UnsatisfiedLinkError - PyTorch 2.0 on Windows 11 #2552

enpasos · 2023-04-18T17:53:55Z

Very cool, that you have started to support PyTorch 2.0!
Here some minor bug ...

Description

Running some application with 0.22.0-SNAPSHOT (today, 18th April 2023) I experience:

Caused by: java.lang.UnsatisfiedLinkError: C:\Users\enpasos\.djl.ai\pytorch\2.0.0-cu118-win-x86_64\nvfuser_codegen.dll: Can't find dependent libraries
        at java.base/jdk.internal.loader.NativeLibraries.load(Native Method)
...
        at ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:53)

Workaround

Deleting .djl.ai\pytorch\2.0.0-cu118-win-x86_64\nvfuser_codegen.dll I only get a warning:

[W C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\torch\csrc\jit\codegen\cuda\interface.cpp:47] Warning: Loading nvfuser library failed with: error in LoadLibrary for nvfuser_codegen.dll. WinError 126: Das angegebene Modul wurde nicht gefunden.
 (function LoadingNvfuserLibrary)

How to Reproduce?

git clone https://github.com/enpasos/muzero.git
cd muzero 
gradlew build -x test
java -jar games/tictactoe/build/libs/tictactoe-0.6.0-SNAPSHOT-exec.jar

Environment Info

DJL: 0.22.0-SNAPSHOT (automatically installed with the app)
- PYTORCH: 2.0.0
Java: Corretto-17.0.6 (needs to be installed)
CUDA (needs to be installed)
- cudnn: 8.9
- CUDA SDK: 11.8
- GPU Driver: 517.89
OS: Microsoft Windows 11
Hardware
- GPU: NVIDIA GeForce RTX 4090
- CPU: Intel Core i9-13900K
- RAM: 128 GB

The text was updated successfully, but these errors were encountered:

frankfliu · 2023-04-18T17:56:39Z

PyTorch 2.0.0 CUDA 118 has multiple issues. We are wait for 2.0.1 release.

farzad-845 · 2023-05-02T12:08:20Z

@frankfliu The same error occurs in my workstation, how can I downgrade to the previous version, and which version works well with Windows 11?

enpasos · 2023-05-02T12:28:00Z

@farzad-845 I am happy with PyTorch 2.0.0 and CUDA SDK 11.8 and use the workaround to simply delete
C:\Users\myuser.djl.ai\pytorch\2.0.0-cu118-win-x86_64\nvfuser_codegen.dll.
Previously I was working with

pytorch = "1.13.1" with djl-pytorch-native-cu117 = { module = "ai.djl.pytorch:pytorch-native-cu117", version.ref = "pytorch" }

Hope this helps.

farzad-845 · 2023-05-02T13:52:37Z

@enpasos Thank you, But I delete this file and get a warning, after the warning get another error that tells Can't load this .dll.

C:\Users\User\.djl.ai\pytorch\2.0.0-cu118-win-x86_64\a.out: Can't load this .dll (machine code=0x4810) on a AMD 64-bit platform

The Full Error:

Exception in thread "main" ai.djl.engine.EngineException: Failed to load PyTorch native library
	at ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:90)
	at ai.djl.pytorch.engine.PtEngineProvider.getEngine(PtEngineProvider.java:40)
	at ai.djl.engine.Engine.getEngine(Engine.java:187)
	at ai.djl.Model.newInstance(Model.java:99)
	at ai.djl.repository.zoo.BaseModelLoader.createModel(BaseModelLoader.java:191)
	at ai.djl.repository.zoo.BaseModelLoader.loadModel(BaseModelLoader.java:154)
	at ai.djl.repository.zoo.Criteria.loadModel(Criteria.java:172)
	at ai.djl.repository.zoo.ModelZoo.loadModel(ModelZoo.java:141)
Caused by: java.lang.UnsatisfiedLinkError: C:\Users\User\.djl.ai\pytorch\2.0.0-cu118-win-x86_64\a.out: Can't load this .dll (machine code=0x4810) on a AMD 64-bit platform
	at java.base/jdk.internal.loader.NativeLibraries.load(Native Method)
	at java.base/jdk.internal.loader.NativeLibraries$NativeLibraryImpl.open(NativeLibraries.java:388)
	at java.base/jdk.internal.loader.NativeLibraries.loadLibrary(NativeLibraries.java:232)
	at java.base/jdk.internal.loader.NativeLibraries.loadLibrary(NativeLibraries.java:174)
	at java.base/java.lang.ClassLoader.loadLibrary(ClassLoader.java:2389)
	at java.base/java.lang.Runtime.load0(Runtime.java:755)
	at java.base/java.lang.System.load(System.java:1953)
	at ai.djl.pytorch.jni.LibUtils.loadNativeLibrary(LibUtils.java:347)
	at java.base/java.util.stream.ForEachOps$ForEachOp$OfRef.accept(ForEachOps.java:183)
	at java.base/java.util.stream.ReferencePipeline$3$1.accept(ReferencePipeline.java:197)
	at java.base/java.util.stream.ReferencePipeline$2$1.accept(ReferencePipeline.java:179)
	at java.base/java.util.stream.ReferencePipeline$3$1.accept(ReferencePipeline.java:197)
	at java.base/java.util.Iterator.forEachRemaining(Iterator.java:133)
	at java.base/java.util.Spliterators$IteratorSpliterator.forEachRemaining(Spliterators.java:1845)
	at java.base/java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:509)
	at java.base/java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:499)
	at java.base/java.util.stream.ForEachOps$ForEachOp.evaluateSequential(ForEachOps.java:150)
	at java.base/java.util.stream.ForEachOps$ForEachOp$OfRef.evaluateSequential(ForEachOps.java:173)
	at java.base/java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:234)
	at java.base/java.util.stream.ReferencePipeline.forEach(ReferencePipeline.java:596)
	at ai.djl.pytorch.jni.LibUtils.loadLibTorch(LibUtils.java:146)
	at ai.djl.pytorch.jni.LibUtils.loadLibrary(LibUtils.java:78)
	at ai.djl.pytorch.engine.PtEngine.newInstance(PtEngine.java:53)
	... 10 more

