Handle error "Failed to load PyTorch native library"

[koppor]

While trying to reproduce #105, I closed JabRef during downloading. Then I restarted JabRef. Then, I got the error

Failed to load PyTorch native library

Click on "Try to rebuild again" causes the same effect.

[koppor]

ERROR: An error occurred while building the embedding model: ai.djl.engine.EngineException: Failed to load PyTorch native library                                                                                              at [email protected]/ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:90)                                                                                                                           at [email protected]/ai.djl.pytorch.engine.PtEngineProvider.getEngine(PtEngineProvider.java:41)                                                                                                             at [email protected]/ai.djl.engine.Engine.getEngine(Engine.java:190)                                                                                                                                                   at [email protected]/ai.djl.Model.newInstance(Model.java:99)                                                                                                                                                           at [email protected]/ai.djl.repository.zoo.BaseModelLoader.createModel(BaseModelLoader.java:196)                                                                                                                       at [email protected]/ai.djl.repository.zoo.BaseModelLoader.loadModel(BaseModelLoader.java:159)                                                                                                                         at [email protected]/ai.djl.repository.zoo.Criteria.loadModel(Criteria.java:174)                                                                                                                                       at [email protected]/org.jabref.logic.ai.models.DeepJavaEmbeddingModel.<init>(DeepJavaEmbeddingModel.java:23)                                                                                                         at [email protected]/org.jabref.logic.ai.models.EmbeddingModel.rebuild(EmbeddingModel.java:128)                                                                                                                       at [email protected]/org.jabref.gui.util.BackgroundTask$2.call(BackgroundTask.java:91)                                                                                                                                at [email protected]/org.jabref.gui.util.BackgroundTask$2.call(BackgroundTask.java:88)                                                                                                                                at [email protected]/org.jabref.gui.util.UiTaskExecutor$1.call(UiTaskExecutor.java:170)                                                                                                                               at [email protected]/javafx.concurrent.Task$TaskCallable.call(Task.java:1399)                                                                                                                                     at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:317)                                                                                                                                          2024-08-05 16:42:54 [JavaFX Application Thread] org.jabref.gui.JabRefDialogService.notify()                                                                                                                                    at java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:572)                                                                                                                                   at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:317)                                                                                                                                          INFO: An error occurred while building the embedding model                                                                                                                                                                     at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144)                                                                                                                           at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642)                                                                                                                           at java.base/java.lang.Thread.run(Thread.java:1583)                                                                                                                                                            Caused by: java.lang.UnsatisfiedLinkError: C:\Users\WDAGUtilityAccount\.djl.ai\pytorch\2.3.1-cpu-win-x86_64\fbgemm.dll: Can't find dependent libraries                                                                         at java.base/java.lang.ClassLoader.loadLibrary(ClassLoader.java:2418)                                                                                                                                                  at java.base/java.lang.Runtime.load0(Runtime.java:852)                                                                                                                                                                 at java.base/java.lang.System.load(System.java:2025)                                                                                                                                                                   at [email protected]/ai.djl.pytorch.jni.LibUtils.loadNativeLibrary(LibUtils.java:379)                                                                                                                       at [email protected]/ai.djl.pytorch.jni.LibUtils.loadLibTorch(LibUtils.java:195)                                                                                                                            at [email protected]/ai.djl.pytorch.jni.LibUtils.loadLibrary(LibUtils.java:82)                                                                                                                              at [email protected]/ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:53)

[koppor]

Caused by: java.lang.UnsatisfiedLinkError: C:\Users\WDAGUtilityAccount\.djl.ai\pytorch\2.3.1-cpu-win-x86_64\fbgemm.dll: Can't find dependent libraries

[InAnYan]

I also had this issue, and I don't know what to do with it...

I tried to restart JabRef and it worked okay

[koppor]

Does not work at my side:

Caused by: java.lang.UnsatisfiedLinkError: C:\Users\vagrant\.djl.ai\pytorch\2.3.1-cpu-win-x86_64\fbgemm.dll: Can't find dependent libraries                                                                                                                       at java.base/java.lang.ClassLoader.loadLibrary(ClassLoader.java:2418)                                                        at java.base/java.lang.Runtime.load0(Runtime.java:852)                                                                       at java.base/java.lang.System.load(System.java:2025)                                                                         at [email protected]/ai.djl.pytorch.jni.LibUtils.loadNativeLibrary(LibUtils.java:379)                             at [email protected]/ai.djl.pytorch.jni.LibUtils.loadLibTorch(LibUtils.java:195)                                  at [email protected]/ai.djl.pytorch.jni.LibUtils.loadLibrary(LibUtils.java:82)                                    at [email protected]/ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:53)                                 ... 18 more                                                                                      

This refs #105.

Maybe if the download can be triggered explicitly, this can be "solved" by a workaround: User clicks download again. - Or a retry of the download by JabRef automatically? (refs #87)

[InAnYan]

But there is a button "Rebuild"? (It was named "Try to rebuild" some commits ago)

[InAnYan]

Hmm, if we only had ways to reproduce this bug. Because I tried to test it, but it doesn't appear again

[koppor]

On Linux Mint, the download is very fast. Thus, it is a Windows issue.

