NEWS.md

In rTorch: R Bindings to 'PyTorch'

0.4.2

• Release to CRAN after fix.

0.4.1.9000

• Returned from CRAN with note: shouldn't link using []() or local link. Must use full URL if want to link to cran-comments.md.

0.4.1

• to CRAN after fixes.

0.4.0.9006

• Try with Solaris: pandoc (>= 2.0), qpdf ( >= 7.0);. Getting now two notes and a warrning.
• add Solaris: pkgutil -y -i qpdf, pkgutil -y -i pandoc" to SysReqs.
• use skip_on_cran() in test_r_torch_share_objects.R, test_types.R, and test-install_rtorch_dryrun.R. Causing errors in Fedora. It doesn't want to install numpy but now errors went away in Fedora and Solaris because is nos tested on numpy.
• remove line importing numpy at the top of test file
• Instead use qpdf, pandoc (>= 2.7.2) on Solaris".
• add numpy ( >= 1.14.0) for Fedora.
• add Solaris solaris-x86-patched platform to rhub.
• Use a different SystemRequirements in DESCRIPTION: SystemRequirements: "conda (python=3.6 pytorch torchvision cpuonly matplotlib pandas -c pytorch), Python (>=3.6), pytorch (>=1.6), torchvision, numpy". It makes no difference in Fedora.
• test with fedora-clang-devel in rhub. Still throwing error ModuleNotFoundNo module named 'numpy'. failed. It seems tha Fedora cannot install numpy as painless as in Debian or Ubuntu.
• start with one of the tests in test_types.R. Put quotes at the beginning of the function parentheses in Python code.
• received message from CRAN after version 0.4.0 been accepted. Errors in Fedora and Solaris. Will use rhub to debug.

0.4.0.9005

• add unit tests for new functions
• add functions sign, abs, sqrt, floor, ceil, round, sin, cos, tan, asin, acos, atan in generics.R. Documented.

0.4.0.9004

• add file tests/testthat/rhub-tests.R that sends rTorch for testing on three different platforms. Use it as well in addition to Travis and Appveyor. Closer to CRAN tests.
• removing data.table and R6 from Imports. Not used yet.
• add more live tests. They work but torch_config does not update output after issuing a install_pytorch() command; they return the previous installed PyTorch info. The purpose of these test was initially see if the installation performed as initially planned. It works outside unit tests. But we cannot enable this test for CRAN because it will take longer time and may not work due to the PyTorch installation process. The major problem I found with these tests is that the torch_config objects do not update after issuing a new install_pytorch.
• Links to tutorials added to README.
• Remove reticulate.R from rTorch code. Some functions were previously customized to accept channel. Now the reticulate package accepts channel as a function parameter.

0.4.0.9003

• Now CRAN WinBuilder tests are passing.
• Add conda to SystemRequirements because CRAN is not passing.

0.4.0.9002

• add skip_if_no_python() so CRAN doesn't throw error in unit test test-install_rtorch_dryrun.R
• replace function name utils.R by helper_utils.R. We also have utils.R under the R folder.

0.4.0.9001

• split test test-install_rtorch_dryrun.R in two files. The second one test-install_rtorch_parse_version.R will only perform the parsing of what is being sent to install_pytorch().
• use skip_if_no_python() in case there is no way Python is installed at the testing point.
• using package rhub for testing before releasing to CRAN.

0.4.0.9000

• use skip_if_no_torch() in tests in cases where PyTorch cannot be installed in CRAN.
• add function skip_if_no_python().
• move live torch example to a separate file.
• change \dontest to \dontrun in examples.
• create branch 0.4.0-fix_examples_problem_in_cran.

rTorch 0.4.0

• Modify tests/testthat/utils.R to include skip_on_cran()
• change version numbering so it is easier to renumber when back to CRAN for fixes.

rTorch 0.0.4.9000

• Returned from CRAN because of errors. Mainly due to lack of Python and PyTorch installation.

rTorch 0.0.4

• Release to CRAN
• Updated version of rTorch to adapt to new PyTorch versions 1.4, 1.5, 1.6.
• This rTorch release has been tested against Travis Linux, macOs, and Appveyor for Windows. All tests passed successfully. Furthermore, the package has been tested under Python 3.6, 3.7 and 3.8. A testing matrix was implemented in Travis and Appveyor to test version combinations of Python, PyTorch and R. The R versions tested were R-3.4.3, R-3.5.3, R-3.6.3, and R-4.0.2.

