Examples with IntIntVector com.tencent.angel.ml.math2.vector.IntIntVector used on opensource projects

Example 46 with IntIntVector

use of com.tencent.angel.ml.math2.vector.IntIntVector in project angel by Tencent.

the class PSManagerTest method testPSError.

@Test
public void testPSError() throws Exception {
    try {
        int heartbeatInterval = LocalClusterContext.get().getConf().getInt(AngelConf.ANGEL_PS_HEARTBEAT_INTERVAL_MS, AngelConf.DEFAULT_ANGEL_PS_HEARTBEAT_INTERVAL_MS);
        AngelApplicationMaster angelAppMaster = LocalClusterContext.get().getMaster().getAppMaster();
        ParameterServerManager psManager = angelAppMaster.getAppContext().getParameterServerManager();
        AMParameterServer amPs = psManager.getParameterServer(psId);
        PSAttempt psAttempt0 = amPs.getPSAttempt(psAttempt0Id);
        ParameterServer ps = LocalClusterContext.get().getPS(psAttempt0Id).getPS();
        Worker worker = LocalClusterContext.get().getWorker(worker0Attempt0Id).getWorker();
        int w1Id = angelAppMaster.getAppContext().getMatrixMetaManager().getMatrix("w1").getId();
        int w2Id = angelAppMaster.getAppContext().getMatrixMetaManager().getMatrix("w2").getId();
        Location masterLoc = LocalClusterContext.get().getMaster().getAppMaster().getAppContext().getMasterService().getLocation();
        TConnection connection = TConnectionManager.getConnection(ps.getConf());
        MasterProtocol master = connection.getMasterService(masterLoc.getIp(), masterLoc.getPort());
        int task0Iteration = 2;
        int task1Iteration = 1;
        int task0w1Clock = 10;
        int task0w2Clock = 20;
        int task1w1Clock = 9;
        int task1w2Clock = 19;
        int w1Clock = (task0w1Clock < task1w1Clock) ? task0w1Clock : task1w1Clock;
        int w2Clock = (task0w2Clock < task1w2Clock) ? task0w2Clock : task1w2Clock;
        TaskContext task0Context = worker.getTaskManager().getRunningTask().get(task0Id).getTaskContext().getContext();
        TaskContext task1Context = worker.getTaskManager().getRunningTask().get(task1Id).getTaskContext().getContext();
        master.taskIteration(null, TaskIterationRequest.newBuilder().setIteration(task0Iteration).setTaskId(ProtobufUtil.convertToIdProto(task0Id)).build());
        master.taskIteration(null, TaskIterationRequest.newBuilder().setIteration(task1Iteration).setTaskId(ProtobufUtil.convertToIdProto(task1Id)).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task0Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w1Id).setClock(task0w1Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task0Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w2Id).setClock(task0w2Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task1Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w1Id).setClock(task1w1Clock).build()).build());
        master.taskClock(null, TaskClockRequest.newBuilder().setTaskId(ProtobufUtil.convertToIdProto(task1Id)).setMatrixClock(MatrixClock.newBuilder().setMatrixId(w2Id).setClock(task1w2Clock).build()).build());
        assertEquals(amPs.getMaxAttempts(), 4);
        PSAttemptId psAttempt1Id = new PSAttemptId(psId, 1);
        PSAttemptId psAttempt2Id = new PSAttemptId(psId, 2);
        PSAttemptId psAttempt3Id = new PSAttemptId(psId, 3);
        // attempt 0
        ps.stop(-1);
        PSErrorRequest request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt0Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        PSAttempt psAttempt1 = amPs.getPSAttempt(psAttempt1Id);
        assertTrue(psAttempt1 != null);
        assertEquals(psAttempt0.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(psAttempt1.getInternalState(), PSAttemptStateInternal.RUNNING);
        assertEquals(amPs.getState(), AMParameterServerState.RUNNING);
        assertEquals(amPs.getNextAttemptNumber(), 2);
        assertEquals(amPs.getRunningAttemptId(), psAttempt1Id);
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 2);
        List<String> diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 1);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        ps = LocalClusterContext.get().getPS(psAttempt1Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock, w2Clock);
        MatrixClient w1Task0Client = worker.getPSAgent().getMatrixClient("w1", 0);
