Windows10下用cuDNN运行DL4J时ND4J压缩器加载失败求助

安全阿岳

2026-7-28

问题

在Windows10系统下使用NVIDIA cuDNN库运行DL4J，配置了对应的pom依赖，并使用maven-assembly-plugin打包成Uber-Jar，运行时出现压缩器加载失败的错误，提示No compressors were found，完整报错信息及配置如下：

POM依赖配置

<properties>
    <java.version>16</java.version>
    <dl4j-master.version>1.0.0-M2.1</dl4j-master.version>
    <nd4j.backend>nd4j-cuda-11.6</nd4j.backend>
    <jcommon.version>1.0.24</jcommon.version>
    <jfreechart.version>1.0.13</jfreechart.version>
    <javacv.version>1.5.5</javacv.version>
    <logback.version>1.2.6</logback.version>
    <maven.compiler.source>16</maven.compiler.source>
    <maven.compiler.target>16</maven.compiler.target>
    <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
</properties>

<dependencies>  
    <dependency>
        <groupId>org.nd4j</groupId>
        <artifactId>${nd4j.backend}</artifactId>
        <version>${dl4j-master.version}</version>
    </dependency>
    <dependency>
        <groupId>org.nd4j</groupId>
        <artifactId>${nd4j.backend}</artifactId>
        <version>${dl4j-master.version}</version>
        <classifier>windows-x86_64-cudnn</classifier>
    </dependency>
    <dependency>
        <groupId>org.bytedeco</groupId>
        <artifactId>cuda-platform-redist</artifactId>
        <version>11.6-8.3-1.5.7</version>
    </dependency>

    <dependency>
        <groupId>org.datavec</groupId>
        <artifactId>datavec-api</artifactId>
        <version>${dl4j-master.version}</version>
    </dependency>

    <dependency>
        <groupId>org.deeplearning4j</groupId>
        <artifactId>deeplearning4j-core</artifactId>
        <version>${dl4j-master.version}</version>
    </dependency>
...

打包配置

<build>
    <plugins>
        <plugin>
            <groupId>org.apache.maven.plugins</groupId>
            <artifactId>maven-assembly-plugin</artifactId>
            <version>3.4.2</version>
            <executions>
                <execution>
                    <phase>package</phase>
                    <goals>
                        <goal>single</goal>
                    </goals>
                </execution>
            </executions>
            <configuration>
                <archive>
                    <manifest>
                        <mainClass>com.somename.Launcher</mainClass>
                    </manifest>
                </archive>
                <descriptorRefs>
                    <descriptorRef>jar-with-dependencies</descriptorRef>
                </descriptorRefs>
            </configuration>
        </plugin>
    </plugins>
</build>

报错信息

java.lang.RuntimeException: Error loading ND4J Compressors via service loader: No compressors were found. This usually occurs when running ND4J UI from an uber-jar, which was built incorrectly (without services resource files being included)

完整日志：

10:39:06.143 [main] INFO org.nd4j.linalg.factory.Nd4jBackend - Loaded [JCublasBackend] backend
10:39:15.316 [main] INFO org.nd4j.nativeblas.NativeOpsHolder - Number of threads used for linear algebra: 32
10:39:15.363 [main] INFO org.nd4j.linalg.api.ops.executioner.DefaultOpExecutioner - Backend used: [CUDA]; OS: [Windows Server 2019]
10:39:15.363 [main] INFO org.nd4j.linalg.api.ops.executioner.DefaultOpExecutioner - Cores: [8]; Memory: [4,0GB];
10:39:15.364 [main] INFO org.nd4j.linalg.api.ops.executioner.DefaultOpExecutioner - Blas vendor: [CUBLAS]
10:39:15.390 [main] INFO org.nd4j.linalg.jcublas.JCublasBackend - ND4J CUDA build version: 11.6.55
10:39:15.391 [main] INFO org.nd4j.linalg.jcublas.JCublasBackend - CUDA device 0: [Tesla T4]; cc: [7.5]; Total memory: [16105865216]
10:39:15.391 [main] INFO org.nd4j.linalg.jcublas.JCublasBackend - Backend build information:
 MSVC: 192930146
STD version: 201402L
DEFAULT_ENGINE: samediff::ENGINE_CUDA
HAVE_FLATBUFFERS
HAVE_CUDNN
10:39:32.606 [main] ERROR org.nd4j.linalg.compression.BasicNDArrayCompressor - Error loading ND4J Compressors via service loader: No compressors were found. This usually occurs when running ND4J UI from an uber-jar, which was built incorrectly (without services resource files being included)
Exception in thread "main" java.lang.ExceptionInInitializerError
        at org.nd4j.linalg.factory.Nd4j.getCompressor(Nd4j.java:5288)
        at org.nd4j.linalg.api.ndarray.BaseNDArray.getDouble(BaseNDArray.java:1785)
        at org.nd4j.linalg.api.ndarray.BaseNDArray.getInt(BaseNDArray.java:1753)
        at org.deeplearning4j.datasets.datavec.RecordReaderMultiDataSetIterator.convertWritablesBatched(RecordReaderMultiDataSetIterator.java:413)
        at org.deeplearning4j.datasets.datavec.RecordReaderMultiDataSetIterator.convertFeaturesOrLabels(RecordReaderMultiDataSetIterator.java:353)
        at org.deeplearning4j.datasets.datavec.RecordReaderMultiDataSetIterator.nextMultiDataSet(RecordReaderMultiDataSetIterator.java:326)
        at org.deeplearning4j.datasets.datavec.RecordReaderMultiDataSetIterator.next(RecordReaderMultiDataSetIterator.java:206)
        at org.deeplearning4j.datasets.datavec.RecordReaderDataSetIterator.next(RecordReaderDataSetIterator.java:346)
        at org.deeplearning4j.datasets.datavec.RecordReaderDataSetIterator.next(RecordReaderDataSetIterator.java:421)
        at org.deeplearning4j.datasets.datavec.RecordReaderDataSetIterator.next(RecordReaderDataSetIterator.java:53)
        at com.packagename.service.FAClassifierLearning.main(FAClassifierLearning.java:69)
Caused by: java.lang.RuntimeException: Error loading ND4J Compressors via service loader: No compressors were found. This usually occurs when running ND4J UI from an uber-jar, which was built incorrectly (without services resource files being included)
        at org.nd4j.linalg.compression.BasicNDArrayCompressor.loadCompressors(BasicNDArrayCompressor.java:66)
        at org.nd4j.linalg.compression.BasicNDArrayCompressor.<init>(BasicNDArrayCompressor.java:46)
        at org.nd4j.linalg.compression.BasicNDArrayCompressor.<clinit>(BasicNDArrayCompressor.java:39)
        ... 11 more

