| Title: | R Bindings to 'PyTorch' |
|---|---|
| Description: | 'R' implementation and interface of the Machine Learning platform 'PyTorch' <https://pytorch.org/> developed in 'Python'. It requires a 'conda' environment with 'torch' and 'torchvision' Python packages to provide 'PyTorch' functions, methods and classes. The key object in 'PyTorch' is the tensor which is in essence a multidimensional array. These tensors are fairly flexible in performing calculations in CPUs as well as 'GPUs' to accelerate tensor operations. |
| Authors: | Alfonso R. Reyes [aut, cre, cph] |
| Maintainer: | Alfonso R. Reyes <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.4.3-9002 |
| Built: | 2025-03-03 03:48:17 UTC |
| Source: | https://github.com/f0nzie/rtorch |
This generic is similar to applying torch$sub(a, b)
## S3 method for class 'torch.Tensor' a - b## S3 method for class 'torch.Tensor' a - b
a |
tensor |
b |
tensor |
Another tensor representing the subtraction of two tensors.
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a - b ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a - b ## End(Not run)
[
Subset tensors with [
## S3 method for class 'torch.Tensor' x[ ..., drop = TRUE, style = getOption("torch.extract.style"), options = torch_extract_opts(style) ]## S3 method for class 'torch.Tensor' x[ ..., drop = TRUE, style = getOption("torch.extract.style"), options = torch_extract_opts(style) ]
x |
a tensor |
... |
slicing specs. See examples and details. |
drop |
whether to drop scalar dimensions |
style |
One of |
options |
An object returned by |
## Not run: x <- torch$arange(0L, 15L)$view(3L, 5L) # by default, numerics supplied to `...` are interpreted R style x[,1] # first column x[1:2,] # first two rows x[,1, drop = FALSE] # strided steps can be specified in R syntax or python syntax x[, seq(1, 5, by = 2)] x[, 1:5:2] # if you are unfamiliar with python-style strided steps, see: # https://docs.scipy.org/doc/numpy-1.13.0/reference/arrays.indexing.html#basic-slicing-and-indexing # missing arguments for python syntax are valid, but they must by backticked # or supplied as NULL x[, `::2`] x[, NULL:NULL:2] x[, `2:`] # Another Python feature that is available is a Python style ellipsis `...` # (not to be confused with R dots `...`), that in R has been defined as # all_dims() expands to the shape of the tensor y <- torch$arange(0L, 3L^5L)$view(3L, 3L, 3L, 3L, 3L) as.logical((all(y[all_dims(), 1] == y[,,,,1]))$numpy()) == TRUE # negative numbers are always interpreted Python style # The first time a negative number is supplied to `[`, a warning is issued # about the non-standard behavior. x[-1,] # last row, with a warning x[-1,] # the warning is only issued once # specifying `style = 'python'` changes the following: # + zero-based indexing is used # + slice sequences in the form of `start:stop` do not include `stop` # in the returned value # + out-of-bounds indices in a slice are valid # The style argument can be supplied to individual calls of `[` or set # as a global option # example of zero based indexing x[0, , style = 'python'] # first row x[1, , style = 'python'] # second row # example of slices with exclusive stop # run the next options() line before the tensor operations options(torch.extract.style = 'python') x[, 0:1] # just the first column x[, 0:2] # first and second column # example of out-of-bounds index x[, 0:10] options(torch.extract.style = NULL) # slicing with tensors is valid too, but note that tensors are never # translated and are always interpreted Python-style. # A warning is issued the first time a tensor is passed to `[` # just as in Python, only scalar tensors are valid # To silence the warnings about tensors being passed as-is and negative numbers # being interpreted python-style, set options(torch.extract.style = 'R') # clean up from examples options(torch.extract.style = NULL) ## End(Not run)## Not run: x <- torch$arange(0L, 15L)$view(3L, 5L) # by default, numerics supplied to `...