Point(x=11, y=22)
>>> p._replace(x=100)               # _replace() is like str.replace() but targets named fields
Point(x=100, y=22)

tvm.topi.nn.nll_loss(predictions, targets, weights, reduction, ignore_index)

Negative log likelihood loss on the input data.

output{n, i_1, i_2, ..., i_k} = -p * w

where t = target{n, i_1, i_2, ..., i_k}

p = predictions{n, t, i_1, i_2, i_k} w = weights{n, i_1, i_2, ..., i_k} if t != ignore_index else 0

result = reduction(output)

参数:

• predictions (tvm.te.Tensor) -- (k+2)-D with shape (N, C, d_1, d_2, ..., d_k), where C is the number of target classes

• targets (tvm.te.Tensor) -- (k+1)-D with shape (N, d_1, d_2, ..., d_k) The target value of the input.

• weights (tvm.te.Tensor) -- 1-D with shape (C,) The weight of each target value.

• reduction (string) -- The reduction method to apply to output. Can be "mean", "sum" or "none".

• ignore_index (int) -- The target value to ignore.

返回:

output -- a scalar if the reduction type is "mean" or "sum", otherwise the same shape as target.

返回类型:

tvm.te.Tensor

tvm.topi.nn.pad(data, pad_before, pad_after=None, pad_value=0.0, name='PadInput', attrs=None)

Pad Input with zeros.

参数:

• data (tvm.te.Tensor) -- n-D input, can be any layout.

• pad_before (list / tuple of n ints) -- Pad width on each dimension to pad the before the axis begin.

• pad_after (list / tuple of n ints, optional) -- Pad width each dimension to pad the after the axis end.

• pad_value (float, optional) -- The value to be padded.

• name (str, optional) -- The name prefix operators generated

返回:

Output -- n-D, the same layout as Input.

返回类型:

tvm.te.Tensor

tvm.topi.nn.pool1d(data, kernel, stride, dilation, padding, pool_type, ceil_mode=False, layout='NCW', count_include_pad=True)

Perform pooling on width dimension of data.

Width axis is determined according to the layout string. in which 'w' means width. Width dimension cannot be split. For example, NCW, NCW16c, etc. are valid for pool, while NCW16w is not. See parameter layout for more information of the layout string convention.

参数:

• data (tvm.te.Tensor) -- n-D with shape of layout

• kernel (list/tuple of one int or int) -- Kernel size, [kernel_width]

• stride (list/tuple of one int or int) -- Stride size, [stride_width]

• dilation (list/tuple of two ints) -- Dilation size, [dilation_height, dilation_width]

• padding (list/tuple of two ints) -- Pad size, [pad_left, pad_right]

• pool_type (str) -- Pool type, 'max' or 'avg'

• ceil_mode (bool) -- Whether to use ceil when calculating output size.

• layout (string) -- Layout of the input data. The layout is supposed to be composed of upper cases, lower cases and numbers, where upper case indicates a dimension and the corresponding lower case with factor size indicates the split dimension. For example, NCW16c can describe a 4-D tensor of [batch_size, channel, width, channel_block], in which channel_block=16 is a split of dimension channel.

• count_include_pad (bool) -- Whether include padding in the calculation when pool_type is 'avg'

返回:

output -- n-D in the same layout

返回类型:

tvm.te.Tensor

tvm.topi.nn.pool2d(data, kernel, stride, dilation, padding, pool_type, ceil_mode=False, layout='NCHW', count_include_pad=True)

Perform pooling on height and width dimension of data.

It decides the height and width dimension according to the layout string, in which 'W' and 'H' means width and height respectively. Width and height dimension cannot be split. For example, NCHW, NCHW16c, etc. are valid for pool, while NCHW16w, NCHW16h are not. See parameter layout for more information of the layout string convention.

参数:

• data (tvm.te.Tensor) -- n-D with shape of layout

• kernel (list/tuple of two ints) -- Kernel size, [kernel_height, kernel_width]

• stride (list/tuple of two ints) -- Stride size, [stride_height, stride_width]

• dilation (list/tuple of two ints) -- Dilation size, [dilation_height, dilation_width]

• padding (list/tuple of four ints) -- Pad size, [pad_top, pad_left, pad_bottom, pad_right]]

• pool_type (str) -- Pool type, 'max' or 'avg'

• ceil_mode (bool) -- Whether to use ceil when calculating output size.

• layout (string) -- Layout of the input data. The layout is supposed to be composed of upper cases, lower cases and numbers, where upper case indicates a dimension and the corresponding lower case with factor size indicates the split dimension. For example, NCHW16c can describe a 5-D tensor of [batch_size, channel, height, width, channel_block], in which channel_block=16 is a split of dimension channel.

• count_include_pad (bool) -- Whether include padding in the calculation when pool_type is 'avg'

返回:

output -- n-D in the same layout

返回类型:

tvm.te.Tensor

tvm.topi.nn.pool3d(data, kernel, stride, dilation, padding, pool_type, ceil_mode=False, layout='NCDHW', count_include_pad=True)

Perform pooling on depth, height and width dimension of data.

It decides the depth, height and width dimension according to the layout string, in which 'D', 'W' and 'H' means depth, width and height respectively. Depth, width and height dimension cannot be split. For example, NCDHW, NCDHW16c, etc. are valid for pool, while NCDHW16d, NCDHW16w, NCDHW16h are not. See parameter layout for more information of the layout string convention.

