eddl_progress_tensor.md

Tensor Routines

Numpy-like operations over a raw-tensor object

Legend: Development status

Image	Meaning
🟢️	Done
🔴️	Todo
⚫️	Not planned

---

Array creation routines

Ones and zeros

---

Functionality	CPU	GPU	Comments
empty	🟢️	🟢️	Returns a tensor filled with uninitialized data
empty_like	🟢️	🟢 ️	Returns a tensor filled with uninitialized data, with the same size as the input tensor
eye	🟢️	🟢️	Return a 2-D array with ones on the diagonal and zeros elsewhere
identity	🟢️	🟢️	Return the identity array (eye with offset=0)
ones	🟢️	🟢️	Return a new array of given shape and type, filled with ones
ones_like	🟢️	🟢	Returns a tensor filled with the scalar value 1, with the same size as the input tensor
zeros	🟢️	🟢️	Return a new array of given shape and type, filled with zeros
zeros_like	🟢️	🟢	Returns a tensor filled with the scalar value 0, with the same size as the input tensor
full	🟢️	🟢️	Return a new array of given shape and type, filled with "value"
full_like	🟢️	🟢	Returns a tensor filled with the given scalar value, with the same size as the input tensor

From existing data

Functionality	CPU	GPU	Comments
Tensor	🟢️	🟢	Constructs a tensor with data
clone	🟢️	🟢️	Creates an identical (but different) tensor from another
copy	🟢️	🟢️	Copy data from Tensor A to B

Numerical ranges

Functionality	CPU	GPU	Comments
arange	🟢️	🟢️	Return evenly spaced values within a given interval [0, n)
range	🟢️	🟢️	Return evenly spaced values within a given interval [0, n]
linspace	🟢️	🟢️	Return evenly spaced numbers over a specified interval
logspace	🟢️	🟢️	Return numbers spaced evenly on a log scale
geomspace	🟢️	🟢️	Return numbers spaced evenly on a log scale (a geometric progression)

Random

Functionality	CPU	GPU	Comments
randu	🟢️	🟢️	Return a uniform random matrix with given shape
randn	🟢️	🟢️	Return a normal random matrix with data from the "standard normal" distribution

Building matrices

Functionality	CPU	GPU	Comments
diag	🟢️	🟢️	Extract a diagonal or construct a diagonal array
tri	⚫️	⚫️️	An array with ones at and below the given diagonal and zeros elsewhere

Array manipulation routines

Changing array shape

Functionality	CPU	GPU	Comments
reshape	🟢️	🟢️	Gives a new shape to an array without changing its data
flatten	🟢️	🟢️	Return a copy of the array collapsed into one dimension

Transpose-like operations

Functionality	CPU	GPU	Comments
moveaxis	🟢️	🟢️	Move axes of an array to new positions (1, 3): [0,1,2,3] => [0,2,3,1]
swapaxes	🟢️	🟢️	Interchange two axes of an array (1, 3): [0,1,2,3] => [0,3,2,1]
permute	🟢️	🟢️	Permute the dimensions of an array

Changing number of dimensions

Functionality	CPU	GPU	Comments
squeeze	🟢️	🟢️	Remove single-dimensional entries from the shape of an array [3, 4, 1, 7] => [3,4,7]
unsqueeze	🟢️	🟢️	Expand the shape of an array [3, 4, 7] => [1, 3, 4, 7]

Value operations

Functionality	CPU	GPU	Comments
fill	🟢️	🟢️	Fills a tensor in-place, with a constant value
fill_rand_uniform	🟢️	🟢️	Fills a tensor in-place, with values randomly sampled from a uniform distribution
fill_rand_signed_uniform	🟢️	🟢️	Fills a tensor in-place, with values randomly sampled from a signed uniform distribution
fill_rand_normal	🟢️	🟢️	Fills a tensor in-place, with values randomly sampled from a normal distribution
fill_rand_binary	🟢️	🟢️	Fills a tensor in-place, with values randomly sampled from a binary distribution

Joining arrays

Functionality	CPU	GPU	Comments
concatenate	🟢️	🟢️	Join a sequence of arrays along an existing axis
stack	🟢️	🟢	Join a sequence of arrays along a new axis

Splitting arrays

Functionality	CPU	GPU	Comments
split	⚫️	⚫️️	Split an array into multiple sub-arrays

Tiling arrays

Functionality	CPU	GPU	Comments
repeat	🟢️	🟢️️️ ️	Repeats the elements of a tensor along the specified dimension
tile	🟢️	🟢️️️ ️	Repeats the elements of a tensor along the specified dimensions
broadcast	🟢️	🟢️️️ ️	Produce an object that mimics broadcasting.

