Version 1.0

Return to site

Version 1.0

To select the target Windows version define BOOSTUSEWINAPIVERSION to the numeric version similar to WIN32WINNT while compiling Boost and user's code. For example: b2 release define=BOOSTUSEWINAPIVERSION=0x0501 stage The list of Windows API version numbers can be seen on this page. Finally, the version attribute specifies the XML version, of which there are two at the moment (see Wikipedia XML versions. There are slight differences between the versions, so an XML parser needs to know what it is dealing with. In most cases (for English speakers anyway), version 1.0 is sufficient.
Version 1.0.0
Version 1.0.180404.58
Installing previous versions of PyTorchWe'd prefer you install the latest version,but old binaries and installation instructions are provided below foryour convenience.
Commands for Versions >= 1.0.0v1.6.0Conda

OSXLinux and WindowsWheelOSXLinux and Windowsv1.5.1CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.5.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.4.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.2.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.1.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.0.1CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.0.0CondaOSXLinux and WindowsWheelOSXLinux and WindowsCommands for Versions < 1.0.0Via condaThis should be used for most previous macOS version installs.
To install a previous version of PyTorch via Anaconda or Miniconda,replace '0.4.1' in the following commands with the desired version(i.e., '0.2.0').
Installing with CUDA 9
conda install pytorch=0.4.1 cuda90 -c pytorch
or
conda install pytorch=0.4.1 cuda92 -c pytorch
Installing with CUDA 8 Er mapper grid format template.

conda install pytorch=0.4.1 cuda80 -c pytorch
Installing with CUDA 7.5
conda install pytorch=0.4.1 cuda75 -c pytorch
Installing without CUDA

conda install pytorch=0.4.1 -c pytorch
From source

It is possible to checkout an older version of PyTorchand build it.You can list tags in PyTorch git repository with git tag and checkout aparticular one (replace ‘0.1.9' with the desired version) with
git checkout v0.1.9
Follow the install from source instructions in the README.md of the PyTorchcheckout.
Via pipDownload the whl file with the desired version from the following html pages:
https://download.pytorch.org/whl/cpu/torch_stable.html # CPU-only build
https://download.pytorch.org/whl/cu80/torch_stable.html # CUDA 8.0 build
https://download.pytorch.org/whl/cu90/torch_stable.html # CUDA 9.0 build
https://download.pytorch.org/whl/cu92/torch_stable.html # CUDA 9.2 build
https://download.pytorch.org/whl/cu100/torch_stable.html # CUDA 10.0 build
Then, install the file with pip install [downloaded file]
Note: most pytorch versions are available only for specific CUDA versions. For example pytorch=1.0.1 is not available for CUDA 9.2
(Old) PyTorch Linux binaries compiled with CUDA 7.5These predate the html page above and have to be manually installed by downloading the wheel file and pip install downloaded_file
Windows binariesMac and misc. binariesFor recent macOS binaries, use conda:
e.g.,
conda install pytorch=0.4.1 cuda90 -c pytorchconda install pytorch=0.4.1 cuda92 -c pytorchconda install pytorch=0.4.1 cuda80 -c pytorchconda install pytorch=0.4.1 -c pytorch # No CUDA
In the world of software management there exists a dread place called'dependency hell.' The bigger your system grows and the more packages youintegrate into your software, the more likely you are to find yourself, oneday, in this pit of despair.
In systems with many dependencies, releasing new package versions can quicklybecome a nightmare. If the dependency specifications are too tight, you are indanger of version lock (the inability to upgrade a package without having torelease new versions of every dependent package). If dependencies arespecified too loosely, you will inevitably be bitten by version promiscuity(assuming compatibility with more future versions than is reasonable).Dependency hell is where you are when version lock and/or version promiscuityprevent you from easily and safely moving your project forward.
As a solution to this problem, I propose a simple set of rules andrequirements that dictate how version numbers are assigned and incremented.For this system to work, you first need to declare a public API. This mayconsist of documentation or be enforced by the code itself. Regardless, it isimportant that this API be clear and precise. Once you identify your publicAPI, you communicate changes to it with specific increments to your versionnumber. Consider a version format of X.Y.Z (Major.Minor.Patch). Bug fixes notaffecting the API increment the patch version, backwards compatible APIadditions/changes increment the minor version, and backwards incompatible APIchanges increment the major version.