参数:

• data (tvm.te.Tensor) -- n-D with shape of layout

• kernel (list/tuple of three ints) -- Kernel size, [kernel_depth, kernel_height, kernel_width]

• stride (list/tuple of three ints) -- Stride size, [stride_depth, stride_height, stride_width]

• dilation (list/tuple of two ints) -- Dilation size, [dilation_height, dilation_width]

• padding (list/tuple of six ints) -- Pad size, [pad_front, pad_top, pad_left, pad_back, pad_bottom, pad_right]

• pool_type (str) -- Pool type, 'max' or 'avg'

• ceil_mode (bool) -- Whether to use ceil when calculating output size.

• layout (string) -- Layout of the input data. The layout is supposed to be composed of upper cases, lower cases and numbers, where upper case indicates a dimension and the corresponding lower case with factor size indicates the split dimension. For example, NCDHW16c can describe a 6-D tensor of [batch_size, channel, depth, height, width, channel_block], in which channel_block=16 is a split of dimension channel.

• count_include_pad (bool) -- Whether include padding in the calculation when pool_type is 'avg'

返回:

output -- n-D in the same layout

返回类型:

tvm.te.Tensor

tvm.topi.nn.pool_grad(grads, data, kernel, stride, padding, pool_type, ceil_mode=False, count_include_pad=True, layout='NCHW')

Gradient of pooling on height and width dimension of data.

It decides the height and width dimension according to the layout string, in which 'W' and 'H' means width and height respectively. Width and height dimension cannot be split. For example, NCHW, NCHW16c, etc. are valid for pool, while NCHW16w, NCHW16h are not. See parameter layout for more information of the layout string convention.

参数:

• grads (tvm.te.Tensor) -- n-D with shape of layout

• data (tvm.te.Tensor) -- n-D with shape of layout

• kernel (list/tuple of two ints) -- Kernel size, [kernel_height, kernel_width]

• stride (list/tuple of two ints) -- Stride size, [stride_height, stride_width]

• padding (list/tuple of four ints) -- Pad size, [pad_top, pad_left, pad_bottom, pad_right]]

• pool_type (str) -- Pool type, 'max' or 'avg'

• ceil_mode (bool) -- Whether to use ceil when calculating output size.

• count_include_pad (bool) -- Whether include padding in the calculation when pool_type is 'avg'

• layout (string) -- Layout of the input data. The layout is supposed to be composed of upper cases, lower cases and numbers, where upper case indicates a dimension and the corresponding lower case with factor size indicates the split dimension. For example, NCHW16c can describe a 5-D tensor of [batch_size, channel, height, width, channel_block], in which channel_block=16 is a split of dimension channel.

返回:

output -- n-D in the same layout

返回类型:

tvm.te.Tensor

tvm.topi.nn.prelu(x, slope, axis=1)

PReLU. It accepts two arguments: an input x and a weight array W and computes the output as \(PReLU(x) y = x > 0 ? x : W * x\), where \(*\) is an elementwise multiplication for each sample in the batch.

参数:

• x (tvm.te.Tensor) -- Input argument.

• slope (tvm.te.Tensor) -- Channelised slope tensor for prelu

• axis (int) -- The axis where the channel data needs to be applied

返回:

• y (tvm.te.Tensor) -- The result.

• Links

• -----

• [http (//arxiv.org/pdf/1502.01852v1.pdf])

tvm.topi.nn.qnn_conv2d_alter_layout(_attrs, _inputs, _tinfos, _out_type)

Change qnn.conv2d layout.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.qnn_dense_alter_layout(_attrs, _inputs, _tinfos, _out_type)

Change qnn.dense layout. Not to change by default

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current dense op

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

tvm.topi.nn.qnn_requantize_alter_layout(_attrs, _inputs, _tinfos, _out_type)

Change requantize layout.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.reduce(function, iterable[, initial]) → value

Apply a function of two arguments cumulatively to the items of an iterable, from left to right.

This effectively reduces the iterable to a single value. If initial is present, it is placed before the items of the iterable in the calculation, and serves as a default when the iterable is empty.

For example, reduce(lambda x, y: x+y, [1, 2, 3, 4, 5]) calculates ((((1 + 2) + 3) + 4) + 5).

tvm.topi.nn.relu(x)

Take relu of input x.

参数:

x (tvm.te.Tensor) -- Input argument.

返回:

y -- The result.

返回类型:

tvm.te.Tensor

tvm.topi.nn.rms_norm(data, weight, axis, epsilon=1e-05)

Root mean square normalization operator. The output will have the same data type as input.

参数:

• data (tvm.te.Tensor) -- N-D with shape (d_0, d_1, ..., d_{N-1})

• weight (tvm.te.Tensor) -- K-D with shape (r_0, r_1, ..., r_{K-1}) where K == len(axis) and d_{axis_k} == r_k

• axis (list of int) -- Axis over the normalization applied

• epsilon (float) -- The epsilon value to avoid division by zero.

返回:

result -- N-D with shape (d_0, d_1, ..., d_{N-1})

返回类型:

tvm.te.Tensor

tvm.topi.nn.scale_shift_nchw(Input, Scale, Shift)

Batch normalization operator in inference.

参数:

• Input (tvm.te.Tensor) -- 4-D input tensor, NCHW layout [batch, channel, height, width]

• Scale (tvm.te.Tensor) -- Scale tensor, 1-D of size channel number

• Shift (tvm.te.Tensor) -- Shift tensor, 1-D of size channel number

返回:

Output -- Output tensor, layout is NCHW

返回类型:

tvm.te.Tensor

tvm.topi.nn.scale_shift_nchwc(Input, Scale, Shift)

Batch normalization operator in inference.

参数:

• Input (tvm.te.Tensor) -- 5-D input tensor, NCHWc layout [batch, channel_chunk, height, width, channel_block]

• Scale (tvm.te.Tensor) -- Scale tensor, 2-D of size [channel_chunk, channel_block]

• Shift (tvm.te.Tensor) -- Shift tensor, 2-D of size [channel_chunk, channel_block]

返回:

Output -- Output tensor, layout is NHWC

返回类型:

tvm.te.Tensor

tvm.topi.nn.scale_shift_nhwc(Input, Scale, Shift)

Batch normalization operator in inference.