rTorch 0.0.3.9013

• Fixed travis.yml by bringing - PYTHON_V="3.7" PYTORCH_V="1.6" near env: metrix. Maybe some space or alignment was preventing ennvronment variables being passed to Travis containers.
• Travis test passing with Python 3.8 in Linux and macOS. Environment variables are not being passed.
• add function is_rtorch_env_name() and env_name object to torch_config() to live unit test install_pytorch()
• add function install_pytorch() to vignette
• add parameter python_version to function conda_install()
• add backticks to roxygen text since now we are using Markdown in Roxygen: list(markdown = TRUE)
• new pkgdown section for Installation. Add two logical functions
• use markdown in roxygen help text
• Travis test PyTorch 1.5 in R-4.0.2 for Linux and macOS with variable shortened.
• Appveyor test PyTorch 1.6 in R-4.0.2 for Windows.
• New Appveyor test with PyTorch 1.5 in R-4.0.2 for Windows. Failing. Definitely PyTorch 1.5 failing in most of the tests.
• Shorten the variable names PYTORCH_VERSION and PYTHON_VERSION in Travis.
• Appveyor test PyTorch 1.6 in R-4.0.2 for Windows. Passed.
• Appveyor test PyTorch 1.5 in R-4.0.2 for Windows. Failed.
• Travis test PyTorch 1.5 in R-4.0.2 for Linux and macOS
• change logical and, or and not to be boolean or uint8 as their inputs.
• do the same for equal and not equal.
• add a parameter to force to return boolean values instead of uint8 types. Currently, AND ("!") and OR ("|") return booleans while NOT and others don't; they return uint8. We should fix this lack of consistency.
• testing on macOS in Travis.
• add condition when PyTorch is 1.1 or lower to compare against uint8. Newer PyTorch versions make the conversion of the comparison and return boolean values. In 1.1 they return uint8.
• modify functions torch$eq() and torch$ne() to validate boolean inputs.
• modify tests for eq() and ne() in PyTorch 1.1. They return tensor(True, dtype=torch.bool) or tensor(False, dtype=torch.bool).
• add more examples in generics.R and properties.R

rTorch 0.0.3.9012

• Finding a problem when using PyTorch 1.1 in logical operations.
• logical generic functions should return uint8 types as original PyTorch functions in generics.R.
• new unit tests for torch$all, torch$any and some generic logicals in test-tensor_comparison.R.
• switch to a couple of Travis and Appveyor tests to save time.
• modify generic !.torch.Tensor to return boolean if input is boolean, otherwise return opriginal type. Fix tests in test_generics.R and test_numpy_logical.R.
• tests for 4 PyTorch versions in R-4.0.2 and Python 3.7,

rTorch 0.0.3.9011

• Because PyTorch 1.1 and 1.2 are failing on Python 3.8, we could install a custom pytorch with install_pytorch(conda_python_version = "3.8", version = "1.2"). Tests failed. But not because of PyTorch but conflict during the conda installation.
• Other custom pytorch with install_pytorch(conda_python_version = "3.8", version = "1.4") with tests passed.
• add numpy version to printout of rtorch_config().
• perform rebase to get rid off wrong settings for matrix jobs in travis. Keep only those settings that worked.
• maybe a good idea to remove tests with Python 3.8 because they fail with all PyTorch versions.
• add environment variable PYTHON_VERSION to conda in build_script of appveyor. Twelve (12) passed.
• Duplicate matrix for Travis tests. Now we have tests for Python 3.6 and Python 3.7, and 3.8 for only PyTorch 1.6, a total of 36 tests. All passed.
• Duplicate matrix for Appveyor tests. Now we have tests for Python 3.6 and Python 3.7, a total of 36 tests. All passed.
• Testing develop branch with Travis and Appveyor. All tests passed.