I call this system 'Semantic Versioning.' Under this scheme, version numbersand the way they change convey meaning about the underlying code and what hasbeen modified from one version to the next.
Semantic Versioning Specification (SemVer)The key words 'MUST', 'MUST NOT', 'REQUIRED', 'SHALL', 'SHALL NOT', 'SHOULD','SHOULD NOT', 'RECOMMENDED', 'MAY', and 'OPTIONAL' in this document are to beinterpreted as described in RFC 2119.
Software using Semantic Versioning MUST declare a public API. This APIcould be declared in the code itself or exist strictly in documentation.However it is done, it should be precise and comprehensive.
A normal version number MUST take the form X.Y.Z where X, Y, and Z areintegers. X is the major version, Y is the minor version, and Z is the patchversion. Each element MUST increase numerically by increments of one. Forinstance: 1.9.0 -> 1.10.0 -> 1.11.0.
When a major version number is incremented, the minor version and patchversion MUST be reset to zero. When a minor version number is incremented, thepatch version MUST be reset to zero. For instance: 1.1.3 -> 2.0.0 and 2.1.7 ->2.2.0.
A pre-release version number MAY be denoted by appending an arbitrarystring immediately following the patch version and a dash. The string MUST becomprised of only alphanumerics plus dash [0-9A-Za-z-]. Pre-release versionssatisfy but have a lower precedence than the associated normal version.Precedence SHOULD be determined by lexicographic ASCII sort order. Forinstance: 1.0.0-alpha1 < 1.0.0-beta1 < 1.0.0-beta2 < 1.0.0-rc1 < 1.0.0.
Once a versioned package has been released, the contents of that versionMUST NOT be modified. Any modifications must be released as a new version.
Major version zero (0.y.z) is for initial development. Anything may changeat any time. The public API should not be considered stable.
Version 1.0.0 defines the public API. The way in which the version numberis incremented after this release is dependent on this public API and how itchanges.
Patch version Z (x.y.Z | x > 0) MUST be incremented if only backwardscompatible bug fixes are introduced. A bug fix is defined as an internalchange that fixes incorrect behavior.
Minor version Y (x.Y.z | x > 0) MUST be incremented if new, backwardscompatible functionality is introduced to the public API. It MAY beincremented if substantial new functionality or improvements are introducedwithin the private code. It MAY include patch level changes. Patch versionMUST be reset to 0 when minor version is incremented.
Major version X (X.y.z | X > 0) MUST be incremented if any backwardsincompatible changes are introduced to the public API. It MAY include minorand patch level changes. Patch and minor version MUST be reset to 0 when majorversion is incremented.
Tagging Specification (SemVerTag)This sub-specification SHOULD be used if you use a version control system(Git, Mercurial, SVN, etc) to store your code. Using this system allowsautomated tools to inspect your package and determine SemVer compliance andreleased versions.
When tagging releases in a version control system, the tag for a versionMUST be 'vX.Y.Z' e.g. 'v3.1.0'.
The first revision that introduces SemVer compliance SHOULD be tagged'semver'. This allows pre-existing projects to assume compliance at anyarbitrary point and for automated tools to discover this fact.
Why Use Semantic Versioning?This is not a new or revolutionary idea. In fact, you probably do somethingclose to this already. The problem is that 'close' isn't good enough. Withoutcompliance to some sort of formal specification, version numbers areessentially useless for dependency management. By giving a name and cleardefinition to the above ideas, it becomes easy to communicate your intentionsto the users of your software. Once these intentions are clear, flexible (butnot too flexible) dependency specifications can finally be made.
A simple example will demonstrate how Semantic Versioning can make dependencyhell a thing of the past. Consider a library called 'Firetruck.' It requires aSemantically Versioned package named 'Ladder.' At the time that Firetruck iscreated, Ladder is at version 3.1.0. Since Firetruck uses some functionalitythat was first introduced in 3.1.0, you can safely specify the Ladderdependency as greater than or equal to 3.1.0 but less than 4.0.0. Now, whenLadder version 3.1.1 and 3.2.0 become available, you can release them to yourpackage management system and know that they will be compatible with existingdependent software.