This is my pom file:

        <!-- https://mvnrepository.com/artifact/ai.djl/api -->
        <dependency>
            <groupId>ai.djl</groupId>
            <artifactId>api</artifactId>
            <version>0.22.1</version>
        </dependency>

        <!-- https://mvnrepository.com/artifact/ai.djl.huggingface/tokenizers -->
        <dependency>
            <groupId>ai.djl.huggingface</groupId>
            <artifactId>tokenizers</artifactId>
            <version>0.22.1</version>
        </dependency>


        <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-model-zoo</artifactId>
            <version>0.22.1</version>
        </dependency>

        <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-native-cu118</artifactId>
            <version>2.0.0</version>
        </dependency>
        <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-jni</artifactId>
            <version>2.0.0-0.22.1</version>
            <scope>runtime</scope>
        </dependency>


        <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-engine</artifactId>
            <version>0.22.1</version>
            <scope>runtime</scope>
        </dependency>

enpasos · 2023-05-22T15:56:07Z

@farzad-845, sorry for the late reply. I do not know what is causing your problem. However, version 2.0.1 of PyTorch has been released. According to @frankfliu's comment, I hope the problems we are experiencing may be fixed once the new version is integrated into DJL.

enpasos · 2023-07-11T17:11:48Z

Just tested with version 2.0.1.
I see the same UnsatisfiedLinkError as in 2.0.0.
Runs for me with the same workaround as in 2.0.0.

eleven-dimension · 2023-09-11T03:01:12Z

@farzad-845 I am happy with PyTorch 2.0.0 and CUDA SDK 11.8 and use the workaround to simply delete C:\Users\myuser.djl.ai\pytorch\2.0.0-cu118-win-x86_64\nvfuser_codegen.dll. Previously I was working with
pytorch = "1.13.1" with djl-pytorch-native-cu117 = { module = "ai.djl.pytorch:pytorch-native-cu117", version.ref = "pytorch" }
Hope this helps.

Same error on windows 10. Strange as it may seem, this method works. Thank you! @enpasos

enpasos · 2023-10-20T07:22:31Z

Maybe we get rid of this UnsatisfiedLinkError with PyTorch 2.1.0 released October 04, 2023.

https://github.com/pytorch/pytorch/releases/tag/v2.1.0

frankfliu · 2023-11-22T23:04:37Z

#2868

enpasos · 2023-11-27T11:01:53Z

Thx, a lot for the fix and for supporting PyTorch 2.1.1!
It works nicely on my stack

https://github.com/enpasos/muzero  with  
djl = "0.26.0-SNAPSHOT"
pytorch = "2.1.1-SNAPSHOT"
on
cudnn-windows-x86_64-8.9.6.50_cuda12
cuda_12.1.1_531.14
Windows 11 Pro, 23H2, 22631.2715

enpasos added the bug Something isn't working label Apr 18, 2023

enpasos closed this as completed Nov 27, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

UnsatisfiedLinkError - PyTorch 2.0 on Windows 11 #2552

UnsatisfiedLinkError - PyTorch 2.0 on Windows 11 #2552

enpasos commented Apr 18, 2023

frankfliu commented Apr 18, 2023

farzad-845 commented May 2, 2023

enpasos commented May 2, 2023

farzad-845 commented May 2, 2023 •

edited

enpasos commented May 22, 2023

enpasos commented Jul 11, 2023

eleven-dimension commented Sep 11, 2023 •

edited

enpasos commented Oct 20, 2023

frankfliu commented Nov 22, 2023

enpasos commented Nov 27, 2023

UnsatisfiedLinkError - PyTorch 2.0 on Windows 11 #2552

UnsatisfiedLinkError - PyTorch 2.0 on Windows 11 #2552

Comments

enpasos commented Apr 18, 2023

Description

Workaround

How to Reproduce?

Environment Info

frankfliu commented Apr 18, 2023

farzad-845 commented May 2, 2023

enpasos commented May 2, 2023

farzad-845 commented May 2, 2023 • edited

enpasos commented May 22, 2023

enpasos commented Jul 11, 2023

eleven-dimension commented Sep 11, 2023 • edited

enpasos commented Oct 20, 2023

frankfliu commented Nov 22, 2023

enpasos commented Nov 27, 2023

farzad-845 commented May 2, 2023 •

edited

eleven-dimension commented Sep 11, 2023 •

edited