[aoti] Add cpp loader #135374
Closed
[aoti] Add cpp loader #135374
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/135374
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 51be90c with merge base eb38ee2 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
desertfire
reviewed
Sep 6, 2024
desertfire
reviewed
Sep 6, 2024
* Added a cpp loader, AOTIModelPackageLoader, which can load the .pt2, build the .so, and create a runner. The python-facing API is that users can directly call the `run` function, whereas in cpp users can directly access the `runner_` if they are more familiar with that. I couldn't figure out how to bind the `get_runner()` function to python... * Added a new config, `aot_inductor.package_cpp_only` which will **not** package the so. This means that whenever the package is loaded, we will need to build the so. This is turned off by default so that new environments do not need to rebuild their so. The `package_cpp_only` is a feature which torchchat intends to use to provide flexibility to users. * Added a new config, `aot_inductor.metadata` which stores user-provided metadata, serialized to the pt2 as a json file. It also stores the device used when exporting, "cuda" or "cpu", so that during load time, we can use that data to determine which AOTIModelContainerRunner to use. The metadata can be accessed through `loader.get_metadata()`. TODO is to move this metadata to the toplevel `package_aoti` function so that we can remove the metadata as a config. * Separated out `package_aoti` as a standalone function, instead of it automatically being called in inductor. This is to prepare for the case where users will compile multiple models, and want to bundle it in one package. The specific use case is in torchchat, where we want to package the separately-exported encoder and decoder layers. An example of how to use this is in `test_multiple_methods`. * `load_package` will load a singular model, given the model name. * The loader doesn't support windows for now, I think I need to add some more casing to make the build commands work on windows? cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
* Added a cpp loader, AOTIModelPackageLoader, which can load the .pt2, build the .so, and create a runner. The python-facing API is that users can directly call the `run` function, whereas in cpp users can directly access the `runner_` if they are more familiar with that. I couldn't figure out how to bind the `get_runner()` function to python... * Added a new config, `aot_inductor.package_cpp_only` which will **not** package the so. This means that whenever the package is loaded, we will need to build the so. This is turned off by default so that new environments do not need to rebuild their so. The `package_cpp_only` is a feature which torchchat intends to use to provide flexibility to users. * Added a new config, `aot_inductor.metadata` which stores user-provided metadata, serialized to the pt2 as a json file. It also stores the device used when exporting, "cuda" or "cpu", so that during load time, we can use that data to determine which AOTIModelContainerRunner to use. The metadata can be accessed through `loader.get_metadata()`. TODO is to move this metadata to the toplevel `package_aoti` function so that we can remove the metadata as a config. * Separated out `package_aoti` as a standalone function, instead of it automatically being called in inductor. This is to prepare for the case where users will compile multiple models, and want to bundle it in one package. The specific use case is in torchchat, where we want to package the separately-exported encoder and decoder layers. An example of how to use this is in `test_multiple_methods`. * `load_package` will load a singular model, given the model name. * The loader doesn't support windows for now, I think I need to add some more casing to make the build commands work on windows? cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
|
@angelayi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
desertfire
reviewed
Sep 7, 2024
| }; | ||
|
|
||
| using CreateAOTIModelRunnerFunc = std::shared_ptr<AOTIModelContainerRunner> (*)( | ||
| using CreateAOTIModelRunnerFunc = std::unique_ptr<AOTIModelContainerRunner> (*)( |
There was a problem hiding this comment.
cc @EikanWang , we are updating this to return unique_ptr.
desertfire
reviewed
Sep 8, 2024
| } | ||
|
|
||
| std::tuple<std::string, std::string> AOTIModelPackageLoader:: | ||
| get_cpp_compile_command( |
There was a problem hiding this comment.
This works, but I wonder why we didn't store the whole compile option as one string in the first place?
There was a problem hiding this comment.
Nikita mentioned it would be better to store it as a set of flags instead of one string for better debugability and hackiness.
|
|
||
| std::string include_dirs_args = ""; | ||
| for (auto& arg : compile_options["include_dirs"]) { | ||
| include_dirs_args += "-I" + arg.get<std::string>() + " "; |
There was a problem hiding this comment.
Are the directories stored as absolute paths or relative paths? This seems problematic if the compilation platform and deployment platform are different...
There was a problem hiding this comment.
Discussed offline. The compiling from cpp work needs more thinking and will not be used in this PR.
There was a problem hiding this comment.
Ah I didn't consider this.. I think I need to take a closer look into the cpp build side. For now this is gated by the flag packge_cpp_only.