参数:

• Input (tvm.te.Tensor) -- 4-D input tensor, NHWC layout [batch, height, width, channel]

• Scale (tvm.te.Tensor) -- Scale tensor, 1-D of size channel number

• Shift (tvm.te.Tensor) -- Shift tensor, 1-D of size channel number

返回:

Output -- Output tensor, layout is NHWC

返回类型:

tvm.te.Tensor

tvm.topi.nn.simplify(expr)

Simplify the expression if it is Expr, directly return if it is int.

参数:

expr (Expr or int) -- The input.

返回:

out -- The simplified output

返回类型:

Expr or int

tvm.topi.nn.simulated_dequantize(data, in_dtype, input_scale=None, input_zero_point=None, axis=-1)

Simulated QNN dequantize operator that mimics QNN outputs without changing datatype. The benefit of this operator over true QNN dequantize is that this operator allows dynamic datatype selection and can operate on both per-channel and scalar scales and zero points while QNN dequantize requires both of these to be fixed at compile time.

参数:

• data (tvm.te.Tensor) -- An N-D input tensor to the operator.

• in_dtype (tvm.te.Tensor) -- A scalar variable that indicates which datatype to simulate dequantization with. Use SQNN_DTYPE_TO_CODE to convert a dtype string into the corresponding variable value.

• input_scale (tvm.te.Tensor, optional) -- A scalar tensor representing the scale to use when dequantizing from integer datatypes. When it contains more than a single value, N must match the number of channels in data.

• input_zero_point (tvm.te.Tensor, optional) -- A 1-D tensor representing the zero point to use when dequantizing from integer datatypes. When it contains more than a single value, N must match the number of channels in data.

• axis (int, optional) -- The channel axis for quantization. Default value is -1 which corresponds to the last axis.

tvm.topi.nn.simulated_quantize(data, out_dtype, output_scale=None, output_zero_point=None, axis=-1)

Simulated QNN quantize operator that mimics QNN outputs without changing datatype. The benefit of this operator over true QNN quantize is that this operator allows dynamic datatype selection and can operate on both per-channel and scalar scales and zero points while QNN quantize requires both of these to be fixed at compile time.

参数:

• data (tvm.te.Tensor) -- An N-D input tensor to the operator.

• out_dtype (tvm.te.Tensor) -- A scalar variable that indicates which datatype to simulate quantization with. Use SQNN_DTYPE_TO_CODE to convert a dtype string into the corresponding variable value.

• output_scale (tvm.te.Tensor, optional) -- A scalar tensor representing the scale to use when quantizing to integer datatypes. When it contains more than a single value, N must match the number of channels in data.

• output_zero_point (tvm.te.Tensor, optional) -- A 1-D tensor representing the zero point to use when quantizing to integer datatypes. When it contains more than a single value, N must match the number of channels in data.

• axis (int, optional) -- The channel axis for quantization. Default value is -1 which corresponds to the last axis.

tvm.topi.nn.softmax(x, axis=-1)

Perform softmax activation on the data.

参数:

• data (tvm.te.Tensor) -- can be any dimension

• axis (int) -- channel axis

返回:

output -- output shape is the same as input

返回类型:

tvm.te.Tensor

tvm.topi.nn.softmax_common(x, axis, use_fast_exp)

The common part of softmax and fast_softmax

tvm.topi.nn.space_to_batch_nd(data, block_shape, pad_before, pad_after, pad_value=0.0)

Perform batch to space transformation on the data

参数:

• data (tvm.te.Tensor) -- N-D Tensor with shape [batch, spatial_shape, remaining_shapes], where spatial_shape has M dimensions.

• block_shape (list of ints) -- list of size [M] where M is number of spatial dims, specifies block size for each spatial dimension.

• pad_before (list of ints) -- list of shape [M] where M is number of spatial dims, specifies zero-padding size before each spatial dimension.

• pad_after (list of ints) -- list of shape [M] where M is number of spatial dims, specifies zero-padding size after each spatial dimension.

• pad_value (float, optional) -- The value used for padding.

返回:

output

返回类型:

tvm.te.Tensor

tvm.topi.nn.space_to_depth(data, block_size, layout='NCHW')

Perform space to depth transformation on the data

参数:

• data (tvm.te.Tensor) -- 4-D tensor in either NCHW or NHWC layout.

• block_size (int) -- Size of blocks to decompose into channel dimension.

• layout (string) -- Either NCHW or NHWC, indicating data layout.

返回:

output -- Output of shape [N, C * block_size**2, H / block_size, W / block_size]

返回类型:

tvm.te.Tensor

tvm.topi.nn.sparse_add(dense_data, sparse_data, sparse_indices, sparse_indptr)

Computes sparse-dense addition

参数:

• dense_data (tvm.te.Tensor) -- 2-D with shape [M, N]

• sparse_data (tvm.te.Tensor) -- 1-D with shape [nnz] (CSR)

• sparse_indices (tvm.te.Tensor) -- 1-D with shape [nnz] (CSR)

• sparse_indptr (tvm.te.Tensor) -- 1-D with shape [M + 1] (CSR)

返回:

output -- 2-D with shape [M, N]

返回类型:

tvm.te.Tensor

tvm.topi.nn.sparse_conv2d(dense_data, sparse_data, sparse_indices, sparse_indptr, layout='NHWC', kernel_size=1)

Computes sparse-conv2d(1*1) of data and (weight_data, weight_indices, weight_indptr)

参数:

