@Override
public Value lazyArrayChunk(final Key key) {
final Key arykey = ValueArray.getArrayKey(key); // From the base file key
final long off = (_iceRoot != null) ? 0 : ValueArray.getChunkOffset(key); // The offset
final Path p =
(_iceRoot != null)
? new Path(_iceRoot, getIceName(key, (byte) 'V'))
: new Path(arykey.toString());
final Size sz = new Size();
run(
new Callable() {
@Override
public Object call() throws Exception {
FileSystem fs = FileSystem.get(p.toUri(), CONF);
long rem = fs.getFileStatus(p).getLen() - off;
sz._value = (rem > ValueArray.CHUNK_SZ * 2) ? (int) ValueArray.CHUNK_SZ : (int) rem;
return null;
}
},
true,
0);
Value val = new Value(key, sz._value, Value.HDFS);
val.setdsk(); // But its already on disk.
return val;
}
/**
* Creates the value header based on the calculated columns.
*
* <p>Also stores the header to its appropriate key. This will be the VA header of the parsed
* dataset.
*/
private void createValueArrayHeader() {
assert (_phase == Pass.TWO);
Column[] cols = new Column[_ncolumns];
int off = 0;
for (int i = 0; i < cols.length; ++i) {
cols[i] = new Column();
cols[i]._n = _numRows - _invalidValues[i];
cols[i]._base = _bases[i];
assert (char) pow10i(-_scale[i]) == pow10i(-_scale[i])
: "scale out of bounds!, col = " + i + ", scale = " + _scale[i];
cols[i]._scale = (char) pow10i(-_scale[i]);
cols[i]._off = (char) off;
cols[i]._size = (byte) COL_SIZES[_colTypes[i]];
cols[i]._domain = _colDomains[i];
cols[i]._max = _max[i];
cols[i]._min = _min[i];
cols[i]._mean = _mean[i];
cols[i]._sigma = _sigma[i];
cols[i]._name = _colNames[i];
off += Math.abs(cols[i]._size);
}
// let any pending progress reports finish
DKV.write_barrier();
// finally make the value array header
ValueArray ary = new ValueArray(_resultKey, _numRows, off, cols);
UKV.put(_resultKey, ary.value());
}
// Read up to 'len' bytes of Value. Value should already be persisted to
// disk. A racing delete can trigger a failure where we get a null return,
// but no crash (although one could argue that a racing load&delete is a bug
// no matter what).
@Override
public byte[] load(Value v) {
long skip = 0;
Key k = v._key;
// Convert an arraylet chunk into a long-offset from the base file.
if (k._kb[0] == Key.ARRAYLET_CHUNK) {
skip = ValueArray.getChunkOffset(k); // The offset
k = ValueArray.getArrayKey(k); // From the base file key
}
if (k._kb[0] == Key.DVEC) {
skip = water.fvec.NFSFileVec.chunkOffset(k); // The offset
}
try {
FileInputStream s = null;
try {
s = new FileInputStream(getFileForKey(k));
FileChannel fc = s.getChannel();
fc.position(skip);
AutoBuffer ab = new AutoBuffer(fc, true, Value.NFS);
byte[] b = ab.getA1(v._max);
ab.close();
assert v.isPersisted();
return b;
} finally {
if (s != null) s.close();
}
} catch (IOException e) { // Broken disk / short-file???