        MatrixClient w1Task1Client = worker.getPSAgent().getMatrixClient("w1", 1);
        int matrixW1Id = w1Task0Client.getMatrixId();
        int[] delta = new int[100000];
        for (int i = 0; i < 100000; i++) {
            delta[i] = 2;
        }
        IntIntVector deltaVec = new IntIntVector(100000, new IntIntDenseVectorStorage(delta));
        deltaVec.setMatrixId(matrixW1Id);
        deltaVec.setRowId(0);
        w1Task0Client.increment(deltaVec);
        deltaVec = new IntIntVector(100000, new IntIntDenseVectorStorage(delta));
        deltaVec.setMatrixId(matrixW1Id);
        deltaVec.setRowId(0);
        w1Task1Client.increment(deltaVec);
        w1Task0Client.clock().get();
        w1Task1Client.clock().get();
        ps = LocalClusterContext.get().getPS(psAttempt1Id).getPS();
        int snapshotInterval = LocalClusterContext.get().getConf().getInt(AngelConf.ANGEL_PS_BACKUP_INTERVAL_MS, AngelConf.DEFAULT_ANGEL_PS_BACKUP_INTERVAL_MS);
        Thread.sleep(snapshotInterval * 2);
        // attempt1
        ps.stop(-1);
        request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt1Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        PSAttempt psAttempt2 = amPs.getPSAttempt(psAttempt2Id);
        assertTrue(psAttempt2 != null);
        assertEquals(psAttempt1.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(psAttempt2.getInternalState(), PSAttemptStateInternal.RUNNING);
        assertEquals(amPs.getState(), AMParameterServerState.RUNNING);
        assertEquals(amPs.getNextAttemptNumber(), 3);
        assertEquals(amPs.getRunningAttemptId(), psAttempt2Id);
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 3);
        diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 2);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(1), psAttempt1Id + " failed due to: out of memory");
        ps = LocalClusterContext.get().getPS(psAttempt2Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock + 1, w2Clock);
        assertEquals(sum((IntIntVector) w1Task0Client.getRow(0)), 400000);
        // attempt1
        ps.stop(-1);
        request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt2Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        PSAttempt psAttempt3 = amPs.getPSAttempt(psAttempt3Id);
        assertTrue(psAttempt3 != null);
        assertEquals(psAttempt2.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(psAttempt3.getInternalState(), PSAttemptStateInternal.RUNNING);
        assertEquals(amPs.getState(), AMParameterServerState.RUNNING);
        assertEquals(amPs.getNextAttemptNumber(), 4);
        assertEquals(amPs.getRunningAttemptId(), psAttempt3Id);
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 4);
        diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 3);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(1), psAttempt1Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(2), psAttempt2Id + " failed due to: out of memory");
        ps = LocalClusterContext.get().getPS(psAttempt3Id).getPS();
        checkMatrixInfo(ps, w1Id, w2Id, w1Clock + 1, w2Clock);
        ps.stop(-1);
        request = PSErrorRequest.newBuilder().setPsAttemptId(ProtobufUtil.convertToIdProto(psAttempt3Id)).setMsg("out of memory").build();
        master.psError(null, request);
        Thread.sleep(heartbeatInterval * 2);
        assertEquals(psAttempt3.getInternalState(), PSAttemptStateInternal.FAILED);
        assertEquals(amPs.getState(), AMParameterServerState.FAILED);
        assertEquals(angelAppMaster.getAppContext().getApp().getExternAppState(), AppState.FAILED);
        assertEquals(amPs.getNextAttemptNumber(), 4);
        assertNull(amPs.getRunningAttemptId());
        assertNull(amPs.getSuccessAttemptId());
        assertEquals(amPs.getPSAttempts().size(), 4);
        diagnostics = amPs.getDiagnostics();
        assertEquals(diagnostics.size(), 4);
        assertEquals(diagnostics.get(0), psAttempt0Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(1), psAttempt1Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(2), psAttempt2Id + " failed due to: out of memory");
        assertEquals(diagnostics.get(3), psAttempt3Id + " failed due to: out of memory");
    } catch (Exception x) {
        LOG.error("run testPSError failed ", x);
        throw x;
    }
}