Adding and removing elements

Functionality	CPU	GPU	Comments
delete	⚫️	⚫️️	Return a new array with sub-arrays along an axis deleted
insert	⚫️	⚫️️	Insert values along the given axis before the given indices
append	⚫️	⚫️️	Append values to the end of an array
trim_zeros	⚫️	⚫️️	Trim the leading and/or trailing zeros from a 1-D array or sequence
unique	⚫️	⚫️️	Find the unique elements of an array

Rearranging elements and transformations

Functionality	CPU	GPU	Comments
shift	🟢️	🟢️
rotate	🟢️	🟢️
scale	🟢️	🟢️
flip	🟢️	🟢️
crop	🟢️	🟢️
crop_scale	🟢️	🟢️
cutout	🟢️	🟢️
pad	🟢️	🟢️
shift_random	🟢️	🟢️
rotate_random	🟢️	🟢️
scale_random	🟢️	🟢️
flip_random	🟢️	🟢️
crop_random	🟢️	🟢️
crop_scale_random	🟢️	🟢️
cutout_random	🟢️	🟢️

Indexing routines

Generating index arrays

Functionality	CPU	GPU	Comments
nonzero	🟢️	🔴️	Return the indices of the elements that are non-zero
where	🟢️	🟢️	Return elements, either from x or y, depending on condition
mask_indices	⚫️	⚫️️	Return the indices to access (n, n) arrays, given a masking function

Indexing-like operations

Functionality	CPU	GPU	Comments
select	🟢️	🟢️	Returns an array with the selected indices Tensor::select(k); k=vector of strings ({"0", ":5", ":", "3:6"})
set_select	🟢️	🟢️	Sets the elements in the array using the selected indices `Tensor::set_select({"0", ":5", ":", "3:6"}, k); //k=float or Tensor
permute	🟢️	🟢️	Permutation of tensor dimensions
slice	🟢️	🟢️	Alias for select
expand	🟢️	🟢️	Returns a new tensor with singleton dimensions expanded to a larger size
index_select	⚫️	⚫️️ ️	Returns a new tensor which indexes the input tensor along dimension dim using the entries in index
masked_select	⚫️	⚫️️ ️	Returns a new 1-D tensor which indexes the input tensor according to the boolean mask
take	⚫️	⚫️️ ️	Returns a new tensor with the elements of input at the given indices. The input tensor is treated as if it were viewed as a 1-D tensor

Input and output

Input

Functionality	CPU	GPU	Comments
load	🟢️	-	Images: jpg, png, bmp, hdr, psd, tga, gif, pic, pgm, ppm Other: bin

Output

Functionality	CPU	GPU	Comments
save	🟢️	-	Images: jpg, png, bmp, hdr, psd, tga, gif, pic, pgm, ppm Text: csv, tsv, txt,... Other: bin
save2txt	🟢️	-

Linear algebra

Matrix and vector products

Functionality	CPU	GPU	Comments
dot	⚫️	⚫️️	Dot product of two arrays
inner	⚫️	⚫️️	Inner product of two arrays
outer	⚫️	⚫️️	Compute the outer product of two vectors
matmul	⚫️	⚫️️	Matrix product of two arrays
tensordot	⚫️	⚫️	Compute tensor dot product along specified axes for arrays >= 1-D
interpolate	🟢️	🟢️	Interpolate two tensors: c*A + (1-c)*B

Norms and other numbers

Functionality	CPU	GPU	Comments
norm	🟢️	🟢️	Matrix or vector norm
trace	🟢️	🟢️	Return the sum along diagonals of the array

Logic functions

Truth value testing

Functionality	CPU	GPU	Comments
all	🟢️	🟢️	Test whether all array elements along a given axis evaluate to True
any	🟢️	🟢️	Test whether any array element along a given axis evaluates to True

Array contents

Functionality	CPU	GPU	Comments
isfinite	🟢️	🟢️	Test element-wise for finiteness (not infinity or not Not a Number)
isinf	🟢️	🟢️	Test element-wise for positive or negative infinity
isnan	🟢️	🟢️	Test element-wise for NaN and return result as a boolean array
isneginf	🟢️	🟢️	Test element-wise for negative infinity, return result as bool array
isposinf	🟢️	🟢️	Test element-wise for positive infinity, return result as bool array

Logical operations

Functionality	CPU	GPU	Comments
logical_and	🟢️	🟢️	Compute the truth value of x1 AND x2 element-wise
logical_or	🟢️	🟢️	Compute the truth value of x1 OR x2 element-wise
logical_not	🟢️	🟢️	Compute the truth value of NOT x element-wise
logical_xor	🟢️	🟢️	Compute the truth value of x1 XOR x2, element-wise