desertfire
approved these changes
Sep 9, 2024
malfet
approved these changes
Sep 9, 2024
|
@angelayi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
* Added a cpp loader, AOTIModelPackageLoader, which can load the .pt2, build the .so, and create a runner. The python-facing API is that users can directly call the `run` function, whereas in cpp users can directly access the `runner_` if they are more familiar with that. I couldn't figure out how to bind the `get_runner()` function to python... * Added a new config, `aot_inductor.package_cpp_only` which will **not** package the so. This means that whenever the package is loaded, we will need to build the so. This is turned off by default so that new environments do not need to rebuild their so. The `package_cpp_only` is a feature which torchchat intends to use to provide flexibility to users. * Added a new config, `aot_inductor.metadata` which stores user-provided metadata, serialized to the pt2 as a json file. It also stores the device used when exporting, "cuda" or "cpu", so that during load time, we can use that data to determine which AOTIModelContainerRunner to use. The metadata can be accessed through `loader.get_metadata()`. TODO is to move this metadata to the toplevel `package_aoti` function so that we can remove the metadata as a config. * Separated out `package_aoti` as a standalone function, instead of it automatically being called in inductor. This is to prepare for the case where users will compile multiple models, and want to bundle it in one package. The specific use case is in torchchat, where we want to package the separately-exported encoder and decoder layers. An example of how to use this is in `test_multiple_methods`. * `load_package` will load a singular model, given the model name. * The loader doesn't support windows for now, I think I need to add some more casing to make the build commands work on windows? cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang Differential Revision: [D62329906](https://our.internmc.facebook.com/intern/diff/D62329906) [ghstack-poisoned]
* Added a cpp loader, AOTIModelPackageLoader, which can load the .pt2, build the .so, and create a runner. The python-facing API is that users can directly call the `run` function, whereas in cpp users can directly access the `runner_` if they are more familiar with that. I couldn't figure out how to bind the `get_runner()` function to python... * Added a new config, `aot_inductor.package_cpp_only` which will **not** package the so. This means that whenever the package is loaded, we will need to build the so. This is turned off by default so that new environments do not need to rebuild their so. The `package_cpp_only` is a feature which torchchat intends to use to provide flexibility to users. * Added a new config, `aot_inductor.metadata` which stores user-provided metadata, serialized to the pt2 as a json file. It also stores the device used when exporting, "cuda" or "cpu", so that during load time, we can use that data to determine which AOTIModelContainerRunner to use. The metadata can be accessed through `loader.get_metadata()`. TODO is to move this metadata to the toplevel `package_aoti` function so that we can remove the metadata as a config. * Separated out `package_aoti` as a standalone function, instead of it automatically being called in inductor. This is to prepare for the case where users will compile multiple models, and want to bundle it in one package. The specific use case is in torchchat, where we want to package the separately-exported encoder and decoder layers. An example of how to use this is in `test_multiple_methods`. * `load_package` will load a singular model, given the model name. * The loader doesn't support windows for now, I think I need to add some more casing to make the build commands work on windows? cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang Differential Revision: [D62329906](https://our.internmc.facebook.com/intern/diff/D62329906) [ghstack-poisoned]
|
@angelayi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Chao1Han
pushed a commit
to Chao1Han/pytorch
that referenced
this pull request
Sep 20, 2024
* Added a cpp loader, AOTIModelPackageLoader, which can load the .pt2, build the .so, and create a runner. The python-facing API is that users can directly call the `run` function, whereas in cpp users can directly access the `runner_` if they are more familiar with that. I couldn't figure out how to bind the `get_runner()` function to python... * Added a new config, `aot_inductor.package_cpp_only` which will **not** package the so. This means that whenever the package is loaded, we will need to build the so. This is turned off by default so that new environments do not need to rebuild their so. The `package_cpp_only` is a feature which torchchat intends to use to provide flexibility to users. * Added a new config, `aot_inductor.metadata` which stores user-provided metadata, serialized to the pt2 as a json file. It also stores the device used when exporting, "cuda" or "cpu", so that during load time, we can use that data to determine which AOTIModelContainerRunner to use. The metadata can be accessed through `loader.get_metadata()`. TODO is to move this metadata to the toplevel `package_aoti` function so that we can remove the metadata as a config. * Separated out `package_aoti` as a standalone function, instead of it automatically being called in inductor. This is to prepare for the case where users will compile multiple models, and want to bundle it in one package. The specific use case is in torchchat, where we want to package the separately-exported encoder and decoder layers. An example of how to use this is in `test_multiple_methods`. * `load_package` will load a singular model, given the model name. * The loader doesn't support windows for now, I think I need to add some more casing to make the build commands work on windows? Differential Revision: [D62329906](https://our.internmc.facebook.com/intern/diff/D62329906) Pull Request resolved: pytorch#135374 Approved by: https://github.com/desertfire, https://github.com/malfet
This was referenced Oct 3, 2024
kit1980
added a commit
that referenced
this pull request
Oct 3, 2024
* Apply changes from #135374 * Fix dependency on filesystem on Linux (#137209) Similar to: #134494 We are seeing come back of #133437 due to use of filesystem on Linux Pull Request resolved: #137209 Approved by: https://github.com/kit1980, https://github.com/malfet --------- Co-authored-by: atalman <atalman@fb.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
-> [aoti] Add cpp loader #135374
Added a cpp loader, AOTIModelPackageLoader, which can load the .pt2, build the .so, and create a runner. The python-facing API is that users can directly call the
runfunction, whereas in cpp users can directly access therunner_if they are more familiar with that. I couldn't figure out how to bind theget_runner()function to python...Added a new config,
aot_inductor.package_cpp_onlywhich will not package the so. This means that whenever the package is loaded, we will need to build the so. This is turned off by default so that new environments do not need to rebuild their so. Thepackage_cpp_onlyis a feature which torchchat intends to use to provide flexibility to users.Added a new config,
aot_inductor.metadatawhich stores user-provided metadata, serialized to the pt2 as a json file. It also stores the device used when exporting, "cuda" or "cpu", so that during load time, we can use that data to determine which AOTIModelContainerRunner to use. The metadata can be accessed throughloader.get_metadata(). TODO is to move this metadata to the toplevelpackage_aotifunction so that we can remove the metadata as a config.Separated out
package_aotias a standalone function, instead of it automatically being called in inductor. This is to prepare for the case where users will compile multiple models, and want to bundle it in one package. The specific use case is in torchchat, where we want to package the separately-exported encoder and decoder layers. An example of how to use this is intest_multiple_methods.load_packagewill load a singular model, given the model name.The loader doesn't support windows for now, I think I need to add some more casing to make the build commands work on windows?
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang
Differential Revision: D62329906