• dense_data (tvm.te.Tensor) --

  4-D with shape [M, H, W, K] (layout=NHWC)

  4-D with shape [M, K, H, W] (layout=NCHW)

• sparse_data (tvm.te.Tensor) --

  2-D with shape [num_blocks, bs_r] (BSR)

  3-D with shape [num_blocks, bs_r, bs_c] (BSR)

• sparse_indices (tvm.te.Tensor) -- 1-D with shape [num_blocks] (BSR)

• sparse_indptr (tvm.te.Tensor) -- 1-D with shape [(N + 1) // bs_r] (BSR)

• layout (str) -- layout of data

返回:

output -- 4-D with shape [M, H, W, N] (layout=NHWC) 4-D with shape [M, N, H ,W] (layout=NCHW)

返回类型:

tvm.te.Tensor

tvm.topi.nn.sparse_dense(dense_data, sparse_data, sparse_indices, sparse_indptr, sparse_lhs=False)

Computes sparse-dense matrix multiplication of data and (weight_data, weight_indices, weight_indptr).T, if sparse_lhs=False or Computes sparse-dense matrix multiplication of (data_data, data_indices, data_indptr) and weight.T, if sparse_lhs=True

参数:

• dense_data (tvm.te.Tensor) -- 2-D with shape [M, K]

• sparse_data (tvm.te.Tensor) -- 1-D with shape [nnz] (CSR) or 3-D with shape [num_blocks, bs_r, bs_c] (BSR)

• sparse_indices (tvm.te.Tensor) -- 1-D with shape [nnz] (CSR) or 1-D with shape [num_blocks] (BSR)

• sparse_indptr (tvm.te.Tensor) -- 1-D with shape [N + 1] (CSR) or 1-D with shape [(N + 1) // bs_r] (BSR)

• sparse_lhs (bool, optional) -- Indicates whether lhs or rhs matrix is sparse. Default value is False.

返回:

output -- 2-D with shape [M, N]

返回类型:

tvm.te.Tensor

tvm.topi.nn.sparse_dense_alter_layout(_attrs, _inputs, _tinfos, _out_type)

Change Sparse Dense layout.

This is used for modifying the inputs weights so they are more amenable for the target.

参数:

• attrs (tvm.ir.Attrs) -- Attributes of current convolution

• inputs (tvm.relay.Expr) -- Grouped input symbols

• tinfos (list) -- Input shape and dtype

• out_type (type) -- The output type

备注

Unlike other TOPI functions, this function operates on both graph level and operator level.

tvm.topi.nn.sparse_dense_sp_lhs(data_data, data_indices, data_indptr, weight)

Computes sparse-dense matrix multiplication of (data_data, data_indices, data_indptr) and weight.T

参数:

• data_data -- 1-D with shape [nnz] (CSR) or 3-D with shape [num_blocks, bs_r, bs_c] (BSR)

• data_indices -- 1-D with shape [nnz] (CSR) or 1-D with shape [num_blocks] (BSR)

• data_indptr -- 1-D with shape [M + 1] (CSR) or 1-D with shape [(M + 1) // bs_r] (BSR)

• weight -- 2-D with shape [N, K]

返回:

output -- 2-D with shape [M, N]

返回类型:

tvm.te.Tensor

tvm.topi.nn.sparse_dense_sp_rhs(data, weight_data, weight_indices, weight_indptr)

Computes sparse-dense matrix multiplication of data and (weight_data, weight_indices, weight_indptr).T

参数:

• data (tvm.te.Tensor) -- 2-D with shape [M, K]

• weight_data (tvm.te.Tensor) -- 1-D with shape [nnz] (CSR) or 3-D with shape [num_blocks, bs_r, bs_c] (BSR)

• weight_indices (tvm.te.Tensor) -- 1-D with shape [nnz] (CSR) or 1-D with shape [num_blocks] (BSR)

• weight_indptr (tvm.te.Tensor) -- 1-D with shape [N + 1] (CSR) or 1-D with shape [(N + 1) // bs_r] (BSR)

返回:

output -- 2-D with shape [M, N]

返回类型:

tvm.te.Tensor

tvm.topi.nn.sparse_transpose(sparse_data, sparse_indices, sparse_indptr)

Transpose a square sparse matrix, A is an n-by-n sparse matrix in the CSR format. ** Currently only support Square Matrices **

参数:

• sparse_data (tvm.te.Tensor) -- 1-D with shape [nonzeros]

• sparse_indices (tvm.te.Tensor) -- 1-D with shape [nonzeros], dtype of 'int32'

• sparse_indptr (tvm.te.Tensor) -- 1-D with shape [n+1], dtype of 'int32'

返回:

• out_data (tvm.te.Tensor) -- 1-D with shape [nonzeros]

• out_indices (tvm.te.Tensor) -- 1-D with shape [nonzeros], dtype of 'int32'

• out_indptr (tvm.te.Tensor) -- 1-D with shape [n+1], dtype of 'int32'

tvm.topi.nn.strided_slice(a, begin, end, strides=None, axes=None, slice_mode='end', assume_inbound=True)

Slice of an array.

参数:

• a (tvm.te.Tensor) -- The tensor to be sliced.

• begin (list of int) -- The indices to begin with in the slicing.

• end (list of int) -- Indices indicating end of the slice.

• strides (list of int, optional) -- Specifies the stride values, it can be negative in that case, the input tensor will be reversed in that particular axis.

• axes (list of int, optional) -- Axes along which slicing is applied. When it is specified, begin, end strides, and axes need to a list of integers of the same length.

• slice_mode (str, optional) -- The slice mode [end, size]. end - The ending indices for the slice [default]. size - The input strides will be ignored, input end in this mode indicates the sizeof a slice starting at the location specified by begin. If end[i] is -1, all remaining elements in that dimension are included in the slice.