解决方法

问题根源是Uber-Jar打包时，多个依赖中的META-INF/services文件被覆盖，ND4J压缩器通过Java ServiceLoader机制加载，这些服务文件丢失导致无法找到压缩器。以下是两种解决方案：

方案1：改用maven-shade-plugin打包

maven-shade-plugin默认支持合并META-INF/services文件，替换原有的assembly-plugin配置：

<build>
    <plugins>
        <plugin>
            <groupId>org.apache.maven.plugins</groupId>
            <artifactId>maven-shade-plugin</artifactId>
            <version>3.4.1</version>
            <executions>
                <execution>
                    <phase>package</phase>
                    <goals>
                        <goal>shade</goal>
                    </goals>
                    <configuration>
                        <transformers>
                            <!-- 合并所有META-INF/services下的文件 -->
                            <transformer implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer"/>
                            <!-- 指定主类 -->
                            <transformer implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer">
                                <mainClass>com.somename.Launcher</mainClass>
                            </transformer>
                        </transformers>
                    </configuration>
                </execution>
            </executions>
        </plugin>
    </plugins>
</build>

方案2：修改assembly-plugin配置，添加服务文件合并规则

如果坚持使用assembly-plugin，需要自定义assembly描述文件来合并服务文件：

在项目根目录创建src/main/assembly/uberjar.xml，内容如下：

<assembly xmlns="http://maven.apache.org/ASSEMBLY/2.1.0"
          xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
          xsi:schemaLocation="http://maven.apache.org/ASSEMBLY/2.1.0 http://maven.apache.org/xsd/assembly-2.1.0.xsd">
    <id>jar-with-dependencies</id>
    <formats>
        <format>jar</format>
    </formats>
    <includeBaseDirectory>false</includeBaseDirectory>
    <dependencySets>
        <dependencySet>
            <outputDirectory>/</outputDirectory>
            <useProjectArtifact>true</useProjectArtifact>
            <unpack>true</unpack>
            <scope>runtime</scope>
        </dependencySet>
    </dependencySets>
    <containerDescriptorHandlers>
        <!-- 合并META-INF/services目录下的所有文件 -->
        <containerDescriptorHandler implementation="org.apache.maven.plugins.assembly.handlers.service.ServiceHandler"/>
    </containerDescriptorHandlers>
    <archive>
        <manifest>
            <mainClass>com.somename.Launcher</mainClass>
        </manifest>
    </archive>
</assembly>

修改pom中的assembly-plugin配置，引用自定义描述文件：

<build>
    <plugins>
        <plugin>
            <groupId>org.apache.maven.plugins</groupId>
            <artifactId>maven-assembly-plugin</artifactId>
            <version>3.4.2</version>
            <executions>
                <execution>
                    <phase>package</phase>
                    <goals>
                        <goal>single</goal>
                    </goals>
                </execution>
            </executions>
            <configuration>
                <descriptors>
                    <descriptor>src/main/assembly/uberjar.xml</descriptor>
                </descriptors>
            </configuration>
        </plugin>
    </plugins>
</build>

额外说明

你的pom依赖中nd4j-cuda-11.6已经包含了默认的ND4J压缩器实现，无需额外添加依赖，只要打包时保证META-INF/services文件不丢失，就能正常加载压缩器。

内容的提问来源于stack exchange，提问作者apollox

Windows10下用cuDNN运行DL4J时ND4J压缩器加载失败求助

问题 #

POM依赖配置 #

打包配置 #

报错信息 #

解决方法 #

方案1：改用maven-shade-plugin打包 #

方案2：修改assembly-plugin配置，添加服务文件合并规则 #

额外说明 #

问题