Java tutorial
/* * Copyright (c) 2018 by Andrew Charneski. * * The author 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 com.simiacryptus.mindseye.layers.cudnn; import com.google.gson.JsonObject; import com.simiacryptus.mindseye.lang.*; import com.simiacryptus.mindseye.lang.cudnn.*; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import javax.annotation.Nonnull; import javax.annotation.Nullable; import java.util.Arrays; import java.util.List; import java.util.Map; import java.util.UUID; import java.util.stream.Stream; /** * Reduces the resolution of the input by selecting a centered window. The output png will have the same number of * color bands. */ @SuppressWarnings("serial") public class ImgCropLayer extends LayerBase implements MultiPrecision<ImgCropLayer> { private static final Logger log = LoggerFactory.getLogger(ImgCropLayer.class); private int sizeX; private int sizeY; private Precision precision = Precision.Double; /** * Instantiates a new Img eval key. */ private ImgCropLayer() { } /** * Instantiates a new Img crop key. * * @param sizeX the size x * @param sizeY the size y */ public ImgCropLayer(int sizeX, int sizeY) { this.sizeX = sizeX; this.sizeY = sizeY; assert 0 < sizeX; assert 0 < sizeY; } /** * Instantiates a new Img eval key. * * @param json the json * @param rs the rs */ protected ImgCropLayer(@Nonnull final JsonObject json, Map<CharSequence, byte[]> rs) { super(json); sizeX = json.get("sizeX").getAsInt(); sizeY = json.get("sizeY").getAsInt(); this.precision = Precision.valueOf(json.getAsJsonPrimitive("precision").getAsString()); assert 0 < sizeX; assert 0 < sizeY; } /** * From json img eval key. * * @param json the json * @param rs the rs * @return the img eval key */ public static ImgCropLayer fromJson(@Nonnull final JsonObject json, Map<CharSequence, byte[]> rs) { return new ImgCropLayer(json, rs); } /** * Copy cuda tensor. * * @param gpu the gpu * @param input the input tensor * @param length the length * @param inputDimensions the input dimensions * @param outputDimensions the output dimensions * @param dirty the dirty * @param precision the precision * @return the cuda tensor */ public static CudaTensor copy(final CudnnHandle gpu, final CudaTensor input, final int length, final int[] inputDimensions, final int[] outputDimensions, final boolean dirty, Precision precision) { if (3 != inputDimensions.length) throw new IllegalArgumentException("inputDimensions.length"); if (3 != outputDimensions.length) throw new IllegalArgumentException("dimOut.length"); if (inputDimensions[2] != outputDimensions[2]) { throw new IllegalArgumentException(String.format("%d != %d", inputDimensions[2], outputDimensions[2])); } //log.info(String.format("offset=%d,%d", offsetX, offsetY)); @Nonnull final int[] viewDim = getViewDimensions(inputDimensions, outputDimensions); int sourceOffset = 0; int destinationOffset = 0; if (inputDimensions[0] < outputDimensions[0]) { destinationOffset += (outputDimensions[0] - inputDimensions[0]) / 2; } else { sourceOffset += (inputDimensions[0] - outputDimensions[0]) / 2; } if (inputDimensions[1] < outputDimensions[1]) { destinationOffset += outputDimensions[0] * ((outputDimensions[1] - inputDimensions[1]) / 2); } else { sourceOffset += input.descriptor.hStride * ((inputDimensions[1] - outputDimensions[1]) / 2); } assert sourceOffset >= 0; assert destinationOffset >= 0; assert sourceOffset + Tensor.length(viewDim) <= Tensor.length(inputDimensions); assert destinationOffset + Tensor.length(viewDim) <= Tensor.length(outputDimensions); @Nonnull final CudaDevice.CudaTensorDescriptor sourceViewDescriptor = gpu.newTensorDescriptor(precision, // length, // viewDim[2], // viewDim[1], // viewDim[0], // input.descriptor.nStride, // input.descriptor.cStride, // input.descriptor.hStride, // input.descriptor.wStride); CudaMemory inputTensorMemory = input.getMemory(gpu); try { if (Arrays.equals(viewDim, outputDimensions)) { assert sourceOffset >= 0; assert destinationOffset == 0; return CudaTensor.wrap(inputTensorMemory.withByteOffset(sourceOffset * precision.size), sourceViewDescriptor, precision); } @Nonnull final CudaDevice.CudaTensorDescriptor destinationViewDescriptor = gpu.newTensorDescriptor(precision, // length, // viewDim[2], // viewDim[1], // viewDim[0], // outputDimensions[2] * outputDimensions[1] * outputDimensions[0], // outputDimensions[1] * outputDimensions[0], // outputDimensions[0], // 1); @Nonnull final CudaMemory outputBuffer = gpu.allocate((long) length * outputDimensions[2] * outputDimensions[1] * outputDimensions[0] * precision.size, MemoryType.Managed.normalize(), dirty); CudaSystem.handle(gpu.cudnnTransformTensor(precision.getPointer(1.0), sourceViewDescriptor.getPtr(), inputTensorMemory.getPtr().withByteOffset(sourceOffset * precision.size), precision.getPointer(0.0), destinationViewDescriptor.getPtr(), outputBuffer.getPtr().withByteOffset(destinationOffset * precision.size))); assert CudaDevice.isThreadDeviceId(gpu.getDeviceId()); inputTensorMemory.dirty(); outputBuffer.dirty(); Stream.