Comparison

Boolean

Functionality	CPU	GPU	Comments
allclose	🟢️	🟢️	Returns True if two arrays are element-wise equal within a tolerance
isclose	🟢️	🟢️	Returns a boolean array where two arrays are element-wise equal within a tolerance
greater	🟢️	🟢️	Return the truth value of (x1 > x2); Tensor-Tensor, Tensor-float
greater_equal	🟢️	🟢️	Return the truth value of (x1 >= x2); Tensor-Tensor, Tensor-float
less	🟢️	🟢️	Return the truth value of (x1 < x2) element-wise; Tensor-Tensor, Tensor-float
less_equal	🟢️	🟢️	Return the truth value of (x1 =< x2) element-wise; Tensor-Tensor, Tensor-float
equal	🟢️	🟢️	Return (x1 == x2) element-wise; Tensor-Tensor, Tensor-float
not_equal	🟢️	🟢️	Return (x1 != x2) element-wise; Tensor-Tensor, Tensor-float

Indices

Functionality	CPU	GPU	Comments
argsort	🟢️	🟢️ ️	Returns the indices that sort a tensor along a given dimension in ascending order by value
kthvalue	⚫️️	⚫️️ ️	Returns a namedtuple (values, indices) where values is the k th smallest element of each row of the input tensor in the given dimension dim
sort	🟢️	🟢️️	Sorts the elements of the input tensor along a given dimension in ascending order by value
topk	⚫️️	⚫️️ ️	Returns the k largest elements of the given input tensor along a given dimension

Mathematical functions

Point-wise

Functionality	CPU	GPU	Comments
abs	🟢️	🟢️
acos	🟢️	🟢️
add	🟢️	🟢️
asin	🟢️	🟢️
atan	🟢️	🟢️
ceil	🟢️	🟢️
clamp	🟢️	🟢️
clampmax	🟢️	🟢️
clampmin	🟢️	🟢️
cos	🟢️	🟢️
cosh	🟢️	🟢️
div	🟢️	🟢️
exp	🟢️	🟢️
floor	🟢️	🟢️
log	🟢️	🟢️
log2	🟢️	🟢️
log10	🟢️	🟢️
logn	🟢️	🟢️
mod	🟢️	🟢️
mult	🟢️	🟢️
neg	🟢️	🟢️
normalize	🟢️	🟢️
pow	🟢️	🟢️
powb	🟢️	🟢️
reciprocal	🟢️	🟢️
remainder	🟢️	🟢️
round	🟢️	🟢️
rsqrt	🟢️	🟢️
sigmoid	🟢️	🟢️
sign	🟢️	🟢️
sin	🟢️	🟢️
sinh	🟢️	🟢️
sqr	🟢️	🟢️
sqrt	🟢️	🟢️
sub	🟢️	🟢️
tan	🟢️	🟢️
tanh	🟢️	🟢️
trunc	🟢️	🟢️

Element-wise

Functionality	CPU	GPU	Comments
add	🟢️	🟢️	Tensor-Tensor, Tensor-float
div	🟢️	🟢️	Tensor-Tensor, Tensor-float
mult	🟢️	🟢️	Tensor-Tensor, Tensor-float
sub	🟢️	🟢️	Tensor-Tensor, Tensor-float
maximum	🟢️	🟢️	Tensor-Tensor, Tensor-float
minimum	🟢️	🟢️	Tensor-Tensor, Tensor-float

Single-value

Functionality	CPU	GPU	Comments
argmax	🟢️	🟢️
argmin	🟢️	🟢️
max	🟢️	🟢️
min	🟢️	🟢️
mean	🟢️	🟢
median	🟢️	🟢
mode	🟢️	🟢️
norm	🟢	🟢️
prod	🟢️	🟢
std	🟢	🟢
sum	🟢️	🟢
sum_abs	🟢️	🟢
var	🟢️	🟢

Reductions

Functionality	CPU	GPU	Comments
argmax	🟢️	🟢️
argmin	🟢️	🟢️
max	🟢️	🟢️
min	🟢️	🟢️
mean	🟢️	🟢
median	🟢️	🟢
mode	🟢️	🟢️
norm	🟢	🟢
prod	🟢️	🟢
std	🟢	🟢
sum	🟢️	🟢
sum_abs	🟢️	🟢
var	🟢️	🟢

Miscellaneous

Functionality	CPU	GPU	Comments
toCPU	🟢️	🟢️	Clone a tensor to the CPU
toGPU	🟢️	🟢️	Clone a tensor to the GPU
isCPU	🟢️	🟢️	Check if the tensor if in CPU
isGPU	🟢️	🟢️	Check if the tensor if in GPU
isFPGA	-	-	Check if the tensor if in FPGA
isSquared	🟢️	🟢️	Check if all dimensions in the tensors are the same
copy	🟢️	🟢️	Copy data from Tensor A to B
clone	🟢️	🟢️	Clone a tensor (same device)
info	🟢️	🟢️	Print shape, device and size information
print	🟢️	🟢️	Prints the tensor values
numel	🟢️	🟢️	Returns the total number of elements in the input tensor