• assume_inbound (bool, optional) -- A flag to indicate if all indices are assumed to be inbound

返回:

ret

返回类型:

tvm.te.Tensor

tvm.topi.nn.try_get_conv2d_sparse_input(args)

Analyze the input data from the given args.

参数:

args (List[Tensor]) -- Input/output Tensor of a TVM subgraph.

返回:

Map from the input Tensor to its buffer name.

返回类型:

Dict[Tensor, str]

备注

The buffer name is specially designed, and these buffer should be provided in SearchTask(..., task_inputs={...}).

tvm.topi.nn.try_get_sparse_input(args)

Analyze the input data from the given args.

参数:

args (List[Tensor]) -- Input/output Tensor of a TVM subgraph.

返回:

Map from the input Tensor to its buffer name.

返回类型:

Dict[Tensor, str]

备注

The buffer name is specially designed, and these buffer should be provided in SearchTask(..., task_inputs={...}).

tvm.topi.nn.unpack_NCHWc_to_nchw(packed_out, out_dtype)

Unpack conv2d_NCHWc output from layout NCHWc to NCHW

参数:

• packed_out (tvm.te.Tensor) -- The output tensor of conv2d_NCHWc.

• out_dtype (str) -- The output dtype.

返回:

unpacked_out -- The unpacked output tensor in NCHW layout.

返回类型:

tvm.te.Tensor

tvm.topi.nn.upsampling(data, scale_h, scale_w, layout='NCHW', method='nearest_neighbor', align_corners=False, output_shape=None)

Perform upsampling on the data.

Nearest neighbor and bilinear upsampling are supported.

参数:

• inputs (tvm.te.Tensor) -- inputs is a 4-D tensor with shape [batch, channel, in_height, in_width] or [batch, in_height, in_width, channel]

• scale_h (float) -- Scaling factor for height

• scale_w (float) -- Scaling factor for width

• layout (string, optional) -- either "NCHW" or "NHWC"

• method ({"bilinear", "nearest_neighbor", "bicubic"}) -- Method to be used for upsampling.

• output_shape (tvm.tir.container.Array, optional) -- Shape to return. If left None will be inferred (If shape is determined dynamically, pass out_dtype.shape as output_shape)

返回:

output -- 4-D with shape [batch, channel, in_height*scale_h, in_width*scale_w] or [batch, in_height*scale, in_width*scale, channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.upsampling3d(data, scale_d, scale_h, scale_w, layout='NCDHW', method='nearest_neighbor', coordinate_transformation_mode='half_pixel', output_shape=None)

Perform upsampling on the data.

Nearest neighbor and bilinear upsampling are supported.

参数:

• inputs (tvm.te.Tensor) -- inputs is a 5-D tensor with shape [batch, channel, in_depth, in_height, in_width] or [batch, in_depth, in_height, in_width, channel]

• scale_d (float) -- Scaling factor for depth

• scale_h (float) -- Scaling factor for height

• scale_w (float) -- Scaling factor for width

• layout (string, optional) -- either "NCDHW" or "NDHWC"

• method ({"trilinear", "nearest_neighbor"}) -- Method to be used for upsampling.

• coordinate_transformation_mode (string, optional) -- Describes how to transform the coordinate in the resized tensor to the coordinate in the original tensor. Refer to the ONNX Resize operator specification for details. Available options are "half_pixel", "align_corners" and "asymmetric".

• output_shape (tvm.tir.container.Array, optional) -- Shape to return. If left None will be inferred (If shape is determined dynamically, pass out_dtype.shape as output_shape)

返回:

output -- 5-D with shape [batch, channel, in_depth*scale, in_height*scale, in_width*scale] or [batch, in_depth*scale, in_height*scale, in_width*scale, channel]

返回类型:

tvm.te.Tensor

tvm.topi.nn.winograd_transform_matrices(tile_size, kernel_size, out_dtype)

Compute the A, B, and G transform matrices for tile_size as a tvm.Expr.

tvm.topi.image

IMAGE network operators

Functions:

affine_grid(data, target_shape)	affine_grid operator that generates 2D sampling grid.
can_convert_multiply_to_intdiv(origin_size, ...)	Check whether can convert multiplication to division
crop_and_resize(data, boxes, box_indices, ...)	Perform crop and resize operation on the data.
dilation2d_nchw(input, filter, stride, ...)	Morphological dilation operator in NCHW layout.
dilation2d_nhwc(input, filter, stride, ...)	Morphological 2d dilation NHWC layout.
get_1d_indices(indices[, layout])	Get 1d indices
get_1d_pixel(data, layout, image_width, n, ...)	Get 1d pixel
get_2d_indices(indices[, layout])	Get 2d indices
get_2d_pixel(data, layout, image_height, ...)	Get 2d pixel
get_3d_indices(indices[, layout])	Get 3d indices
get_3d_pixel(data, layout, image_depth, ...)	Get 3d pixel
get_closest_index(in_x, rounding_method, boxes)	get the closest index to a value based on a certain rounding method
get_inx(x, image_width, target_width, ...[, ...])	Infer input x from output x with various coordinate transformation methods
get_pad_tuple(padding, kernel)	Common code to get the pad option
grid_sample(data, grid[, method, layout, ...])	Applies grid sampling to input feature map.
nchw_pack_layout(layout_info)	Check whether the layout type is NCHWinic
nchw_xc_layout(layout_info)	Check whether the layout type is NCHWxc
pad(data, pad_before[, pad_after, ...])	Pad Input with zeros.
resize1d(data, roi, size[, layout, method, ...])	Perform resize operation on the data.
resize2d(data, roi, size[, layout, method, ...])	Perform resize operation on the data.
resize3d(data, roi, size[, layout, method, ...])	Perform resize operation on the data.
simplify(expr)	Simplify the expression if it is Expr, directly return if it is int.

tvm.topi.image.affine_grid(data, target_shape)

affine_grid operator that generates 2D sampling grid.