H2O.ignore(e);
return null;
}
}
protected void testScalarExpression(String expr, double result) {
Key key = executeExpression(expr);
ValueArray va = ValueArray.value(key);
assertEquals(va.numRows(), 1);
assertEquals(va.numCols(), 1);
assertEquals(result, va.datad(0, 0), 0.0);
UKV.remove(key);
}
// Test kaggle/creditsample-test data
@org.junit.Test
public void kaggle_credit() {
Key okey = loadAndParseFile("credit.hex", "smalldata/kaggle/creditsample-training.csv.gz");
UKV.remove(Key.make("smalldata/kaggle/creditsample-training.csv.gz_UNZIPPED"));
UKV.remove(Key.make("smalldata\\kaggle\\creditsample-training.csv.gz_UNZIPPED"));
ValueArray val = DKV.get(okey).get();
// Check parsed dataset
final int n = new int[] {4, 2, 1}[ValueArray.LOG_CHK - 20];
assertEquals("Number of chunks", n, val.chunks());
assertEquals("Number of rows", 150000, val.numRows());
assertEquals("Number of cols", 12, val.numCols());
// setup default values for DRF
int ntrees = 3;
int depth = 30;
int gini = StatType.GINI.ordinal();
int seed = 42;
StatType statType = StatType.values()[gini];
final int cols[] =
new int[] {0, 2, 3, 4, 5, 7, 8, 9, 10, 11, 1}; // ignore column 6, classify column 1
// Start the distributed Random Forest
final Key modelKey = Key.make("model");
DRFJob result =
hex.rf.DRF.execute(
modelKey,
cols,
val,
ntrees,
depth,
1024,
statType,
seed,
true,
null,
-1,
Sampling.Strategy.RANDOM,
1.0f,
null,
0,
0,
false);
// Wait for completion on all nodes
RFModel model = result.get();
assertEquals("Number of classes", 2, model.classes());
assertEquals("Number of trees", ntrees, model.size());
model.deleteKeys();
UKV.remove(modelKey);
UKV.remove(okey);
}
@Override
public double score(ValueArray data, AutoBuffer ab, int row) {
int j = 0;
for (int i = 0; i < _xCols.length; ++i)
_row[j++] =
data.isNA(ab, row, _xCols[i])
? Double.NaN
: (_catMap == null || _catMap[i] == null)
? data.datad(ab, row, _xCols[i])
: translateCat(i, (int) data.data(ab, row, _xCols[i]));
return M.score0(_row);
}
/**
* Executes the phase one of the parser.
*
* <p>First phase detects the encoding and basic statistics of the parsed dataset.
*
* <p>For CSV parsers it detects the parser setup and then launches the distributed computation on
* per chunk basis.
*
* <p>For XLS and XLSX parsers that do not work in distrubuted way parses the whole datasets.
*
* @throws Exception
*/
public void passOne(CsvParser.Setup setup) throws Exception {
switch (_parserType) {
case CSV:
// precompute the parser setup, column setup and other settings
byte[] bits = _sourceDataset.getFirstBytes(); // Can limit to eg 256*1024
if (setup == null) setup = CsvParser.guessCsvSetup(bits);
if (setup._data == null) {
_error = "Unable to determine the separator or number of columns on the dataset";
return;
}
_colNames = setup._data[0];
setColumnNames(_colNames);
_skipFirstLine = setup._header;
// set the separator
this._sep = setup._separator;
// if parsing value array, initialize the nrows array
if (_sourceDataset._isArray != 0) {
ValueArray ary = ValueArray.value(_sourceDataset);
_nrows = new int[(int) ary.chunks()];
}
// launch the distributed parser on its chunks.
this.invoke(_sourceDataset._key);
break;
case XLS:
// XLS parsing is not distributed, just obtain the value stream and
// run the parser
CustomParser p = new XlsParser(this);
p.parse(_sourceDataset._key);
--_myrows; // do not count the header
break;
case XLSX:
// XLS parsing is not distributed, just obtain the value stream and
// run the parser
CustomParser px = new XlsxParser(this);
px.parse(_sourceDataset._key);
break;
default:
throw new Error("NOT IMPLEMENTED");
}
// calculate proper numbers of rows for the chunks
if (_nrows != null) {
_numRows = 0;
for (int i = 0; i < _nrows.length; ++i) {
_numRows += _nrows[i];
_nrows[i] = _numRows;
}
} else {
_numRows = _myrows;
}
// normalize mean
for (int i = 0; i < _ncolumns; ++i) _mean[i] = _mean[i] / (_numRows - _invalidValues[i]);
}
static Value lazyArrayChunk(Key key) {
Key arykey = ValueArray.getArrayKey(key); // From the base file key
long off = ValueArray.getChunkOffset(key); // The offset
long size = getFileForKey(arykey).length();
long rem = size - off;
// the last chunk can be fat, so it got packed into the earlier chunk
if (rem < ValueArray.CHUNK_SZ && off > 0) return null;
int sz = (rem >= ValueArray.CHUNK_SZ * 2) ? (int) ValueArray.CHUNK_SZ : (int) rem;
Value val = new Value(key, sz, Value.NFS);
val.setdsk(); // But its already on disk.