rTorch 0.0.3.9010

• 1. branch 0.0.3.9010-fix-appveyor
• modify appveyor script. currently failing
• remove suffix -cpu from pytorch and torchvision packages from appveyor.yml. still failing because of python version is 3.6.1.
• change python version to 3.6 in appveyor.yml. Passed.
• appveyor still failing with error package 'remotes' was installed before R 4.0.0: please re-install it. repo f0nzie/r-appveyor requires some changes.
• updating file appveyor-tool.ps1 in r-appveyor repo. changes related to Rtools4.
• Windows tests with appveyor have been so far with pytorch=1.1.0. Will change to pytorch=1.4.
• Testing Linux with python=3.7 and pytorch=1.4. All R versions passed.
• clean up DESCRIPTION. remove ctb. will credit them in README.
• Testing Linux with python=3.7 and pytorch=1.6. All R versions passed.
• Windows tests with pytorch=1.4 and R-4.0.2. Passed.
• Windows tests (matrix) with pytorch=1.2, 1.4, 1.6 and R-4.0.2. Passed.
• Windows tests (matrix) with two version of R: 4.0.2 and 3.6.3 over pytorch, 1.1, 1.2, 1.4, and 1.6. using appveyor variable R_VERSION.
• Linux tests at python=3.8 and pytorch=1.1 failing for all R versions. Rest of tests passed.
• Windows tests (matrix) with two version of R: 4.0.2, 3.6.3 and 3.5.3 over pytorch, 1.1, 1.2, 1.4, and 1.6. using appveyor variable R_VERSION.

rTorch 0.0.3.9009

• 20200911
• add new vignette for PyTorch installation details
• clean up and imporve README
• tested on PyTorch 1.4 on macOS and Linux. All passed
• tested on PyTorch 1.6 on Linux. Passed.
• tested on PyTorch 1.2 on Linux. Passed.
• tested on PyTorch 1.2 on macOS. Passed but R-3.4.3.

rTorch 0.0.3.9008

• 1. Branch 0.0.3.9008-implement-todo-items
• add test in test_numpy_logical.R to check sample tensors
• add test in test_info.R add test to check the version three components
• regenerate pkgdown site. add make_copy function

rTorch 0.0.3.9007

• 1. Branch 0.0.3.9007-fix-auto-load-torch
• simplify imports in package.R
• provide function for help handler after change in on_load()
• fix function make_copy() to consider when an object have multiple classes. Use any for the logical selection
• test o Travis for macOS and R-4.0.2, R-3.6.3 with pytorch=1.4. PASSED
• test o Travis for Linux Xenial and R-4.0.2, R-3.6.3 with pytorch=1.4. PASSED

rTorch 0.0.3.9006

• 1. Fix install_pytorch() and parse_torch_version().
• Modify functions install_pytorch() and parse_torch_version()
• add new unit tests for install_pytorch() and parse_torch_version()
• add the dry_run option to install_pytorch() to use output values in unit tests
• new unit tests file test-install_commands.R

rTorch 0.0.3.9005

• 20200829
• rename function in tests from tensor_dim_ to tensor_ndim
• Dockerfile now using environment variables for name and version of the package insaide the script.
• export function make_copy() moved from unit test utilities.
• Update README. Remove mention to torch$index (not applicable).
• Add more installation instructions for PyTorch.
• Clarify some examples in the README. Use message() instead of print()

rTorch 0.0.3.9004

• 20200828
• All tests are passing in Travis on R-4.0.0, R-3.6.3, R-3.5.3 and R-3.4.3.
• Tests that are failing are in the examples.
• Error is RuntimeError: Expected object of scalar type Byte but got scalar type Long in [.torch.Tensor at generics functions.
• Example causing error is a verification of the tensor (all(y[all_dims(), 1] == y[,,,,1]) == torch$tensor(1L))$numpy(). In older versions of R it works. We could change the test to something like as.logical((all(y[all_dims(), 1] == y[,,,,1]))$numpy()) == TRUE. Tested in R-3.6.3 locally and PASSED. Will test via Travis.
• All tests in Ubuntu xenial with PyTorch 1.1 using Travis passed.
• Testing R-4.0.0 with PyTorch 1.1 generates 28 warnings test_torch_core.R:211: warning: narrow the condition has length > 1 and only the first element will be used but all test passed.
• integrating Docker with rTorch. The Docker container will create an equivalent Travis machine to save time during tests.
• Adding option - if [ "$TRAVIS_OS_NAME" = "osx" ]; then conda install nomkl;fi in .travis.yml to be able to get rid off an error related to OMP
• merging branch 003.9004-fix-examples-torch-byte-to-long with develop.
• will start testing with PyTorch 1.4 as the average version. Installing PyTorch 1.4 with