` are interpreted R style x[,1] # first column x[1:2,] # first two rows x[,1, drop = FALSE] # strided steps can be specified in R syntax or python syntax x[, seq(1, 5, by = 2)] x[, 1:5:2] # if you are unfamiliar with python-style strided steps, see: # https://docs.scipy.org/doc/numpy-1.13.0/reference/arrays.indexing.html#basic-slicing-and-indexing # missing arguments for python syntax are valid, but they must by backticked # or supplied as NULL x[, `::2`] x[, NULL:NULL:2] x[, `2:`] # Another Python feature that is available is a Python style ellipsis `...` # (not to be confused with R dots `...`), that in R has been defined as # all_dims() expands to the shape of the tensor y <- torch$arange(0L, 3L^5L)$view(3L, 3L, 3L, 3L, 3L) as.logical((all(y[all_dims(), 1] == y[,,,,1]))$numpy()) == TRUE # negative numbers are always interpreted Python style # The first time a negative number is supplied to `[`, a warning is issued # about the non-standard behavior. x[-1,] # last row, with a warning x[-1,] # the warning is only issued once # specifying `style = 'python'` changes the following: # + zero-based indexing is used # + slice sequences in the form of `start:stop` do not include `stop` # in the returned value # + out-of-bounds indices in a slice are valid # The style argument can be supplied to individual calls of `[` or set # as a global option # example of zero based indexing x[0, , style = 'python'] # first row x[1, , style = 'python'] # second row # example of slices with exclusive stop # run the next options() line before the tensor operations options(torch.extract.style = 'python') x[, 0:1] # just the first column x[, 0:2] # first and second column # example of out-of-bounds index x[, 0:10] options(torch.extract.style = NULL) # slicing with tensors is valid too, but note that tensors are never # translated and are always interpreted Python-style. # A warning is issued the first time a tensor is passed to `[` # just as in Python, only scalar tensors are valid # To silence the warnings about tensors being passed as-is and negative numbers # being interpreted python-style, set options(torch.extract.style = 'R') # clean up from examples options(torch.extract.style = NULL) ## End(Not run)
This generic is similar to torch$mul(a, b)
## S3 method for class 'torch.Tensor' a * b## S3 method for class 'torch.Tensor' a * b
a |
tensor |
b |
tensor |
Another tensor representing the multiplication of two tensors.
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a * b ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a * b ## End(Not run)
This generic is similar to torch$div(a, b)
## S3 method for class 'torch.Tensor' a / b## S3 method for class 'torch.Tensor' a / b
a |
tensor |
b |
tensor |
Another tensor representing the division of two tensors.
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a / b ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a / b ## End(Not run)
This generic is similar to torch$dot(a, b)
a %.*% ba %.*% b
a |
tensor |
b |
tensor |
a scalar
## Not run: p <- torch$Tensor(list(2, 3)) q <- torch$Tensor(list(2, 1)) p %.*% q ## End(Not run)## Not run: p <- torch$Tensor(list(2, 3)) q <- torch$Tensor(list(2, 1)) p %.*% q ## End(Not run)
This generic is similar to torch$matmul(a, b)
a %**% ba %**% b
a |
tensor |
b |
tensor |
a scalar or a tensor
## Not run: p <- torch$randn(3L) q <- torch$randn(3L) p %**% q ## End(Not run)## Not run: p <- torch$randn(3L) q <- torch$randn(3L) p %**% q ## End(Not run)
Computes the element-wise remainder of division.
## S3 method for class 'torch.Tensor' a %% b## S3 method for class 'torch.Tensor' a %% b
a |
a tensor |
b |
a scalar or a tensor |
the reminder of the division between tensor by a scalar or tensor
## Not run: x <- torch$Tensor(list(-3., -2, -1, 1, 2, 3)) y <- torch$Tensor(list(1., 2, 3, 4, 5)) torch$remainder(x, 2) torch$remainder(y, 1.5) x %% 2 y %% 1.5 ## End(Not run)## Not run: x <- torch$Tensor(list(-3., -2, -1, 1, 2, 3)) y <- torch$Tensor(list(1., 2, 3, 4, 5)) torch$remainder(x, 2) torch$remainder(y, 1.5) x %% 2 y %% 1.5 ## End(Not run)
This generic is similar to applying torch$add(a, b)
## S3 method for class 'torch.Tensor' a + b## S3 method for class 'torch.Tensor' a + b
a |
tensor |
b |
tensor |
Another tensor representing the addition of two tensors.