return val;
}
// Recursively remove, gathering all the pending remote key-deletes
private static void remove(Key key, Futures fs) {
Value val = DKV.get(key, 32); // Get the existing Value, if any
if (val == null) return; // Trivial delete
if (val._isArray != 0) { // See if this is an Array
ValueArray ary = ValueArray.value(val);
for (long i = 0; i < ary.chunks(); i++) // Delete all the chunks
remove(ary.getChunkKey(i), fs);
}
if (key._kb[0] == Key.KEY_OF_KEYS) // Key-of-keys?
for (Key k : val.flatten()) // Then recursively delete
remove(k, fs);
DKV.remove(key, fs);
}
public static Response redirect(JsonObject fromPageResponse, Key rfModelKey) {
RFModel rfModel = DKV.get(rfModelKey).get();
ValueArray data = DKV.get(rfModel._dataKey).get();
return redirect(
fromPageResponse,
null,
rfModelKey,
rfModel._dataKey,
rfModel._totalTrees,
data.numCols() - 1,
null,
true,
false);
}
/*@org.junit.Test*/ public void covtype() {
// Key okey = loadAndParseFile("covtype.hex", "smalldata/covtype/covtype.20k.data");
// Key okey = loadAndParseFile("covtype.hex", "../datasets/UCI/UCI-large/covtype/covtype.data");
// Key okey = loadAndParseFile("covtype.hex", "/home/0xdiag/datasets/standard/covtype.data");
Key okey = loadAndParseFile("mnist.hex", "smalldata/mnist/mnist8m.10k.csv.gz");
// Key okey = loadAndParseFile("mnist.hex", "/home/0xdiag/datasets/mnist/mnist8m.csv");
ValueArray val = UKV.get(okey);
// setup default values for DRF
int ntrees = 8;
int depth = 999;
int gini = StatType.ENTROPY.ordinal();
int seed = 42;
StatType statType = StatType.values()[gini];
final int cols[] = new int[val.numCols()];
for (int i = 1; i < cols.length; i++) cols[i] = i - 1;
cols[cols.length - 1] = 0; // Class is in column 0 for mnist
// Start the distributed Random Forest
final Key modelKey = Key.make("model");
DRFJob result =
hex.rf.DRF.execute(
modelKey,
cols,
val,
ntrees,
depth,
1024,
statType,
seed,
true,
null,
-1,
Sampling.Strategy.RANDOM,
1.0f,
null,
0,
0,
false);
// Wait for completion on all nodes
RFModel model = result.get();
assertEquals("Number of classes", 10, model.classes());
assertEquals("Number of trees", ntrees, model.size());
model.deleteKeys();
UKV.remove(modelKey);
UKV.remove(okey);
}
public static void put(Key key, Value val, Futures fs) {
Value res = DKV.put(key, val, fs);
// If the old Value was a large array, we need to delete the leftover
// chunks - they are unrelated to the new Value which might be either
// bigger or smaller than the old Value.
if (res != null && res._isArray != 0) {
ValueArray ary = ValueArray.value(res);
for (long i = 0; i < ary.chunks(); i++) // Delete all the chunks
DKV.remove(ary.getChunkKey(i), fs);
}
if (key._kb[0] == Key.KEY_OF_KEYS) // Key-of-keys?