> rTorch:::install_conda(package="pytorch=1.4", envname="r-torch", conda="auto", conda_python_version = "3.6", pip=FALSE, channel="pytorch", extra_packages=c("torchvision", "cpuonly", "matplotlib", "pandas"))

rTorch 0.0.3.9003

• 20200814
• creating branch, make active fix-readme-add-tests.
• Using https://travis-ci.org/
• combine tensor_functions.R and utils.R
• unit tests for transpose and permute
• Getting this warning during check: checkRd: (5) rTorch.Rd:0-7: Must have a \description. Also stops in travis-ci.org.
• Switching from PyTorch 1.6 to 1.1 to debug error in rTorch.Rd
• Fixed problem with rTorch.Rd. Block in package.R needed description. Added this extra line below the title: #' PyTorch bindings for R. The problem originated by the new R version.
• Re-install PyTorch 1.6 with rTorch:::install_conda(package="pytorch=1.6", envname="r-torch", conda="auto", conda_python_version = "3.6", pip=FALSE, channel="pytorch", extra_packages=c("torchvision", "cpuonly", "matplotlib", "pandas")). Run tests. Run devtools::check(). All passed.
• Add --run-donttest option to check() arguments. Getting errors.
• Fix all_dims() examples in generics.R.
• Fix logical_not() examples in generics.R.
• Fix [.torch.Tensor examples in extract.R.
• Fix torch_extract_opts examples in extract.R.
• Travis stopping on error in dontrun examples that passed in the local machine. What is different is the PyTorch version specified in .travis.yml. Changing variable from "1.1"" to PYTORCH_VERSION="1.6".
• Travis stopping on error related to suffix pytorch-cpu==1.6 in command 'rTorch::install_pytorch(method="conda", version=Sys.getenv("PYTORCH_VERSION"), channel="pytorch", conda_python_version="3.6")'. We need to modify function install_pytorch().
• tests to be performed with R version 4.0.0 (2020-04-24) -- "Arbor Day"
• first, remove installation of gcc or libstdc++
• remove rTorch::pytorch_install(). Use instead rTorch:::conda_install().
• create environment variables for PYTORCH_VERSION, PYTHON_VERSION and LD_LIBRARY_PATH.
• remove symbolic link to libstdc++.so.6 in the Linux installation. This is confusing Python.
• export LD_LIBRARY_PATH=${TRAVIS_HOME}/miniconda/lib.
• install required packages with Rscript -e 'install.packages(c("logging", "reticulate", "jsonlite", "R6", "rstudioapi", "data.table"))
• reduce size of tensor in test_tensor_dim.R because throwing error due to lack of memory.
• after careful revision no more errors in Linux. Only one NOTE: * checking for future file timestamps ... NOTE. unable to verify current time.
• all tests running fine with R-4.0.0. Will change version to R-3.6.3.
• all tests running fine with R-3.5.3. Multiple R versions through Travis.
• all tests running fine with R-3.4.3.
• all tests passed in macOS with versions 3.6.3, 3.5.3 and 3.4.3.

rTorch 0.0.3.9002

• 20200810
• creating branch fix-elimination-cpu-suffix to address removal of suffix by developer.
• Installed PyTorch 1.1 with rTorch:::install_conda(package="pytorch=1.1", envname="r-torch", conda="auto", conda_python_version = "3.6", pip=FALSE, channel="pytorch", extra_packages=c("torchvision", "cpuonly", "matplotlib", "pandas"))
• revise unit tests and fix version dependence. Two test failing since last release PyTorch 1.1. Four tests failing with PyTorch 1.6 installed. Related to versioning checks.
• All tests in README passing and running.
• fixed tests in test_types.R. Minor changes in reticulate makes it more sensitive.
• set aside check on mnist dataset until internal tests are resolved
• install PyTorch 1.6 on Python 3.6`. Restart RStudio.
• fix version test with VERSIONS <- c("1.1", "1.0", "1.2", "1.3", "1.4", "1.5", "1.6") in test_info.R
• With PyTorch 1.6 we are getting the warning extract syntaxsys:1: UserWarning: The given NumPy array is not writeable, and PyTorch does not support non-writeable tensors. This means you can write to the underlying (supposedly non-writeable) NumPy array using the tensor. You may want to copy the array to protect its data or make it writeable before converting it to a tensor. This type of warning will be suppressed for the rest of this program. (Triggered internally at /opt/conda/conda-bld/pytorch_1595629417679/work/torch/csrc/utils/tensor_numpy.cpp:141.).
• add test custom function expect_all_true() in utils.R that shortens a test with multiple TRUE returning from condition
• Fix overwriting warning by adding r_to_py to R array and then copying with r_to_py(r_array)$copy() before converting to tensor
• five test files giving NumPy overwrite warning test-generic-methods.R, test_generics.R, test_numpy_logical.R, test_tensor_slicing.R, test_torch_core.R
• change functions tensor_logical_and() and tensor_logical_or() in generics.R -which use NumPy logical functions - to make a copy before converting the numpy array to a tensor
• change as_tensor() function in tensor_functions.R with torch$as_tensor(). Use make_copy() to prevent PyTorch warning.