Reproduce:

1. Install VirtualBox (Windows howto at https://github.com/JabRef/jabref/tree/main/scripts/vms - you can update for Linux if you run linux)
2. Install Vagrant
3. cd scripts/vm/windows
4. vagrant up
5. Wait until Windows VM is up (approx 10 Minuts)
6. Login using vagrant as Password
7. Open cmd
8. git clone https://github.com/JabRef/jabref.git
9. cd jabref
10. git checkout ai-pr-1
11. gradlew run
12. Go to Settings -> Ai
13. Enable AI
14. Click on Save
15. Quit JabRef
16. Kill the running downloading jobs
17. Press Ctrl+C at the gradle output to be sure, all is cut

Result:

2024-08-07 10:53:47 [pool-2-thread-3] ai.djl.pytorch.jni.LibUtils.downloadPyTorch()                                                                                                                  INFO: Downloading https://publish.djl.ai/pytorch/2.3.1/cpu/win-x86_64/native/lib/asmjit.dll.gz ...                                                                                                   2024-08-07 10:53:47 [main] org.jabref.Launcher.main()                                                                                                                                                ERROR: Unexpected exception: java.lang.RuntimeException: Exception in Application stop method                                                                                                                at [email protected]/com.sun.javafx.application.LauncherImpl.launchApplication1(LauncherImpl.java:898)                                                                                          at [email protected]/com.sun.javafx.application.LauncherImpl.lambda$launchApplication$2(LauncherImpl.java:196)                                                                                  at java.base/java.lang.Thread.run(Thread.java:1583)                                                                                                                                          Caused by: java.util.NoSuchElementException: java.lang.IndexOutOfBoundsException                                                                                                                             at java.base/java.util.AbstractList$Itr.next(AbstractList.java:379)                                                                                                                                  at java.base/java.util.Iterator.forEachRemaining(Iterator.java:133)                                                                                                                                  at java.base/java.util.Spliterators$IteratorSpliterator.forEachRemaining(Spliterators.java:1939)                                                                                                     at java.base/java.util.stream.ReferencePipeline$Head.forEach(ReferencePipeline.java:762)                                                                                                             at [email protected]/org.jabref.gui.util.UiTaskExecutor.shutdown(UiTaskExecutor.java:131)                                                                                                           at [email protected]/org.jabref.gui.JabRefGUI.shutdownThreadPools(JabRefGUI.java:375)                                                                                                               at [email protected]/org.jabref.gui.JabRefGUI.stop(JabRefGUI.java:365)                                                                                                                              at [email protected]/com.sun.javafx.application.LauncherImpl.lambda$launchApplication1$10(LauncherImpl.java:858)                                                                                at [email protected]/com.sun.javafx.application.PlatformImpl.lambda$runAndWait$12(PlatformImpl.java:483)                                                                                        at [email protected]/com.sun.javafx.application.PlatformImpl.lambda$runLater$10(PlatformImpl.java:456)                                                                                          at java.base/java.security.AccessController.doPrivileged(AccessController.java:400)                                                                                                                  at [email protected]/com.sun.javafx.application.PlatformImpl.lambda$runLater$11(PlatformImpl.java:455)                                                                                          at [email protected]/com.sun.glass.ui.InvokeLaterDispatcher$Future.run(InvokeLaterDispatcher.java:95)                                                                                           at [email protected]/com.sun.glass.ui.win.WinApplication._runLoop(Native Method)                                                                                                                at [email protected]/com.sun.glass.ui.win.WinApplication.lambda$runLoop$3(WinApplication.java:184)                                                                                              ... 1 more                                                                                                                                                                                   Caused by: java.lang.IndexOutOfBoundsException                                                                                                                                                               at [email protected]/javafx.collections.transformation.FilteredList.get(FilteredList.java:169)                                                                                                      at [email protected]/com.tobiasdiez.easybind.MappedList.get(MappedList.java:31)                                                                                                 at java.base/java.util.AbstractList$Itr.next(AbstractList.java:373)                                                                                                                                  ... 15 more                                                                                                                                                                                  2024-08-07 10:53:47 [pool-2-thread-3] ai.djl.pytorch.jni.LibUtils.downloadPyTorch()                                                                                                                  INFO: Downloading https://publish.djl.ai/pytorch/2.3.1/cpu/win-x86_64/native/lib/libiompstubs5md.dll.gz ...                                                                                          <===========--> 90% EXECUTING [47s]                                                                                                                                                                  > :run                                                                                                                                                                                               ^CTerminate batch job (Y/N)? ^C                                                                                      

Then Re-Run JabRef (using gradlew run)

Result:

024-08-07 11:01:33 [JavaFX Application Thread] sun.util.logging.internal.LoggingProviderImpl$JULWrapper.log()                                                                                       WARN: Resource "" not found.                                                                                                                                                                         Loading:     100% |========================================|                                                                                                                                         2024-08-07 11:01:34 [JavaFX Application Thread] org.jabref.logic.ai.models.EmbeddingModel.lambda$startRebuildingTask$1()                                                                             ERROR: An error occurred while building the embedding model: ai.djl.engine.EngineException: Failed to load PyTorch native library                                                                            at [email protected]/ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:90)                                                                                                         at [email protected]/ai.djl.pytorch.engine.PtEngineProvider.getEngine(PtEngineProvider.java:41)                                                                                           at [email protected]/ai.djl.engine.Engine.getEngine(Engine.java:190)                                            

---

This refs #132 - a broken embedded model should be fixed with a download!

[InAnYan]

OMG so that's a Windows issue?

[InAnYan]

Okay, little bit of research led to, that the user needs to install VC++ redistributable

https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170

[InAnYan]

Wait until Windows VM is up (approx 10 Minuts)

😔

[InAnYan]

From dev call:

1. Check if VC++ should be really installed (if needed, then update docs (not blog, only link to docs))
2. Add documentation if embedding model download was interrupted in the middle and AI can't be used at all (fix: delete .djl.ai dir)

[koppor]

Windows: choco install vcredist140

Link for self-guided installation: https://aka.ms/vs/16/release/vc_redist.x64.exe

[InAnYan]

Yes, they really need VC++. So, I updated the documentation about that.

And I also added guide, if user closed JabRef in the middle of downloading the embedding model