for (Key k : res.flatten()) // Then recursively delete
remove(k, fs);
if (res != null) res.freeMem();
}
@Override
public byte[] load(final Value v) {
final byte[] b = MemoryManager.malloc1(v._max);
long skip = 0;
Key k = v._key;
final Path p;
if (_iceRoot != null) {
p = new Path(_iceRoot, getIceName(v));
} else {
// Convert an arraylet chunk into a long-offset from the base file.
if (k._kb[0] == Key.ARRAYLET_CHUNK) {
skip = ValueArray.getChunkOffset(k); // The offset
k = ValueArray.getArrayKey(k); // From the base file key
if (k.toString().endsWith(Extensions.HEX)) { // Hex file?
int value_len = DKV.get(k).memOrLoad().length; // How long is the ValueArray header?
skip += value_len;
}
}
p = new Path(k.toString());
}
final long skip_ = skip;
run(
new Callable() {
@Override
public Object call() throws Exception {
FileSystem fs = FileSystem.get(p.toUri(), CONF);
FSDataInputStream s = null;
try {
s = fs.open(p);
// NOTE:
// The following line degrades performance of HDFS load from S3 API:
// s.readFully(skip,b,0,b.length);
// Google API's simple seek has better performance
// Load of 300MB file via Google API ~ 14sec, via s.readFully ~ 5min (under the same
// condition)
ByteStreams.skipFully(s, skip_);
ByteStreams.readFully(s, b);
assert v.isPersisted();
} finally {
Utils.close(s);
}
return null;
}
},
true,
v._max);
return b;
}
/**
* Stores the stream to its chunk using the atomic union. After the data from the stream is
* stored, its memory is freed up.
*/
public void store() {
assert _ab.eof();
Key k = ValueArray.getChunkKey(_chunkIndex, _resultKey);
AtomicUnion u = new AtomicUnion(_ab.bufClose(), _chunkOffset);
alsoBlockFor(u.fork(k));
_ab = null; // free mem
}
public static String store2Hdfs(Key srcKey) {
assert srcKey._kb[0] != Key.ARRAYLET_CHUNK;
assert PersistHdfs.getPathForKey(srcKey) != null; // Validate key name
Value v = DKV.get(srcKey);
if (v == null) return "Key " + srcKey + " not found";
if (!v.isArray()) { // Simple chunk?
v.setHdfs(); // Set to HDFS and be done
return null; // Success
}
// For ValueArrays, make the .hex header
ValueArray ary = v.get();
String err = PersistHdfs.freeze(srcKey, ary);
if (err != null) return err;
// The task managing which chunks to write next,
// store in a known key
TaskStore2HDFS ts = new TaskStore2HDFS(srcKey);
Key selfKey = ts.selfKey();
UKV.put(selfKey, ts);
// Then start writing chunks in-order with the zero chunk
H2ONode chk0_home = ValueArray.getChunkKey(0, srcKey).home_node();
RPC.call(ts.chunkHome(), ts);
// Watch the progress key until it gets removed or an error appears
long idx = 0;
while ((ts = UKV.get(selfKey, TaskStore2HDFS.class)) != null) {
if (ts._indexFrom != idx) {
System.out.print(" " + idx + "/" + ary.chunks());
idx = ts._indexFrom;
}
if (ts._err != null) { // Found an error?