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a + b ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a + b ## End(Not run)
This generic is similar to torch$lt(a, b)
## S3 method for class 'torch.Tensor' a < b## S3 method for class 'torch.Tensor' a < b
a |
tensor |
b |
tensor |
A tensor of booleans representing the logical result of the comparison.
False to represent 0, and True to represent 1 in a tensor of data type torch$uint8.
## Not run: A <- torch$ones(28L, 28L) C <- A * 0.5 A < C ## End(Not run)## Not run: A <- torch$ones(28L, 28L) C <- A * 0.5 A < C ## End(Not run)
This generic is similar to torch$le(a, b)
## S3 method for class 'torch.Tensor' a <= b## S3 method for class 'torch.Tensor' a <= b
a |
tensor |
b |
tensor |
A tensor of booleans representing the logical result of the comparison.
False to represent 0, and True to represent 1 in a tensor of data type torch$uint8.
## Not run: A <- torch$ones(5L, 5L) C <- torch$as_tensor(np$random$randint(2L, size=c(5L, 5L)), dtype=torch$float32) A <= C C <= A ## End(Not run)## Not run: A <- torch$ones(5L, 5L) C <- torch$as_tensor(np$random$randint(2L, size=c(5L, 5L)), dtype=torch$float32) A <= C C <= A ## End(Not run)
This generic is approximately similar to torch$eq(a, b), with the
difference that the generic returns a tensor of booleans instead of
a tensor of data type torch$uint8.
## S3 method for class 'torch.Tensor' x == y## S3 method for class 'torch.Tensor' x == y
x |
tensor |
y |
tensor |
A tensor of booleans, where False corresponds to 0, and 1 to True
in a tensor of data type torch$bool.
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) a == b ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) a == b ## End(Not run)
This generic is similar to torch$gt(a, b)
## S3 method for class 'torch.Tensor' a > b## S3 method for class 'torch.Tensor' a > b
a |
tensor |
b |
tensor |
A tensor of booleans representing the logical result of the comparison.
False to represent 0, and True to represent 1 in a tensor of data type torch$uint8.
## Not run: A <- torch$ones(5L, 5L) C <- torch$as_tensor(np$random$randint(2L, size=c(5L, 5L)), dtype=torch$float32) A > C C > A ## End(Not run)## Not run: A <- torch$ones(5L, 5L) C <- torch$as_tensor(np$random$randint(2L, size=c(5L, 5L)), dtype=torch$float32) A > C C > A ## End(Not run)
This generic is similar to torch$ge(a, b)
## S3 method for class 'torch.Tensor' a >= b## S3 method for class 'torch.Tensor' a >= b
a |
tensor |
b |
tensor |
A tensor of booleans representing the logical result of the comparison.
False to represent 0, and True to represent 1 in a tensor of data type torch$uint8.
## Not run: A <- torch$ones(5L, 5L) C <- torch$as_tensor(np$random$randint(2L, size=c(5L, 5L)), dtype=torch$float32) A >= C C >= A ## End(Not run)## Not run: A <- torch$ones(5L, 5L) C <- torch$as_tensor(np$random$randint(2L, size=c(5L, 5L)), dtype=torch$float32) A >= C C >= A ## End(Not run)
This function returns an object that can be used when subsetting tensors with
[. If you are familiar with Python, this is equivalent to the Python Ellipsis
..., (not to be confused with ... in R). In Python, if x is a numpy
array or a torch tensor, in x[..., i] the ellipsis means "expand to match number of
dimension of x".
To translate the above Python expression to R, write:
x[all_dims(), i].
all_dims()all_dims()
## Not run: # Run this d <- torch$tensor(list(list(0, 0), list(0, 0), list(0, 1), list(1, 1)), dtype=torch$uint8) d[all_dims(), 1] f <- torch$arange(9L)$reshape(c(3L, 3L)) f f[all_dims()] f[all_dims(), 1L] ## End(Not run)## Not run: # Run this d <- torch$tensor(list(list(0, 0), list(0, 0), list(0, 1), list(1, 1)), dtype=torch$uint8) d[all_dims(), 1] f <- torch$arange(9L)$reshape(c(3L, 3L)) f f[all_dims()] f[all_dims(), 1L] ## End(Not run)
Returns True if all elements in the tensor are non-zero, False otherwise.