As a responsible developer you will, of course, want to verify that anypackage upgrades function as advertised. The real world is a messy place;there's nothing we can do about that but be vigilant. What you can do is letSemantic Versioning provide you with a sane way to release and upgradepackages without having to roll new versions of dependent packages, saving youtime and hassle.
If all of this sounds desirable, all you need to do to start using SemanticVersioning is to declare that you are doing so and then follow the rules. Linkto this website from your README so others know the rules and can benefit fromthem.
FAQHow should I deal with revisions in the 0.y.z initial development phase?The simplest thing to do is start your initial development release at 0.1.0and then increment the minor version for each subsequent release.
How do I know when to release 1.0.0?If your software is being used in production, it should probably already be1.0.0. If you have a stable API on which users have come to depend, you shouldbe 1.0.0. If you're worrying a lot about backwards compatibility, you shouldprobably already be 1.0.0.

To select the target Windows version define BOOSTUSEWINAPIVERSION to the numeric version similar to WIN32WINNT while compiling Boost and user's code. For example: b2 release define=BOOSTUSEWINAPIVERSION=0x0501 stage The list of Windows API version numbers can be seen on this page. Finally, the version attribute specifies the XML version, of which there are two at the moment (see Wikipedia XML versions. There are slight differences between the versions, so an XML parser needs to know what it is dealing with. In most cases (for English speakers anyway), version 1.0 is sufficient.
Version 1.0.0
Version 1.0.180404.58
Installing previous versions of PyTorchWe'd prefer you install the latest version,but old binaries and installation instructions are provided below foryour convenience.
Commands for Versions >= 1.0.0v1.6.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.5.1CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.5.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.4.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.2.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.1.0CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.0.1CondaOSXLinux and WindowsWheelOSXLinux and Windowsv1.0.0CondaOSXLinux and WindowsWheelOSXLinux and WindowsCommands for Versions < 1.0.0Via condaThis should be used for most previous macOS version installs.
To install a previous version of PyTorch via Anaconda or Miniconda,replace '0.4.1' in the following commands with the desired version(i.e., '0.2.0').
Installing with CUDA 9
conda install pytorch=0.4.1 cuda90 -c pytorch
or
conda install pytorch=0.4.1 cuda92 -c pytorch
Installing with CUDA 8 Er mapper grid format template.
conda install pytorch=0.4.1 cuda80 -c pytorch
Installing with CUDA 7.5
conda install pytorch=0.4.1 cuda75 -c pytorch
Installing without CUDA
conda install pytorch=0.4.1 -c pytorch
From sourceIt is possible to checkout an older version of PyTorchand build it.You can list tags in PyTorch git repository with git tag and checkout aparticular one (replace ‘0.1.9' with the desired version) with
git checkout v0.1.9
Follow the install from source instructions in the README.md of the PyTorchcheckout.
Via pipDownload the whl file with the desired version from the following html pages:
https://download.pytorch.org/whl/cpu/torch_stable.html # CPU-only build
https://download.pytorch.org/whl/cu80/torch_stable.html # CUDA 8.0 build
https://download.pytorch.org/whl/cu90/torch_stable.html # CUDA 9.0 build
https://download.pytorch.org/whl/cu92/torch_stable.html # CUDA 9.2 build
https://download.pytorch.org/whl/cu100/torch_stable.html # CUDA 10.0 build
Then, install the file with pip install [downloaded file]
Note: most pytorch versions are available only for specific CUDA versions. For example pytorch=1.0.1 is not available for CUDA 9.2
(Old) PyTorch Linux binaries compiled with CUDA 7.5These predate the html page above and have to be manually installed by downloading the wheel file and pip install downloaded_file
Windows binariesMac and misc. binariesFor recent macOS binaries, use conda:
e.g.,
conda install pytorch=0.4.1 cuda90 -c pytorchconda install pytorch=0.4.1 cuda92 -c pytorchconda install pytorch=0.4.1 cuda80 -c pytorchconda install pytorch=0.4.1 -c pytorch # No CUDA
In the world of software management there exists a dread place called'dependency hell.' The bigger your system grows and the more packages youintegrate into your software, the more likely you are to find yourself, oneday, in this pit of despair.