UKV.remove(selfKey); // Cleanup & report
return ts._err;
}
try {
Thread.sleep(100);
} catch (InterruptedException e) {
}
}
System.out.println(" " + ary.chunks() + "/" + ary.chunks());
// PersistHdfs.refreshHDFSKeys();
return null;
}
protected void testKeyValues(
Key k, double n1, double n2, double n3, double nx3, double nx2, double nx1) {
ValueArray v = ValueArray.value(k);
assertEquals(v.datad(0, 0), n1, 0.0);
assertEquals(v.datad(1, 0), n2, 0.0);
assertEquals(v.datad(2, 0), n3, 0.0);
assertEquals(v.datad(v.numRows() - 3, 0), nx3, 0.0);
assertEquals(v.datad(v.numRows() - 2, 0), nx2, 0.0);
assertEquals(v.datad(v.numRows() - 1, 0), nx1, 0.0);
}
public static KMeansScore score(KMeansModel model, ValueArray ary) {
KMeansScore kms = new KMeansScore();
kms._arykey = ary._key;
kms._cols = model.columnMapping(ary.colNames());
kms._clusters = model._clusters;
kms._normalized = model._normalized;
kms.invoke(ary._key);
return kms;
}
// User-Weak-Get a Key from the distributed cloud.
// Right now, just gets chunk#0 from a ValueArray, or a normal Value otherwise.
public static Value get(Key key) {
Value val = DKV.get(key);
if (val != null && val._isArray != 0) {
Key k2 = ValueArray.getChunkKey(0, key);
Value vchunk0 = DKV.get(k2);
assert vchunk0 != null : "missed looking for key " + k2 + " from " + key;
return vchunk0; // Else just get the prefix asked for
}
return val;
}
@Override
public void map(Key key) {
_rows = new long[_clusters.length];
_dist = new double[_clusters.length];
assert key.home();
ValueArray va = DKV.get(_arykey).get();
AutoBuffer bits = va.getChunk(key);
int rows = va.rpc(ValueArray.getChunkIndex(key));
double[] values = new double[_cols.length - 1];
ClusterDist cd = new ClusterDist();
for (int row = 0; row < rows; row++) {
KMeans.datad(va, bits, row, _cols, _normalized, values);
KMeans.closest(_clusters, values, cd);
_rows[cd._cluster]++;
_dist[cd._cluster] += cd._dist;
}
_arykey = null;
_cols = null;
_clusters = null;
}
@Override
public byte[] load(final Value v) {
final byte[] b = MemoryManager.malloc1(v._max);
long skip = 0;
Key k = v._key;
if (k._kb[0] == Key.ARRAYLET_CHUNK) {
skip = ValueArray.getChunkOffset(k); // The offset
k = ValueArray.getArrayKey(k); // From the base file key
} else if (k._kb[0] == Key.DVEC) {
skip = water.fvec.NFSFileVec.chunkOffset(k); // The offset
}
final Path p =
_iceRoot == null ? new Path(getPathForKey(k)) : new Path(_iceRoot, getIceName(v));
final long skip_ = skip;
run(
new Callable() {
@Override
public Object call() throws Exception {
FileSystem fs = FileSystem.get(p.toUri(), CONF);
FSDataInputStream s = null;
try {
s = fs.open(p);
// NOTE:
// The following line degrades performance of HDFS load from S3 API:
// s.readFully(skip,b,0,b.length);
// Google API's simple seek has better performance
// Load of 300MB file via Google API ~ 14sec, via s.readFully ~ 5min (under the same
// condition)
ByteStreams.skipFully(s, skip_);
ByteStreams.readFully(s, b);
assert v.isPersisted();
} finally {
Utils.close(s);
}
return null;
}
},
true,
v._max);
return b;
}
@Override
public void compute() {
String path = null; // getPathFromValue(val);
ValueArray ary = ValueArray.value(_arykey);
Key self = selfKey();
while (_indexFrom < ary.chunks()) {
Key ckey = ary.getChunkKey(_indexFrom++);
if (!ckey.home()) { // Next chunk not At Home?
RPC.call(chunkHome(), this); // Hand the baton off to the next node/chunk
return;
}
Value val = DKV.get(ckey); // It IS home, so get the data
_err = PersistHdfs.appendChunk(_arykey, val);
if (_err != null) return;
UKV.put(self, this); // Update the progress/self key
}
// We did the last chunk. Removing the selfKey is the signal to the web
// thread that All Done.