## S3 method for class 'torch.Tensor' all(x, dim, ...)## S3 method for class 'torch.Tensor' all(x, dim, ...)
x |
tensor |
dim |
dimension to reduce |
... |
other parameters (yet to be developed) |
A tensor of type torch.uint8 representing the boolean result: 1 for TRUE and 0 for FALSE.
## Not run: a <- torch$BoolTensor(list(TRUE, TRUE, TRUE, TRUE)) b <- torch$BoolTensor(list(FALSE, TRUE, TRUE, TRUE)) c <- torch$BoolTensor(list(TRUE, TRUE, TRUE, FALSE)) all(a) all(b) all(c) d <- torch$tensor(list(list(0, 0), list(0, 0), list(0, 1), list(1, 1)), dtype=torch$uint8) all(d) all(d, dim=0L) all(d, dim=1L) ## End(Not run)## Not run: a <- torch$BoolTensor(list(TRUE, TRUE, TRUE, TRUE)) b <- torch$BoolTensor(list(FALSE, TRUE, TRUE, TRUE)) c <- torch$BoolTensor(list(TRUE, TRUE, TRUE, FALSE)) all(a) all(b) all(c) d <- torch$tensor(list(list(0, 0), list(0, 0), list(0, 1), list(1, 1)), dtype=torch$uint8) all(d) all(d, dim=0L) all(d, dim=1L) ## End(Not run)
Returns True if any elements in the tensor are non-zero, False otherwise.
## S3 method for class 'torch.Tensor' any(x, dim, ...)## S3 method for class 'torch.Tensor' any(x, dim, ...)
x |
tensor |
dim |
dimension to reduce |
... |
other params (yet to be developed) |
A tensor of type torch.uint8 representing the boolean result: 1 for TRUE and 0 for FALSE.
## Not run: a <- torch$BoolTensor(list(TRUE, TRUE, TRUE, TRUE)) b <- torch$BoolTensor(list(FALSE, TRUE, TRUE, TRUE)) c <- torch$BoolTensor(list(TRUE, TRUE, TRUE, FALSE)) any(a) any(b) any(c) d <- torch$tensor(list(list(1, 0), list(0, 0), list(0, 1), list(0, 0)), dtype=torch$uint8) any(d) any(d, dim=0L) any(d, dim=1L) ## End(Not run)## Not run: a <- torch$BoolTensor(list(TRUE, TRUE, TRUE, TRUE)) b <- torch$BoolTensor(list(FALSE, TRUE, TRUE, TRUE)) c <- torch$BoolTensor(list(TRUE, TRUE, TRUE, FALSE)) any(a) any(b) any(c) d <- torch$tensor(list(list(1, 0), list(0, 0), list(0, 1), list(0, 0)), dtype=torch$uint8) any(d) any(d, dim=0L) any(d, dim=1L) ## End(Not run)
Convert a tensor to a boolean equivalent tensor
as_boolean(x)as_boolean(x)
x |
a torch tensor |
## Not run: uo <- torch$ones(3L) as_boolean(uo) # tensor([True, True, True], dtype=torch.bool) ## End(Not run)## Not run: uo <- torch$ones(3L) as_boolean(uo) # tensor([True, True, True], dtype=torch.bool) ## End(Not run)
Dataset of 60,000 28x28 grayscale images of the 10 digits, along with a test set of 10,000 images.
dataset_mnist_digits(ntrain = 60000L, ntest = 10000L, onehot = TRUE)dataset_mnist_digits(ntrain = 60000L, ntest = 10000L, onehot = TRUE)
ntrain |
number of training samples |
ntest |
number of test samples |
onehot |
boolean |
Get the dimensions of a tensor displaying it as a vector.