In systems with many dependencies, releasing new package versions can quicklybecome a nightmare. If the dependency specifications are too tight, you are indanger of version lock (the inability to upgrade a package without having torelease new versions of every dependent package). If dependencies arespecified too loosely, you will inevitably be bitten by version promiscuity(assuming compatibility with more future versions than is reasonable).Dependency hell is where you are when version lock and/or version promiscuityprevent you from easily and safely moving your project forward.
As a solution to this problem, I propose a simple set of rules andrequirements that dictate how version numbers are assigned and incremented.For this system to work, you first need to declare a public API. This mayconsist of documentation or be enforced by the code itself. Regardless, it isimportant that this API be clear and precise. Once you identify your publicAPI, you communicate changes to it with specific increments to your versionnumber. Consider a version format of X.Y.Z (Major.Minor.Patch). Bug fixes notaffecting the API increment the patch version, backwards compatible APIadditions/changes increment the minor version, and backwards incompatible APIchanges increment the major version.
I call this system 'Semantic Versioning.' Under this scheme, version numbersand the way they change convey meaning about the underlying code and what hasbeen modified from one version to the next.
Semantic Versioning Specification (SemVer)The key words 'MUST', 'MUST NOT', 'REQUIRED', 'SHALL', 'SHALL NOT', 'SHOULD','SHOULD NOT', 'RECOMMENDED', 'MAY', and 'OPTIONAL' in this document are to beinterpreted as described in RFC 2119.
Software using Semantic Versioning MUST declare a public API. This APIcould be declared in the code itself or exist strictly in documentation.However it is done, it should be precise and comprehensive.
A normal version number MUST take the form X.Y.Z where X, Y, and Z areintegers. X is the major version, Y is the minor version, and Z is the patchversion. Each element MUST increase numerically by increments of one. Forinstance: 1.9.0 -> 1.10.0 -> 1.11.0.
When a major version number is incremented, the minor version and patchversion MUST be reset to zero. When a minor version number is incremented, thepatch version MUST be reset to zero. For instance: 1.1.3 -> 2.0.0 and 2.1.7 ->2.2.0.
A pre-release version number MAY be denoted by appending an arbitrarystring immediately following the patch version and a dash. The string MUST becomprised of only alphanumerics plus dash [0-9A-Za-z-]. Pre-release versionssatisfy but have a lower precedence than the associated normal version.Precedence SHOULD be determined by lexicographic ASCII sort order. Forinstance: 1.0.0-alpha1 < 1.0.0-beta1 < 1.0.0-beta2 < 1.0.0-rc1 < 1.0.0.
Once a versioned package has been released, the contents of that versionMUST NOT be modified. Any modifications must be released as a new version.
Major version zero (0.y.z) is for initial development. Anything may changeat any time. The public API should not be considered stable.
Version 1.0.0 defines the public API. The way in which the version numberis incremented after this release is dependent on this public API and how itchanges.
Patch version Z (x.y.Z | x > 0) MUST be incremented if only backwardscompatible bug fixes are introduced. A bug fix is defined as an internalchange that fixes incorrect behavior.
Minor version Y (x.Y.z | x > 0) MUST be incremented if new, backwardscompatible functionality is introduced to the public API. It MAY beincremented if substantial new functionality or improvements are introducedwithin the private code. It MAY include patch level changes. Patch versionMUST be reset to 0 when minor version is incremented.
Major version X (X.y.z | X > 0) MUST be incremented if any backwardsincompatible changes are introduced to the public API. It MAY include minorand patch level changes. Patch and minor version MUST be reset to 0 when majorversion is incremented.
Tagging Specification (SemVerTag)This sub-specification SHOULD be used if you use a version control system(Git, Mercurial, SVN, etc) to store your code. Using this system allowsautomated tools to inspect your package and determine SemVer compliance andreleased versions.
When tagging releases in a version control system, the tag for a versionMUST be 'vX.Y.Z' e.g. 'v3.1.0'.
The first revision that introduces SemVer compliance SHOULD be tagged'semver'. This allows pre-existing projects to assume compliance at anyarbitrary point and for automated tools to discover this fact.