UKV.remove(self);
}
// Bridge from new Model scoring to old Model scoring
public Frame score(Frame data) {
final double threshold = getThreshold();
String[][] ds = _va.domains();
if (ds[ds.length - 1] == null && !Double.isNaN(threshold)) {
// This is a binomial classifier
ds[ds.length - 1] = new String[] {"F", "T"};
}
Model m =
new Model(null, null, _va.colNames(), ds) {
@Override
protected float[] score0(double data[ /*ncols*/], float preds[ /*nclasses*/]) {
float s = (float) OldModel.this.score0(data);
if (preds.length == 1) preds[0] = s;
else {
assert preds.length == 2;
preds[0] = 1 - s;
preds[1] = s;
}
return preds;
}
};
return m.score(data);
}
private void updateClusters(
int[] clusters, int count, long chunk, long numrows, int rpc, long updatedRow) {
final int offset = (int) (updatedRow - (rpc * chunk));
final Key chunkKey = ValueArray.getChunkKey(chunk, _job.dest());
final int[] message;
if (count == clusters.length) message = clusters;
else {
message = new int[count];
System.arraycopy(clusters, 0, message, 0, message.length);
}
final int rows = ValueArray.rpc(chunk, rpc, numrows);
new Atomic() {
@Override
public Value atomic(Value val) {
assert val == null || val._key.equals(chunkKey);
AutoBuffer b = new AutoBuffer(rows * ROW_SIZE);
if (val != null) b._bb.put(val.memOrLoad());
for (int i = 0; i < message.length; i++) b.put4((offset + i) * 4, message[i]);
b.position(b.limit());
return new Value(chunkKey, b.buf());
}
}.invoke(chunkKey);
}
/**
* Adapt model for the given dataset. Default behavior is to map columns and categoricals to their
* original indexes. Categorical values we have not seen when building the model are translated as
* NaN.
*
* <p>Override this to get custom adapt behavior (eg. handle unseen cats differently).
*
* @param ary - tst dataset
* @return OldModel - model adapted to be applied on the given data
*/
public OldModel adapt(ValueArray ary) {
boolean id = true;
final int[] colMap = columnMapping(ary.colNames());
if (!isCompatible(colMap))
throw new IllegalArgumentException("This model uses different columns than those provided");
int[][] catMap = new int[colMap.length][];
for (int i = 0; i < colMap.length - 1; ++i) {
Column c = ary._cols[colMap[i]];
if (c.isEnum() && !Arrays.deepEquals(_va._cols[i]._domain, c._domain)) {
id = false;
catMap[i] = new int[c._domain.length];
for (int j = 0; j < c._domain.length; ++j)
catMap[i][j] = find(c._domain[j], _va._cols[i]._domain);
}
}
if (id && identityMap(colMap)) catMap = null;
return new ModelDataAdaptor(
this, colMap[colMap.length - 1], Arrays.copyOf(colMap, colMap.length - 1), catMap);
}
public static Job run(final Key dest, final KMeansModel model, final ValueArray ary) {
final ChunkProgressJob job = new ChunkProgressJob(ary.chunks(), dest);
new ValueArray(dest, 0).delete_and_lock(job.self());
final H2OCountedCompleter fjtask =
new H2OCountedCompleter() {
@Override
public void compute2() {
KMeansApply kms = new KMeansApply();
kms._job = job;
kms._arykey = ary._key;
kms._cols = model.columnMapping(ary.colNames());
kms._clusters = model._clusters;
kms._normalized = model._normalized;
kms.invoke(ary._key);
Column c = new Column();
c._name = Constants.RESPONSE;
c._size = ROW_SIZE;
c._scale = 1;
c._min = 0;
c._max = model._clusters.length;
c._mean = Double.NaN;
c._sigma = Double.NaN;
c._domain = null;
c._n = ary.numRows();
ValueArray res = new ValueArray(dest, ary.numRows(), c._size, new Column[] {c});
res.unlock(job.self());
job.remove();
tryComplete();
}
@Override
public boolean onExceptionalCompletion(Throwable ex, CountedCompleter caller) {
job.onException(ex);
return super.onExceptionalCompletion(ex, caller);
}
};
job.start(fjtask);
H2O.submitTask(fjtask);
return job;
}
private H2ONode chunkHome() {
return ValueArray.getChunkKey(_indexFrom, _arykey).home_node();
}
/**
* Creates a new ValueArray with classes. New ValueArray is not aligned with source one
* unfortunately so have to send results to each chunk owner using Atomic.