## S3 method for class 'torch.Tensor' dim(x)## S3 method for class 'torch.Tensor' dim(x)
x |
tensor |
a vector of integers with the dimensions of the tensor
## Not run: uo = torch$ones(3L, 5L) # it is a 3x5 tensor dim(uo) ## End(Not run)## Not run: uo = torch$ones(3L, 5L) # it is a 3x5 tensor dim(uo) ## End(Not run)
Install PyTorch and its dependencies
install_pytorch( method = c("conda", "virtualenv", "auto"), conda = "auto", version = "default", envname = "r-torch", extra_packages = NULL, restart_session = TRUE, conda_python_version = "3.6", pip = FALSE, channel = "stable", cuda_version = NULL, dry_run = FALSE, ... )install_pytorch( method = c("conda", "virtualenv", "auto"), conda = "auto", version = "default", envname = "r-torch", extra_packages = NULL, restart_session = TRUE, conda_python_version = "3.6", pip = FALSE, channel = "stable", cuda_version = NULL, dry_run = FALSE, ... )
method |
Installation method. By default, "auto" automatically finds a method that will work in the local environment. Change the default to force a specific installation method. Note that the "virtualenv" method is not available on Windows (as this isn't supported by PyTorch). Note also that since this command runs without privillege the "system" method is available only on Windows. |
conda |
The path to a |
version |
PyTorch version to install. The "default" version is 1.4.
You can specify a specific PyTorch version with |
envname |
Name of Python or conda environment to install within.
The default environment name is |
extra_packages |
Additional Python packages to install along with
PyTorch. If more than one package use a character vector:
|
restart_session |
Restart R session after installing (note this will only occur within RStudio). |
conda_python_version |
the Python version installed in the created conda
environment. Python 3.4 is installed by default. But you could specify for instance:
|
pip |
logical |
channel |
conda channel. The default channel is |
cuda_version |
string for the cuda toolkit version to install. For example,
to install a specific CUDA version use |
dry_run |
logical, set to TRUE for unit tests, otherwise will execute the command. |
... |
other arguments passed to |
## Not run: # install PyTorch 1.6 on Python 3.7 including pandas install_pytorch(version = "1.6", conda_python_version = "3.7", extra_packages = "pandas") # Install PyTorch 1.4, Python 3.6, pandas, matplotlib install from the console install_pytorch(version = "1.4", conda_python_version = "3.6", extra_packages = c("pandas", "matplotlib")) # Install PyTorch 1.3 on Python 3.6 including pandas, matplotlib install_pytorch(version = "1.3", conda_python_version = "3.6", extra_packages = c("pandas", "matplotlib"), dry_run = FALSE) ## End(Not run)## Not run: # install PyTorch 1.6 on Python 3.7 including pandas install_pytorch(version = "1.6", conda_python_version = "3.7", extra_packages = "pandas") # Install PyTorch 1.4, Python 3.6, pandas, matplotlib install from the console install_pytorch(version = "1.4", conda_python_version = "3.6", extra_packages = c("pandas", "matplotlib")) # Install PyTorch 1.3 on Python 3.6 including pandas, matplotlib install_pytorch(version = "1.3", conda_python_version = "3.6", extra_packages = c("pandas", "matplotlib"), dry_run = FALSE) ## End(Not run)
Determine if the object is a tensor by looking inheritance
is_tensor(obj)is_tensor(obj)
obj |
an object |
This function is equivalent to torch$numel()
## S3 method for class 'torch.Tensor' length(x)## S3 method for class 'torch.Tensor' length(x)
x |
tensor |
the number of elements of a tensor as an integer
## Not run: uo = torch$ones(3L, 5L) # tensor with 15 elements length(uo) ## End(Not run)## Not run: uo = torch$ones(3L, 5L) # tensor with 15 elements length(uo) ## End(Not run)
Logarithm of a tensor given the tensor and the base
## S3 method for class 'torch.Tensor' log(x, base = exp(1L))## S3 method for class 'torch.Tensor' log(x, base = exp(1L))
x |
a tensor |
base |
the base of the logarithm |
## Not run: x <- torch$tensor(c(512, 1024, 2048, 4096)) # tensor([ 9., 10., 11., 12.]) base <- 2 log(x, base) x <- torch$tensor(c(1, 10, 100, 1000)) # tensor([0., 1., 2., 3.]) log(x, 10) ## End(Not run)## Not run: x <- torch$tensor(c(512, 1024, 2048, 4096)) # tensor([ 9., 10., 11., 12.]) base <- 2 log(x, base) x <- torch$tensor(c(1, 10, 100, 1000)) # tensor([0., 1., 2., 3.]) log(x, 10) ## End(Not run)
Logarithm of a tensor in base 10
## S3 method for class 'torch.Tensor' log10(x)## S3 method for class 'torch.Tensor' log10(x)
x |
a tensor |
## Not run: x <- torch$tensor(c(1, 10, 100, 1000)) # tensor([0., 1., 2., 3.]) ## End(Not run)## Not run: x <- torch$tensor(c(1, 10, 100, 1000)) # tensor([0., 1., 2., 3.]) ## End(Not run)
Logarithm of a tensor in base 2
## S3 method for class 'torch.Tensor' log2(x)## S3 method for class 'torch.Tensor' log2(x)
x |
a tensor |
## Not run: x <- torch$tensor(c(512, 1024, 2048, 4096)) # tensor([ 9., 10., 11., 12.]) ## End(Not run)## Not run: x <- torch$tensor(c(512, 1024, 2048, 4096)) # tensor([ 9., 10., 11., 12.]) ## End(Not run)
There is not equivalent function in PyTorch for this generic.