Why Use Semantic Versioning?This is not a new or revolutionary idea. In fact, you probably do somethingclose to this already. The problem is that 'close' isn't good enough. Withoutcompliance to some sort of formal specification, version numbers areessentially useless for dependency management. By giving a name and cleardefinition to the above ideas, it becomes easy to communicate your intentionsto the users of your software. Once these intentions are clear, flexible (butnot too flexible) dependency specifications can finally be made.
A simple example will demonstrate how Semantic Versioning can make dependencyhell a thing of the past. Consider a library called 'Firetruck.' It requires aSemantically Versioned package named 'Ladder.' At the time that Firetruck iscreated, Ladder is at version 3.1.0. Since Firetruck uses some functionalitythat was first introduced in 3.1.0, you can safely specify the Ladderdependency as greater than or equal to 3.1.0 but less than 4.0.0. Now, whenLadder version 3.1.1 and 3.2.0 become available, you can release them to yourpackage management system and know that they will be compatible with existingdependent software.
As a responsible developer you will, of course, want to verify that anypackage upgrades function as advertised. The real world is a messy place;there's nothing we can do about that but be vigilant. What you can do is letSemantic Versioning provide you with a sane way to release and upgradepackages without having to roll new versions of dependent packages, saving youtime and hassle.
If all of this sounds desirable, all you need to do to start using SemanticVersioning is to declare that you are doing so and then follow the rules. Linkto this website from your README so others know the rules and can benefit fromthem.
FAQHow should I deal with revisions in the 0.y.z initial development phase?The simplest thing to do is start your initial development release at 0.1.0and then increment the minor version for each subsequent release.
How do I know when to release 1.0.0?If your software is being used in production, it should probably already be1.0.0. If you have a stable API on which users have come to depend, you shouldbe 1.0.0. If you're worrying a lot about backwards compatibility, you shouldprobably already be 1.0.0.
Doesn't this discourage rapid development and fast iteration?Major version zero is all about rapid development. If you're changing the APIevery day you should either still be in version 0.x.x or on a separatedevelopment branch working on the next major version.
If even the tiniest backwards incompatible changes to the public API require a major version bump, won't I end up at version 42.0.0 very rapidly?This is a question of responsible development and foresight. Incompatiblechanges should not be introduced lightly to software that has a lot ofdependent code. The cost that must be incurred to upgrade can be significant.Having to bump major versions to release incompatible changes means you'llthink through the impact of your changes, and evaluate the cost/benefit ratioinvolved.
Documenting the entire public API is too much work!It is your responsibility as a professional developer to properly documentsoftware that is intended for use by others. Managing software complexity is ahugely important part of keeping a project efficient, and that's hard to do ifnobody knows how to use your software, or what methods are safe to call. Inthe long run, Semantic Versioning, and the insistence on a well defined publicAPI can keep everyone and everything running smoothly.
What do I do if I accidentally release a backwards incompatible change as a minor version?As soon as you realize that you've broken the Semantic Versioning spec, fixthe problem and release a new minor version that corrects the problem andrestores backwards compatibility. Remember, it is unacceptable to modifyversioned releases, even under this circumstance. If it's appropriate,document the offending version and inform your users of the problem so thatthey are aware of the offending version.
What should I do if I update my own dependencies without changing the public API?That would be considered compatible since it does not affect the public API.Software that explicitly depends on the same dependencies as your packageshould have their own dependency specifications and the author will notice anyconflicts. Determining whether the change is a patch level or minor levelmodification depends on whether you updated your dependencies in order to fixa bug or introduce new functionality. I would usually expect additional codefor the latter instance, in which case it's obviously a minor level increment.
Version 1.0.0What should I do if the bug that is being fixed returns the code to being compliant with the public API (i.e. the code was incorrectly out of sync with the public API documentation)?Use your best judgment. If you have a huge audience that will be drasticallyimpacted by changing the behavior back to what the public API intended, thenit may be best to perform a major version release, even though the fix couldstrictly be considered a patch release. Remember, Semantic Versioning is allabout conveying meaning by how the version number changes. If these changesare important to your users, use the version number to inform them.
AboutThe Semantic Versioning specification is authored by Tom Preston-Werner, inventor of Gravatars and cofounder of GitHub.
If you'd like to leave feedback, please open an issue on GitHub.
LicenseVersion 1.0.180404.58Creative Commons ― CC BY 3.0http://creativecommons.org/licenses/by/3.0/