<ReferenceCounting>of(sourceViewDescriptor, destinationViewDescriptor) .forEach(ReferenceCounting::freeRef); CudaDevice.CudaTensorDescriptor descriptorCudaResource = gpu.newTensorDescriptor(precision, // length, // outputDimensions[2], // outputDimensions[1], // outputDimensions[0], // outputDimensions[2] * outputDimensions[1] * outputDimensions[0], // outputDimensions[1] * outputDimensions[0], // outputDimensions[0], // 1); return CudaTensor.wrap(outputBuffer, descriptorCudaResource, precision); } finally { inputTensorMemory.freeRef(); } } /** * Get view dimensions int [ ]. * * @param sourceDimensions the source dimensions * @param destinationDimensions the destination dimensions * @return the int [ ] */ @Nonnull public static int[] getViewDimensions(int[] sourceDimensions, int[] destinationDimensions) { @Nonnull final int[] viewDim = new int[3]; Arrays.parallelSetAll(viewDim, i -> Math.min(sourceDimensions[i], destinationDimensions[i])); return viewDim; } /** * Gets compatibility key. * * @return the compatibility key */ @Nonnull public Layer getCompatibilityLayer() { return this.as(com.simiacryptus.mindseye.layers.java.ImgCropLayer.class); } @Nullable @Override public Result evalAndFree(@Nonnull final Result... inObj) { if (!CudaSystem.isEnabled()) return getCompatibilityLayer().evalAndFree(inObj); assert 1 == inObj.length; final Result input = inObj[0]; final TensorList inputData = input.getData(); assert 3 == inputData.getDimensions().length; final int length = inputData.length(); @Nonnull int[] dimIn = inputData.getDimensions(); if (dimIn[0] == sizeX && dimIn[1] == sizeY) { return input; } @Nonnull final int[] dimOut = Arrays.copyOf(dimIn, 3); dimOut[0] = sizeX; dimOut[1] = sizeY; final TensorList outputData = CudaSystem.run(gpu -> { @Nullable final CudaTensor inputTensor = gpu.getTensor(inputData, precision, MemoryType.Device, false); inputData.freeRef(); boolean dirty = dimOut[0] <= dimIn[0] && dimOut[1] <= dimIn[1]; assert dimOut[0] > 0; assert dimOut[1] > 0; assert dimOut[2] > 0; CudaTensor cudaTensor = copy(gpu, inputTensor, length, dimIn, dimOut, dirty, precision); Stream.<ReferenceCounting>of(inputTensor).forEach(ReferenceCounting::freeRef); return CudaTensorList.wrap(cudaTensor, length, dimOut, precision); }, inputData); int[] output_dimensions = outputData.getDimensions(); int output_length = outputData.length(); return new Result(outputData, (@Nonnull final DeltaSet<UUID> buffer, @Nonnull final TensorList delta) -> { if (!Arrays.equals(delta.getDimensions(), output_dimensions)) { throw new AssertionError( Arrays.toString(delta.getDimensions()) + " != " + Arrays.toString(output_dimensions)); } if (delta.length() != output_length) { throw new AssertionError(delta.length() + " != " + output_length); } assert delta.length() == length; if (input.isAlive()) { final TensorList passbackTensorList = CudaSystem.run(gpu -> { @Nullable final CudaTensor errorPtr = gpu.getTensor(delta, precision, MemoryType.Device, false); delta.freeRef(); boolean dirty = dimOut[0] >= dimIn[0] && dimOut[1] >= dimIn[1]; CudaTensor cudaTensor = copy(gpu, errorPtr, length, dimOut, dimIn, dirty, precision); Stream.<ReferenceCounting>of(errorPtr).forEach(ReferenceCounting::freeRef); return CudaTensorList.wrap(cudaTensor, length, dimIn, precision); }, delta); input.accumulate(buffer, passbackTensorList); } else { delta.freeRef(); } }) { @Override public void accumulate(final DeltaSet<UUID> buffer, final TensorList delta) { getAccumulator().accept(buffer, delta); } @Override protected void _free() { Arrays.stream(inObj).forEach(nnResult -> nnResult.freeRef()); } @Override public boolean isAlive() { return Arrays.stream(inObj).anyMatch(x -> x.isAlive()); } }; } @Nonnull @Override public JsonObject getJson(Map<CharSequence, byte[]> resources, DataSerializer dataSerializer) { @Nonnull final JsonObject json = super.getJsonStub(); json.addProperty("sizeY", sizeY); json.addProperty("sizeX", sizeX); json.addProperty("precision", precision.name()); return json; } @Nonnull @Override public List<double[]> state() { return Arrays.asList(); } @Override public Precision getPrecision() { return precision; } @Nonnull @Override public ImgCropLayer setPrecision(final Precision precision) { this.precision = precision; return this; } }