This operation is described in https://arxiv.org/pdf/1506.02025.pdf. It generates a uniform sampling grid within the target shape and normalizes it to [-1, 1]. The provided affine transformation is then applied on the sampling grid.

参数:

• data (tvm.Tensor) -- 3-D with shape [batch, 2, 3]. The affine matrix.

• target_shape (list/tuple of two int) -- Specifies the output shape (H, W).

返回:

Output -- 4-D with shape [batch, 2, target_height, target_width]

返回类型:

tvm.Tensor

tvm.topi.image.can_convert_multiply_to_intdiv(origin_size, scaled_size)

Check whether can convert multiplication to division

tvm.topi.image.crop_and_resize(data, boxes, box_indices, crop_size, layout='NCHW', method='bilinear', extrapolation_value=None, out_dtype=None)

Perform crop and resize operation on the data.

参数:

• data (tvm.te.Tensor) -- inputs is a 4-D tensor with shape [batch, channel, in_height, in_width] or [batch, in_height, in_width, channel]

• boxes (tvm.te.Tensor) -- A 2-D tensor of shape [num_boxes, 4]. Each row of the tensor specifies the coordinates of a box.

• box_indices (tvm.te.Tensor) -- A 1-D tensor of shape [num_boxes], box_indices[i] specifies the data that the i-th box refers to.

• crop_size (Tuple) -- The target size of each box.

• layout (string, optional) -- "NCHW", "NHWC"

• method ({"bilinear", "nearest_neighbor"}) -- Method to be used for resizing.

• extrapolation_value (float, optional) -- Value used for extrapolation, when applicable.

• out_dtype (string, optional) -- Type to return. If left None will be same as input type.

返回:

output -- 4-D with shape [num_boxes, channel, crop_height, crop_width] or [num_boxes, crop_height, crop_width, channel]

返回类型:

tvm.te.Tensor

tvm.topi.image.dilation2d_nchw(input, filter, stride, padding, dilations, out_dtype=None)

Morphological dilation operator in NCHW layout.

参数:

• input (tvm.te.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width]

• filter (tvm.te.Tensor) -- 3-D with shape [ in_channel, filter_height, filter_width]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int or str) -- Padding size

• dilations (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (Optional[str]) -- Specifies the output data type.

返回:

Output -- 4-D with shape [batch, in_channel, out_height, out_width]

返回类型:

tvm.te.Tensor

tvm.topi.image.dilation2d_nhwc(input, filter, stride, padding, dilations, out_dtype=None)

Morphological 2d dilation NHWC layout.

参数:

• input (tvm.te.Tensor) -- 4-D with shape [batch, in_height, in_width, in_channel]

• filter (tvm.te.Tensor) -- 3-D with shape [filter_height, filter_width, in_channel]

• stride (int or a list/tuple of two ints) -- Stride size, or [stride_height, stride_width]

• padding (int) -- Padding size

• dilations (int or a list/tuple of two ints) -- dilation size, or [dilation_height, dilation_width]

• out_dtype (Optional[str]) -- Specifies the output data type.

返回:

Output -- 4-D with shape [batch, out_height, out_width, in_channel]

返回类型:

tvm.te.Tensor

tvm.topi.image.get_1d_indices(indices, layout='NCW')

Get 1d indices

tvm.topi.image.get_1d_pixel(data, layout, image_width, n, c, x, cc, ib, ic)

Get 1d pixel

tvm.topi.image.get_2d_indices(indices, layout='NCHW')

Get 2d indices

tvm.topi.image.get_2d_pixel(data, layout, image_height, image_width, n, c, y, x, cc, ib, ic)

Get 2d pixel

tvm.topi.image.get_3d_indices(indices, layout='NCDHW')

Get 3d indices

tvm.topi.image.get_3d_pixel(data, layout, image_depth, image_height, image_width, n, c, z, y, x, cc)

Get 3d pixel

tvm.topi.image.get_closest_index(in_x, rounding_method, boxes, use_int_div=False)

get the closest index to a value based on a certain rounding method

tvm.topi.image.get_inx(x, image_width, target_width, coordinate_transformation_mode, start_x=0, end_x=-1, use_int_div=False)

Infer input x from output x with various coordinate transformation methods

tvm.topi.image.get_pad_tuple(padding, kernel)

Common code to get the pad option

参数:

• padding (int or str) -- Padding size, or ['VALID', 'SAME']

• kernel (tuple of int) -- Conv kernel size

返回:

• pad_top (int) -- Padding size on top

• pad_left (int) -- Padding size on left

• pad_down (int) -- Padding size on down.

• pad_right (int) -- Padding size on right.

tvm.topi.image.grid_sample(data, grid, method='bilinear', layout='NCHW', padding_mode='zeros', align_corners=True)

Applies grid sampling to input feature map.

Given \(data\) and \(grid\), then for 4-D the output is computed by

\[x_{src} = grid[batch, 0, y_{dst}, x_{dst}] \ y_{src} = grid[batch, 1, y_{dst}, x_{dst}] \ output[batch, channel, y_{dst}, x_{dst}] = G(data[batch, channel, y_{src}, x_{src}])\]

\(x_{dst}\), \(y_{dst}\) enumerate all spatial locations in \(output\), and \(G()\) denotes the interpolation function.

The out-boundary points will be padded with zeros if padding_mode is "zeros", or border pixel value if padding_mode is "border", or inner pixel value if padding_mode is "reflection".