rTorch 0.0.3.9001

• 20190918
• Rename tensorflow old labels to pytorch

rTorch 0.0.3.9000

• 20190805
• Now in CRAN. 10:00
• Announce in LinkedIn and Twitter

rTorch 0.0.3

• 20190802
• Released to CRAN at 15:20

rTorch 0.0.2.9000

• 20190802
• Returned from CRAN with notes
• Fix single quotes in DESCRIPTION
• Change \dontrun by \donttest where applicable
• Get rid of a warning message
• Replace print/cat by message/warning
• Add \value to all functions with @return
• Added cran-comments.md

rTorch 0.0.2

• July 31 2019
• Submitted to CRAN at 15:30. Received. Waiting manual inspection.
• Test with appveyor.yml
• Created repository r-appveyor at Oil Gains Analytics GitHub account. appveyor scripts now are called from this repo. Original source is at krlmlr/r-appveyor
• Test with .travis.yml
• Copy three functions from reticulate to customize it and be able to specify the conda channel. Using pytorch channel in reticulate.R.
• Specify torch-cpu and torchvision-cpu in install.R
• Move out vignettes to reduce testing time. Will ship separately using rsuite.

rTorch 0.0.1.9013

• July 26 2019
• Vignettes temporarily moved to inst/vignettes to reduce build time of package
• Add function remainder for tensors. Equivalent to a %% b
• Change unit tests in test_generics.R to use new function expect_true_tensor
• Enhance functions any and all. Add examples
• Add roxygen documentation to two tensor operations
• Change download folders for MNIST datasets under project folder

rTorch 0.0.1.9012

• July 24 2019
• Change MNIST download folder to ~/raw_data instead of inst/
• On vignette mnist_fashion_inference.Rmd:
• Add dropout class to reduce overfitting
• Add a training loop for the dropout class
• Added/remove experimental code to replicate the Python function to visualize the image along with the bar plot. Unsuccessful because R (image) and Python image (plt.imshow) functions use different array dimensions.

rTorch 0.0.1.9011

• July 22 2019
• Added vignette mnist_fashion_inference.Rmd.
• Added vignette simple_linear_regression.Rmd.
• Add generic ! (logical not)
• Fix generics any, all using as_tensor() instead of tensor()

rTorch 0.0.1.9010

• July 22 2019
• New vignette using PyTorch builtin functions and classes. Rainfall dataset: linear_regression_rainfall_builtins.Rmd
• Add comments to linear_regression_rainfall.Rmd

rTorch 0.0.1.9009

• July 22 2019
• Fix version numbers. Missing the number one.

rTorch 0.0.1.9008

• July 22 2019
• Refresh pkgdown
• Export html files for pkgdown. Modify .gitignore.

rTorch 0.0.1.9006

• July 22 2019
• Add pkgdown website

rTorch 0.0.1.9005

• July 22 2019
• Add vignette png_images_minist_digits.Rmd. It uses PBG images in a local folder instead of downloading MNIST idx format images.
• Add logical operators to README.

rTorch 0.0.1.9004

• July 22 2019
• Add vignette idx_images_minist_digits.Rmd

rTorch 0.0.1.9003

• July 21 2019
• New vignette two_layer_neural_network.Rmd. Had some problem with the tensor types. Fixed by using shorter generic version of the tensor gradient operation.

rTorch 0.0.1.9002

• July 21 2019
• Add two more vignettes.
• Get rid of a warning on roxygen documentation
• Remove old code from generics.R

rTorch 0.0.1.9001

• July 21 2019
• Adding first example as a vignette.
• import Python torch with py_run_string("import torch")

rTorch 0.0.1

• July 21 2019
• alpha version
• first release to Github
• package coming after publication of rpystats-apollo11
• still examples to be added

rTorch 0.0.0.9000

• Added a NEWS.md file to track changes to the package.

rTorch documentation built on Jan. 13, 2021, 4:32 p.m.