Example 47 with IntIntVector

use of com.tencent.angel.ml.math2.vector.IntIntVector in project angel by Tencent.

the class UpdatePSFTest method testSparseIntUDF.

public void testSparseIntUDF() throws Exception {
    Worker worker = LocalClusterContext.get().getWorker(workerAttempt0Id).getWorker();
    MatrixClient client1 = worker.getPSAgent().getMatrixClient(SPARSE_INT_MAT, 0);
    int matrixW1Id = client1.getMatrixId();
    int[] index = genIndexs(feaNum, nnz);
    IntIntVector deltaVec = new IntIntVector(feaNum, new IntIntSparseVectorStorage(feaNum, nnz));
    for (int i = 0; i < index.length; i++) {
        deltaVec.set(index[i], index[i]);
    }
    // for (int i = 0; i < feaNum; i++) {
    // deltaVec.set(i, i);
    // }
    deltaVec.setRowId(0);
    Vector[] updates = new Vector[1];
    updates[0] = deltaVec;
    client1.asyncUpdate(new IncrementRows(new IncrementRowsParam(matrixW1Id, updates))).get();
    IntIntVector row = (IntIntVector) client1.getRow(0);
    for (int id : index) {
        // System.out.println("id=" + id + ", value=" + row.get(id));
        Assert.assertTrue(row.get(id) == deltaVec.get(id));
    }
    Assert.assertTrue(index.length == row.size());
}

Example 48 with IntIntVector

use of com.tencent.angel.ml.math2.vector.IntIntVector in project angel by Tencent.

the class UpdatePSFTest method testDenseIntUDF.

public void testDenseIntUDF() throws Exception {
    Worker worker = LocalClusterContext.get().getWorker(workerAttempt0Id).getWorker();
    MatrixClient client1 = worker.getPSAgent().getMatrixClient(DENSE_INT_MAT, 0);
    int matrixW1Id = client1.getMatrixId();
    int[] index = genIndexs(feaNum, nnz);
    IntIntVector deltaVec = new IntIntVector(feaNum, new IntIntDenseVectorStorage(feaNum));
    for (int i = 0; i < feaNum; i++) {
        deltaVec.set(i, i);
    }
    deltaVec.setRowId(0);
    Vector[] updates = new Vector[1];
    updates[0] = deltaVec;
    client1.asyncUpdate(new IncrementRows(new IncrementRowsParam(matrixW1Id, updates))).get();
    IntIntVector row = (IntIntVector) client1.getRow(0);
    for (int id : index) {
        Assert.assertTrue(row.get(id) == deltaVec.get(id));
    }
    Assert.assertTrue(feaNum == row.size());
}

Example 49 with IntIntVector

use of com.tencent.angel.ml.math2.vector.IntIntVector in project angel by Tencent.

the class SnapshotFormat method save.