*/
@Override
public void map(Key key) {
assert key.home();
if (Job.isRunning(_job.self())) {
ValueArray va = DKV.get(_arykey).get();
AutoBuffer bits = va.getChunk(key);
long startRow = va.startRow(ValueArray.getChunkIndex(key));
int rows = va.rpc(ValueArray.getChunkIndex(key));
int rpc = (int) (ValueArray.CHUNK_SZ / ROW_SIZE);
long chunk = ValueArray.chknum(startRow, va.numRows(), ROW_SIZE);
long updatedChk = chunk;
long updatedRow = startRow;
double[] values = new double[_cols.length - 1];
ClusterDist cd = new ClusterDist();
int[] clusters = new int[rows];
int count = 0;
for (int row = 0; row < rows; row++) {
KMeans.datad(va, bits, row, _cols, _normalized, values);
KMeans.closest(_clusters, values, cd);
chunk = ValueArray.chknum(startRow + row, va.numRows(), ROW_SIZE);
if (chunk != updatedChk) {
updateClusters(clusters, count, updatedChk, va.numRows(), rpc, updatedRow);
updatedChk = chunk;
updatedRow = startRow + row;
count = 0;
}
clusters[count++] = cd._cluster;
}
if (count > 0) updateClusters(clusters, count, chunk, va.numRows(), rpc, updatedRow);
_job.updateProgress(1);
}
_job = null;
_arykey = null;
_cols = null;
_clusters = null;
}
/**
* Simple GLM wrapper to enable launching GLM from command line.
*
* <p>Example input: java -jar target/h2o.jar -name=test -runMethod water.util.GLMRunner
* -file=smalldata/logreg/prostate.csv -y=CAPSULE -family=binomial
*
* @param args
* @throws InterruptedException
*/
public static void main(String[] args) throws InterruptedException {
try {
GLMArgs ARGS = new GLMArgs();
new Arguments(args).extract(ARGS);
System.out.println("==================<GLMRunner START>===================");
ValueArray ary = Utils.loadAndParseKey(ARGS.file);
int ycol;
try {
ycol = Integer.parseInt(ARGS.y);
} catch (NumberFormatException e) {
ycol = ary.getColumnIds(new String[] {ARGS.y})[0];
}
int ncols = ary.numCols();
if (ycol < 0 || ycol >= ary.numCols()) {
System.err.println("invalid y column: " + ycol);
H2O.exit(-1);
}
int[] xcols;
if (ARGS.xs.equalsIgnoreCase("all")) {
xcols = new int[ncols - 1];
for (int i = 0; i < ycol; ++i) xcols[i] = i;
for (int i = ycol; i < ncols - 1; ++i) xcols[i] = i + 1;
} else {
System.out.println("xs = " + ARGS.xs);
String[] names = ARGS.xs.split(",");
xcols = new int[names.length];
try {
for (int i = 0; i < names.length; ++i) xcols[i] = Integer.valueOf(names[i]);
} catch (NumberFormatException e) {
xcols = ary.getColumnIds(ARGS.xs.split(","));
}
}
for (int x : xcols)
if (x < 0) {
System.err.println("Invalid predictor specification " + ARGS.xs);
H2O.exit(-1);
}
GLMJob j =
DGLM.startGLMJob(
DGLM.getData(ary, xcols, ycol, null, true),
new ADMMSolver(ARGS.lambda, ARGS._alpha),
new GLMParams(Family.valueOf(ARGS.family)),
null,
ARGS.xval,
true);
System.out.print("[GLM] computing model...");
int progress = 0;
while (!j.isDone()) {
int p = (int) (100 * j.progress());
int dots = p - progress;
progress = p;
for (int i = 0; i < dots; ++i) System.out.print('.');
Thread.sleep(250);
}
Log.debug(Sys.GENLM, "DONE.");
GLMModel m = j.get();
String[] colnames = ary.colNames();
System.out.println("Intercept" + " = " + m._beta[ncols - 1]);
for (int i = 0; i < xcols.length; ++i) {
System.out.println(colnames[i] + " = " + m._beta[i]);
}
} catch (Throwable t) {
Log.err(t);
} finally { // we're done. shutdown the cloud
Log.debug(Sys.GENLM, "==================<GLMRunner DONE>===================");
UDPRebooted.suicide(UDPRebooted.T.shutdown, H2O.SELF);
}
}
/**
* Map function for distributed parsing of the CSV files.