To generate this generic we use the function np$logical_and().
## S3 method for class 'torch.Tensor' x & y## S3 method for class 'torch.Tensor' x & y
x |
tensor |
y |
tensor |
A tensor of booleans representing the logical result of the comparison.
False to represent 0, and True to represent 1 in a tensor of data type torch$uint8.
## Not run: A <- torch$BoolTensor(list(0L, 1L)) B <- torch$BoolTensor(list(1L, 0L)) C <- torch$BoolTensor(list(1L, 1L)) A & B C & A B & C ## End(Not run)## Not run: A <- torch$BoolTensor(list(0L, 1L)) B <- torch$BoolTensor(list(1L, 0L)) C <- torch$BoolTensor(list(1L, 1L)) A & B C & A B & C ## End(Not run)
There is not equivalent function in PyTorch for this generic.
To generate This generic we use the function np$logical_not(x).
## S3 method for class 'torch.Tensor' !x## S3 method for class 'torch.Tensor' !x
x |
tensor |
A tensor of booleans, where False corresponds to 0, and 1 to True
in a tensor of data type torch$bool.
## Not run: A <- torch$ones(5L) !A Z <- torch$zeros(5L) !Z ## End(Not run)## Not run: A <- torch$ones(5L) !A Z <- torch$zeros(5L) !Z ## End(Not run)
There is not equivalent function in PyTorch for this generic.
To generate this generic we use the function np$logical_or().
## S3 method for class 'torch.Tensor' x | y## S3 method for class 'torch.Tensor' x | y
x |
tensor |
y |
tensor |
A tensor of booleans representing the logical result of the comparison.
False to represent 0, and True to represent 1 in a tensor of data type torch$uint8.
## Not run: A <- torch$BoolTensor(list(0L, 1L)) B <- torch$BoolTensor(list(1L, 0L)) C <- torch$BoolTensor(list(1L, 1L)) A | B C | A B | C ## End(Not run)## Not run: A <- torch$BoolTensor(list(0L, 1L)) B <- torch$BoolTensor(list(1L, 0L)) C <- torch$BoolTensor(list(1L, 1L)) A | B C | A B | C ## End(Not run)
A copy of an array or tensor might be needed to prevent warnings by new PyTorch versions on overwriting the numpy object
make_copy(object, ...)make_copy(object, ...)
object |
a torch tensor or numpy array or R array |
... |
additional parameters |
This generic is approximately similar to torch$ne(a, b), with the
difference that the generic returns a tensor of booleans instead of
a tensor of data type torch$uint8.
## S3 method for class 'torch.Tensor' x != y## S3 method for class 'torch.Tensor' x != y
x |
tensor |
y |
tensor |
A tensor of booleans, where False corresponds to 0, and 1 to True
in a tensor of data type torch$bool.