private void save(ServerLongIntRow row, PSMatrixSaveContext saveContext, MatrixPartitionMeta meta, DataOutputStream out) throws IOException {
    long startCol = meta.getStartCol();
    if (ServerRowUtils.getVector(row) instanceof IntIntVector) {
        IntIntVector vector = (IntIntVector) ServerRowUtils.getVector(row);
        if (vector.isDense()) {
            int[] data = vector.getStorage().getValues();
            for (int i = 0; i < data.length; i++) {
                out.writeInt(data[i]);
            }
        } else if (vector.isSorted()) {
            int[] indices = vector.getStorage().getIndices();
            int[] values = vector.getStorage().getValues();
            for (int i = 0; i < indices.length; i++) {
                out.writeLong(indices[i] + startCol);
                out.writeInt(values[i]);
            }
        } else {
            ObjectIterator<Int2IntMap.Entry> iter = vector.getStorage().entryIterator();
            Int2IntMap.Entry entry;
            while (iter.hasNext()) {
                entry = iter.next();
                out.writeLong(entry.getIntKey() + startCol);
                out.writeInt(entry.getIntValue());
            }
        }
    } else {
        LongIntVector vector = (LongIntVector) ServerRowUtils.getVector(row);
        if (vector.isSorted()) {
            long[] indices = vector.getStorage().getIndices();
            int[] values = vector.getStorage().getValues();
            for (int i = 0; i < indices.length; i++) {
                out.writeLong(indices[i] + startCol);
                out.writeInt(values[i]);
            }
        } else {
            ObjectIterator<Long2IntMap.Entry> iter = vector.getStorage().entryIterator();
            Long2IntMap.Entry entry;
            while (iter.hasNext()) {
                entry = iter.next();
                out.writeLong(entry.getLongKey() + startCol);
                out.writeInt(entry.getIntValue());
            }
        }
    }
}

Example 50 with IntIntVector

use of com.tencent.angel.ml.math2.vector.IntIntVector in project angel by Tencent.

the class HashRouterUtils method split.

/**
 * Split keys by matrix partition
 *
 * @param matrixMeta matrix meta data
 * @param vector Matrix vector
 * @return partition key to key partition map
 */
public static KeyValuePart[] split(MatrixMeta matrixMeta, Vector vector) {
    KeyHash hasher = HasherFactory.getHasher(matrixMeta.getRouterHash());
    PartitionKey[] matrixParts = matrixMeta.getPartitionKeys();
    KeyValuePart[] dataParts = new KeyValuePart[matrixParts.length];
    int estSize = (int) (vector.getSize() / matrixMeta.getPartitionNum());
    for (int i = 0; i < dataParts.length; i++) {
        dataParts[i] = generateDataPart(vector.getRowId(), vector.getType(), estSize);
    }
    switch(vector.getType()) {
        case T_DOUBLE_DENSE:
        case T_DOUBLE_SPARSE:
            {
                splitIntDoubleVector(hasher, matrixMeta, (IntDoubleVector) vector, dataParts);
                break;
            }
        case T_FLOAT_DENSE:
        case T_FLOAT_SPARSE:
            {
                splitIntFloatVector(hasher, matrixMeta, (IntFloatVector) vector, dataParts);
                break;
            }
        case T_INT_DENSE:
        case T_INT_SPARSE:
            {
                splitIntIntVector(hasher, matrixMeta, (IntIntVector) vector, dataParts);
                break;
            }
        case T_LONG_DENSE:
        case T_LONG_SPARSE:
            {
                splitIntLongVector(hasher, matrixMeta, (IntLongVector) vector, dataParts);
                break;
            }
        case T_DOUBLE_SPARSE_LONGKEY:
            {
                splitLongDoubleVector(hasher, matrixMeta, (LongDoubleVector) vector, dataParts);
                break;
            }
        case T_FLOAT_SPARSE_LONGKEY:
            {
                splitLongFloatVector(hasher, matrixMeta, (LongFloatVector) vector, dataParts);
                break;
            }
        case T_INT_SPARSE_LONGKEY:
            {
                splitLongIntVector(hasher, matrixMeta, (LongIntVector) vector, dataParts);
                break;
            }
        case T_LONG_SPARSE_LONGKEY:
            {
                splitLongLongVector(hasher, matrixMeta, (LongLongVector) vector, dataParts);
                break;
            }
        default:
            {
                throw new UnsupportedOperationException("Unsupport vector type " + vector.getType());
            }
    }
    for (int i = 0; i < dataParts.length; i++) {
        if (dataParts[i] != null) {
            dataParts[i].setRowId(vector.getRowId());
        }
    }
    return dataParts;
}