The left-top corner (-1, -1) and right-bottom corner (1, 1) in grid will be map to (0, 0) and (h - 1, w - 1) of data if align_corners is "True", or (-0.5, -0.5) and (h - 0.5, w - 0.5) of data if align_corners is "False".

The shape of the output will be 4-D (data.shape[0], data.shape[1], grid.shape[2], grid.shape[3]), or 5-D (data.shape[0], data.shape[1], grid.shape[2], grid.shape[3], grid.shape[4]).

The operator assumes that \(grid\) has been normalized to [-1, 1].

grid_sample often cooperates with affine_grid which generates sampling grids for grid_sample.

参数:

• data (tvm.Tensor) -- 4-D with shape [batch, in_channel, in_height, in_width], or 5-D with shape [batch, in_channel, in_depth, in_height, in_width]

• grid (tvm.Tensor) -- 4-D with shape [batch, 2, out_height, out_width], or 5-D with shape [batch, 3, out_depth, out_height, out_width]

• method (str) -- The interpolation method, 4-D "nearest", "bilinear", "bicubic" and 5-D "nearest", "bilinear"("trilinear") are supported.

• layout (str) -- The layout of input data and the output.

• padding_mode (str) -- The padding mode for outside grid values, "zeros", "border", "reflection" are supported.

• align_corners (bool) -- Geometrically, we consider the pixels of the input as squares rather than points. If set to "True", the extrema ("-1" and "1") are considered as referring to the center points of the input corner pixels. If set to "False", they are instead considered as referring to the corner points of the input corner pixels, making the sampling more resolution agnostic.

返回:

Output -- 4-D with shape [batch, in_channel, out_height, out_width], or 5-D with shape [batch, in_channel, out_depth, out_height, out_width]

返回类型:

tvm.Tensor

tvm.topi.image.nchw_pack_layout(layout_info)

Check whether the layout type is NCHWinic

tvm.topi.image.nchw_xc_layout(layout_info)

Check whether the layout type is NCHWxc

tvm.topi.image.pad(data, pad_before, pad_after=None, pad_value=0.0, name='PadInput', attrs=None)

Pad Input with zeros.

参数:

• data (tvm.te.Tensor) -- n-D input, can be any layout.

• pad_before (list / tuple of n ints) -- Pad width on each dimension to pad the before the axis begin.

• pad_after (list / tuple of n ints, optional) -- Pad width each dimension to pad the after the axis end.

• pad_value (float, optional) -- The value to be padded.

• name (str, optional) -- The name prefix operators generated

返回:

Output -- n-D, the same layout as Input.

返回类型:

tvm.te.Tensor

tvm.topi.image.resize1d(data, roi, size, layout='NCW', method='linear', coordinate_transformation_mode='half_pixel', rounding_method='', bicubic_alpha=-0.5, bicubic_exclude=0, extrapolation_value=0.0, out_dtype=None, output_shape=None)

Perform resize operation on the data.

参数:

• data (tvm.te.Tensor) -- inputs is a 3-D tensor with shape [batch, channel in_width] or [batch in_width, channel]

• roi (Tuple of Float or Expr) -- The region of interest for cropping the input image. Expected to be of size 2, and format [start_w, end_w]. Only used if coordinate_transformation_mode is tf_crop_and_resize.

• size (Tuple) -- Output resolution scale to

• layout (string, optional) -- "NCW", "NWC", or "NCWc".

• coordinate_transformation_mode (string, optional) -- Describes how to transform the coordinate in the resized tensor to the coordinate in the original tensor. Refer to the ONNX Resize operator specification for details. Available options are "half_pixel", "align_corners" and "asymmetric".

• method (string, optional) -- method of interpolation ("nearest", "linear", "bicubic")

• coordinate_transformation_mode -- Describes how to transform the coordinate in the resized tensor to the coordinate in the original tensor. [half_pixel, align_corners, asymmetric, pytorch_half_pixel, tf_half_pixel_for_nn, and tf_crop_and_resize].

• rounding_method -- Method for rounding coordinate locations

• bicubic_alpha (float, optional) -- Bicubic spline coefficient

• bicubic_exclude (bool, optional:) -- Exclude values outside the image fdor bicubic interpolation

• extrapolation_value (float, optional) -- Value used for extrapolation, when applicable.

• out_dtype (string, optional) -- Type to return. If left None will be same as input type.

• output_shape (tvm.tir.container.Array, optional) -- Shape to return. If left None will be inferred (If shape is determined dynamically, pass out_dtype.shape as output_shape)

返回:

output -- 4-D with shape [batch, chananel, in_width*scale] or [batch, in_width*scale, channel] or 5-D with shape [batch, channel-major, in_width*scale, channel-minor]

返回类型:

tvm.te.Tensor

tvm.topi.image.resize2d(data, roi, size, layout='NCHW', method='linear', coordinate_transformation_mode='half_pixel', rounding_method='', bicubic_alpha=-0.5, bicubic_exclude=0, extrapolation_value=0.0, out_dtype=None, output_shape=None)

Perform resize operation on the data.

参数:

• data (tvm.te.Tensor) -- inputs is a 4-D tensor with shape [batch, channel, in_height, in_width] or [batch, in_height, in_width, channel]

• roi (Tuple of Float or Expr) -- The region of interest for cropping the input image. Expected to be of size 4, and format [start_h, start_w, end_h, end_w]. Only used if coordinate_transformation_mode is tf_crop_and_resize.

• size (Tuple) -- Output resolution scale to

• layout (string, optional) -- "NCHW", "NHWC", or "NCHWc".

• method (string, optional) -- method of interpolation ("nearest", "linear", "bicubic")

• coordinate_transformation_mode (string, optional) -- Describes how to transform the coordinate in the resized tensor to the coordinate in the original tensor. [half_pixel, align_corners, asymmetric, pytorch_half_pixel, tf_half_pixel_for_nn, and tf_crop_and_resize].