*
* <p>In first phase it calculates the min, max, means, encodings and other statistics about the
* dataset, determines the number of columns.
*
* <p>The second pass then encodes the parsed dataset to the result key, splitting it into equal
* sized chunks.
*/
@Override
public void map(Key key) {
try {
Key aryKey = null;
boolean arraylet = key._kb[0] == Key.ARRAYLET_CHUNK;
boolean skipFirstLine = _skipFirstLine;
if (arraylet) {
aryKey = ValueArray.getArrayKey(key);
_chunkId = ValueArray.getChunkIndex(key);
skipFirstLine = skipFirstLine || (ValueArray.getChunkIndex(key) != 0);
}
switch (_phase) {
case ONE:
assert (_ncolumns != 0);
// initialize the column statistics
phaseOneInitialize();
// perform the parse
CsvParser p = new CsvParser(aryKey, _ncolumns, _sep, _decSep, this, skipFirstLine);
p.parse(key);
if (arraylet) {
long idx = ValueArray.getChunkIndex(key);
int idx2 = (int) idx;
assert idx2 == idx;
assert (_nrows[idx2] == 0)
: idx
+ ": "
+ Arrays.toString(_nrows)
+ " ("
+ _nrows[idx2]
+ " -- "
+ _myrows
+ ")";
_nrows[idx2] = _myrows;
}
break;
case TWO:
assert (_ncolumns != 0);
// initialize statistics - invalid rows, sigma and row size
phaseTwoInitialize();
// calculate the first row and the number of rows to parse
int firstRow = 0;
int lastRow = _myrows;
_myrows = 0;
if (arraylet) {
long origChunkIdx = ValueArray.getChunkIndex(key);
firstRow = (origChunkIdx == 0) ? 0 : _nrows[(int) origChunkIdx - 1];
lastRow = _nrows[(int) origChunkIdx];
}
int rowsToParse = lastRow - firstRow;
// create the output streams
_outputStreams2 = createRecords(firstRow, rowsToParse);
assert (_outputStreams2.length > 0);
_ab = _outputStreams2[0].initialize();
// perform the second parse pass
CsvParser p2 = new CsvParser(aryKey, _ncolumns, _sep, _decSep, this, skipFirstLine);
p2.parse(key);
// store the last stream if not stored during the parse
if (_ab != null) _outputStreams2[_outputIdx].store();
break;
default:
assert (false);
}
ParseStatus.update(_resultKey, DKV.get(key).length(), _phase);
} catch (Exception e) {
e.printStackTrace();
_error = e.getMessage();
}
}
public void testDataFrameStructure(Key k, int rows, int cols) {
ValueArray v = ValueArray.value(k);
assertEquals(v.numRows(), rows);
assertEquals(v.numCols(), cols);
}
