krickert commented on code in PR #1462: URL: https://github.com/apache/opennlp/pull/1462#discussion_r4243896811
########## opennlp-core/opennlp-ml/opennlp-dl/src/test/java/opennlp/dl/vectors/TinyVectorModels.java: ########## @@ -0,0 +1,147 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package opennlp.dl.vectors; + +import java.io.File; +import java.io.IOException; +import java.io.InputStream; +import java.nio.file.Files; +import java.nio.file.Path; +import java.nio.file.StandardCopyOption; +import java.util.List; +import java.util.Objects; + +import ai.onnxruntime.OrtEnvironment; +import ai.onnxruntime.OrtException; +import ai.onnxruntime.OrtProvider; + +import opennlp.dl.InferenceOptions; + +import static org.junit.jupiter.api.Assumptions.abort; +import static org.junit.jupiter.api.Assumptions.assumeTrue; + +/** + * Test fixtures for the bundled tiny ONNX models. {@code tiny-vectors.onnx} (see + * {@code gen_tiny_vectors_model.py}) computes {@code output[b][t] = float(input_ids[b][t]) * W} + * with {@code W = [0.5, -1, 2]}; {@code tiny-pooled.onnx} (see {@code gen_tiny_pooled_model.py}) + * adds a {@code sentence_embedding} output holding the sum of the ids times {@code W}. + */ +public final class TinyVectorModels { + + /** The model with a token output only. */ + public static final String TINY_VECTORS = "tiny-vectors.onnx"; + + /** The model with an additional {@code sentence_embedding} output. */ + public static final String TINY_POOLED = "tiny-pooled.onnx"; + + private static final String VECTORS_RESOURCES = "/opennlp/dl/vectors/"; + + private TinyVectorModels() { + } + + /** + * Copies a classpath resource into a directory. The resource is copied rather than resolved in + * place, because inside the {@code opennlp-dl} test-jar, as used by {@code opennlp-dl-gpu}, it + * is not a file. + * + * @param dir The directory to copy into. + * @param resource The absolute classpath name of the resource. + * @param name The file name of the copy. + * @return The copied file. + * @throws IOException Thrown if the resource cannot be copied. + */ + public static File copy(final Path dir, final String resource, final String name) + throws IOException { + final Path file = dir.resolve(name); + try (InputStream is = Objects.requireNonNull( + TinyVectorModels.class.getResourceAsStream(resource), resource)) { + Files.copy(is, file, StandardCopyOption.REPLACE_EXISTING); + } + return file.toFile(); + } + + /** + * Copies a tiny model into a directory. + * + * @param dir The directory to copy into. + * @param name {@link #TINY_VECTORS} or {@link #TINY_POOLED}. + * @return The model file. + * @throws IOException Thrown if the model cannot be copied. + */ + public static File model(final Path dir, final String name) throws IOException { + return copy(dir, VECTORS_RESOURCES + name, name); + } + + /** + * Writes the vocabulary of the tiny models, where the line number is the id: + * {@code [UNK]=2, [SEP]=3, hello=4, world=5, [CLS]=7}. + * + * @param dir The directory to write into. + * @return The vocabulary file. + * @throws IOException Thrown if the vocabulary cannot be written. + */ + public static File vocab(final Path dir) throws IOException { + final Path file = dir.resolve("vocab.txt"); + Files.write(file, List.of("[PAD]", "unused1", "[UNK]", "[SEP]", "hello", "world", + "unused2", "[CLS]")); + return file.toFile(); + } + + /** + * {@return {@code W} times {@code factor}, the expected vector for a mean or summed id} + * + * @param factor The mean or summed token id. + */ + public static float[] scale(final float factor) { + return new float[] {0.5f * factor, -1f * factor, 2f * factor}; + } + + /** + * {@return inference options that request the GPU with the given device id} + * + * @param deviceId The CUDA device id. + */ + public static InferenceOptions gpu(final int deviceId) { + final InferenceOptions options = new InferenceOptions(); + options.setGpu(true); + options.setGpuDeviceId(deviceId); + return options; + } + + /** + * Opens a session on CUDA device {@code 0}, or aborts the calling test where the ONNX Runtime + * on the classpath has no CUDA execution provider or the device is not usable. + * + * @param model The model file. + * @param vocab The vocabulary file. + * @param pooling The pooling. + * @param normalize {@code true} to scale every vector to unit length. + * @return The session on the GPU. + * @throws IOException Thrown if the model or vocabulary cannot be read. + */ + public static SentenceVectorsDL gpuOrAbort(final File model, final File vocab, + final Pooling pooling, final boolean normalize) throws IOException { + assumeTrue(OrtEnvironment.getAvailableProviders().contains(OrtProvider.CUDA), Review Comment: Done, `OnnxTestSupport.hasCudaProvider()` is used in both places. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