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) a != b ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) a != b ## End(Not run)
One tensor operation
one_tensor_op(x) ## S3 method for class 'torch.Tensor' exp(x)one_tensor_op(x) ## S3 method for class 'torch.Tensor' exp(x)
x |
tensor |
torch.Tensor: Exponential of a tensor
## Not run: A <- torch$ones(c(60000L, 1L, 28L, 28L)) dim(A) ## End(Not run)## Not run: A <- torch$ones(c(60000L, 1L, 28L, 28L)) dim(A) ## End(Not run)
Two tensor operations
tensor_ops(a, b) ## S3 method for class 'torch.Tensor' a ^ btensor_ops(a, b) ## S3 method for class 'torch.Tensor' a ^ b
a |
tensor |
b |
tensor |
torch.Tensor: A tensor 'a' to the power of 'b'
## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a + b b - a a * b a / s a == b a == a a != a x <- torch$Tensor(list(list(2, 2, 2), list(4, 4, 4))) y <- torch$Tensor(list(list(1, 2, 1), list(3, 4, 5))) x > y x < y x >= y y <= x diag <- torch$eye(3L) zeros <- torch$zeros(c(3L, 3L)) diag & zeros diag & diag diag | diag zeros | zeros zeros & zeros diag & zeros diag | zeros ## End(Not run) ## Not run: x <- torch$arange(1,11) torch$pow(x, 2) # x^(2) torch$pow(x, -2) # x^(1/2) ## End(Not run)## Not run: a <- torch$Tensor(list(1, 1, 1)) b <- torch$Tensor(list(2, 2, 2)) s <- 2.0 a + b b - a a * b a / s a == b a == a a != a x <- torch$Tensor(list(list(2, 2, 2), list(4, 4, 4))) y <- torch$Tensor(list(list(1, 2, 1), list(3, 4, 5))) x > y x < y x >= y y <= x diag <- torch$eye(3L) zeros <- torch$zeros(c(3L, 3L)) diag & zeros diag & diag diag | diag zeros | zeros zeros & zeros diag & zeros diag | zeros ## End(Not run) ## Not run: x <- torch$arange(1,11) torch$pow(x, 2) # x^(2) torch$pow(x, -2) # x^(1/2) ## End(Not run)
Interface to main PyTorch module. Provides access to top level classes
Interface to numpy module.
Interface to Torchvision module.
torch np torchvisiontorch np torchvision
PyTorch module
numpy module
Torchvision module
Tensor extract options
torch_extract_opts( style = getOption("torch.extract.style"), ..., one_based = getOption("torch.extract.one_based", TRUE), inclusive_stop = getOption("torch.extract.inclusive_stop", TRUE), disallow_out_of_bounds = getOption("torch.extract.dissallow_out_of_bounds", TRUE), warn_tensors_passed_asis = getOption("torch.extract.warn_tensors_passed_asis", TRUE), warn_negatives_pythonic = getOption("torch.extract.warn_negatives_pythonic", TRUE) )torch_extract_opts( style = getOption("torch.extract.style"), ..., one_based = getOption("torch.extract.one_based", TRUE), inclusive_stop = getOption("torch.extract.inclusive_stop", TRUE), disallow_out_of_bounds = getOption("torch.extract.dissallow_out_of_bounds", TRUE), warn_tensors_passed_asis = getOption("torch.extract.warn_tensors_passed_asis", TRUE), warn_negatives_pythonic = getOption("torch.extract.warn_negatives_pythonic", TRUE) )
style |
one of |
... |
ignored |
one_based |
TRUE or FALSE, if one-based indexing should be used |
inclusive_stop |
TRUE or FALSE, if slices like |
disallow_out_of_bounds |
TRUE or FALSE, whether checks are performed on the slicing index to ensure it is within bounds. |
warn_tensors_passed_asis |
TRUE or FALSE, whether to emit a warning the
first time a tensor is supplied to |
warn_negatives_pythonic |
TRUE or FALSE, whether to emit
a warning the first time a negative number is supplied to |
an object with class "torch_extract_opts", suitable for passing to
[.torch.tensor()
## Not run: x <- torch$arange(1L, 10L) opts <- torch_extract_opts("R") x[1, options = opts] # or for more fine-grained control opts <- torch_extract_opts( one_based = FALSE, warn_tensors_passed_asis = FALSE, warn_negatives_pythonic = FALSE ) x[0:2, options = opts] ## End(Not run)## Not run: x <- torch$arange(1L, 10L) opts <- torch_extract_opts("R") x[1, options = opts] # or for more fine-grained control opts <- torch_extract_opts( one_based = FALSE, warn_tensors_passed_asis = FALSE, warn_negatives_pythonic = FALSE ) x[0:2, options = opts] ## End(Not run)
Get the size of a torch tensor or of torch.size object
torch_size(obj)torch_size(obj)
obj |
a torch tensor object |