• rounding_method -- Method for rounding coordinate locations

• bicubic_alpha (float, optional) -- Bicubic spline coefficient

• bicubic_exclude (bool, optional:) -- Exclude values outside the image fdor bicubic interpolation

• extrapolation_value (float, optional) -- Value used for extrapolation, when applicable.

• out_dtype (string, optional) -- Type to return. If left None will be same as input type.

• output_shape (tvm.tir.container.Array, optional) -- Shape to return. If left None will be inferred (If shape is determined dynamically, pass out_dtype.shape as output_shape)

返回:

output -- 4-D with shape [batch, channel, in_height*scale, in_width*scale] or [batch, in_height*scale, in_width*scale, channel] or 5-D with shape [batch, channel-major, in_height*scale, in_width*scale, channel-minor]

返回类型:

tvm.te.Tensor

tvm.topi.image.resize3d(data, roi, size, layout='NCDHW', method='linear', coordinate_transformation_mode='half_pixel', rounding_method='', bicubic_alpha=-0.5, bicubic_exclude=0, extrapolation_value=0.0, out_dtype=None, output_shape=None)

Perform resize operation on the data.

参数:

• data (tvm.te.Tensor) -- inputs is a 5-D tensor with shape [batch, channel, in_depth, in_height, in_width] or [batch, in_depth, in_height, in_width, channel]

• roi (Tuple of Float or Expr) -- The region of interest for cropping the input image. Expected to be of size 6, and format [start_d, start_h, start_w, end_d, end_h, end_w]. Only used if coordinate_transformation_mode is tf_crop_and_resize.

• size (Tuple) -- Output resolution scale to

• layout (string, optional) -- "NCDHW", "NDHWC", or "NCDHWc".

• method (string, optional) -- method of interpolation ("nearest", "linear", "bicubic")

• coordinate_transformation_mode (string, optional) -- Describes how to transform the coordinate in the resized tensor to the coordinate in the original tensor. [half_pixel, align_corners, asymmetric, pytorch_half_pixel, tf_half_pixel_for_nn, and tf_crop_and_resize].

• rounding_method -- Method for rounding coordinate locations

• bicubic_alpha (float, optional) -- Bicubic spline coefficient

• bicubic_exclude (bool, optional:) -- Exclude values outside the image fdor bicubic interpolation

• extrapolation_value (float, optional) -- Value used for extrapolation, when applicable.

• out_dtype (string, optional) -- Type to return. If left None will be same as input type.

• output_shape (tvm.tir.container.Array, optional) -- Shape to return. If left None will be inferred (If shape is determined dynamically, pass out_dtype.shape as output_shape)

返回:

output -- 4-D with shape [batch, channel, in_depth*scale, in_height*scale, in_width*scale] or [batch, in_depth*scale, in_height*scale, in_width*scale, channel] or 5-D with shape [batch, channel-major, in_depth*scale, in_height*scale, in_width*scale, channel-minor]

返回类型:

tvm.te.Tensor

tvm.topi.image.simplify(expr)

Simplify the expression if it is Expr, directly return if it is int.

参数:

expr (Expr or int) -- The input.

返回:

out -- The simplified output

返回类型:

Expr or int

tvm.topi.sparse

Sparse operators

Functions:

csrmm(a, b[, c])	The csrmm routine performs a matrix-matrix operation defined as \(C := A*B + C\), where B and C are dense matrices, A is an m-by-k sparse matrix in the CSR format.
csrmv(a, x[, y])	The csrmv routine performs a matrix-vector operation defined as \(y := A*x + y\), where x and y are vectors, A is an m-by-k sparse matrix in the CSR format.
dense(data, weight[, bias])	Applies a linear transformation: \(Y = XW^T + b\).

tvm.topi.sparse.csrmm(a, b, c=None)

The csrmm routine performs a matrix-matrix operation defined as \(C := A*B + C\), where B and C are dense matrices, A is an m-by-k sparse matrix in the CSR format.

参数:

• a (tvm.contrib.sparse.CSRNDArray) -- 2-D sparse matrix with shape [m, k]

• b (tvm.te.Tensor) -- 2-D dense matrix with shape [k, n]

• c (tvm.te.Tensor, optional) -- 1-D dense vector with shape [n]

返回:

output -- 2-D with shape [m, n]

返回类型:

tvm.te.Tensor

tvm.topi.sparse.csrmv(a, x, y=None)

The csrmv routine performs a matrix-vector operation defined as \(y := A*x + y\), where x and y are vectors, A is an m-by-k sparse matrix in the CSR format.

参数:

• a (tvm.contrib.sparse.CSRNDArray) -- 2-D sparse matrix with shape [m, k]

• x (tvm.te.Tensor) -- 2-D dense matrix with shape [k, 1]

• y (tvm.te.Tensor, optional) -- 1-D dense vector with shape [1]

返回:

output -- 2-D dense matrix with shape [m, 1]

返回类型:

tvm.te.Tensor

tvm.topi.sparse.dense(data, weight, bias=None)

Applies a linear transformation: \(Y = XW^T + b\). Either data or weight should be tvm.contrib.sparse.CSRNDArray.

参数:

• data (tvm.contrib.sparse.CSRNDArray or te.tensor.Tensor) -- 2-D with shape [batch, in_dim]

• weight (te.tensor.Tensor or tvm.contrib.sparse.CSRNDArray) -- 2-D with shape [out_dim, in_dim]

• bias (te.tensor.Tensor, optional) -- 1-D with shape [out_dim]

返回:

output -- 2-D with shape [batch, out_dim]

返回类型:

tvm.te.Tensor