public static ValueArray loadAndParseKey(Key okey, String path) {
FileIntegrityChecker c = FileIntegrityChecker.check(new File(path),false);
Key k = c.syncDirectory(null,null,null,null);
ParseDataset.forkParseDataset(okey, new Key[] { k }, null).get();
UKV.remove(k);
ValueArray res = DKV.get(okey).get();
return res;
}
protected Frame parse_test_file(Key outputKey, String fname, boolean guessSetup) {
File f = find_test_file(fname);
assert f != null && f.exists() : " file not found: " + fname;
NFSFileVec nfs = NFSFileVec.make(f);
return ParseDataset.parse(
outputKey,
new Key[] {nfs._key},
true,
ParseSetup.guessSetup(new Key[] {nfs._key}, false, 1));
}
/**
* Find & parse a folder of CSV files. NPE if file not found.
*
* @param fname Test filename
* @return Frame or NPE
*/
protected Frame parse_test_folder(String fname) {
File folder = find_test_file(fname);
assert folder.isDirectory();
File[] files = folder.listFiles();
Arrays.sort(files);
ArrayList<Key> keys = new ArrayList<Key>();
for (File f : files) if (f.isFile()) keys.add(NFSFileVec.make(f)._key);
Key[] res = new Key[keys.size()];
keys.toArray(res);
return ParseDataset.parse(Key.make(), res);
}
@Test
public void testChicago() {
Frame weather = null, crimes = null, census = null;
String oldtz = Exec.exec("(getTimeZone)").getStr();
try {
weather = parse_test_file(Key.make("weather.hex"), "smalldata/chicago/chicagoAllWeather.csv");
crimes =
parse_test_file(Key.make("crimes.hex"), "smalldata/chicago/chicagoCrimes10k.csv.zip");
String fname = "smalldata/chicago/chicagoCensus.csv";
File f = find_test_file(fname);
assert f != null && f.exists() : " file not found: " + fname;
NFSFileVec nfs = NFSFileVec.make(f);
ParseSetup ps = ParseSetup.guessSetup(new Key[] {nfs._key}, false, 1);
ps.getColumnTypes()[1] = Vec.T_ENUM;
census = ParseDataset.parse(Key.make("census.hex"), new Key[] {nfs._key}, true, ps);
census =
exec_str(
"(colnames= census.hex [0 1 2 3 4 5 6 7 8] [\"Community.Area.Number\" \"COMMUNITY.AREA.NAME\" \"PERCENT.OF.HOUSING.CROWDED\" \"PERCENT.HOUSEHOLDS.BELOW.POVERTY\" \"PERCENT.AGED.16..UNEMPLOYED\" \"PERCENT.AGED.25..WITHOUT.HIGH.SCHOOL.DIPLOMA\" \"PERCENT.AGED.UNDER.18.OR.OVER.64\" \"PER.CAPITA.INCOME.\" \"HARDSHIP.INDEX\"])",
"census.hex");
crimes =
exec_str(
"(colnames= crimes.hex [0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21] [\"ID\" \"Case.Number\" \"Date\" \"Block\" \"IUCR\" \"Primary.Type\" \"Description\" \"Location.Description\" \"Arrest\" \"Domestic\" \"Beat\" \"District\" \"Ward\" \"Community.Area\" \"FBI.Code\" \"X.Coordinate\" \"Y.Coordinate\" \"Year\" \"Updated.On\" \"Latitude\" \"Longitude\" \"Location\"])",
"crimes.hex");
exec_str("(setTimeZone \"Etc/UTC\")", null);
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= unary_op_6 (day (tmp= nary_op_5 (as.Date (cols crimes.hex [2]) \"%m/%d/%Y %I:%M:%S %p\")))) [22] [0:9999]) 22 \"Day\")",
"crimes.hex");
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= binary_op_31 (+ (tmp= unary_op_7 (month nary_op_5)) #1)) [23] [0:9999]) 23 \"Month\")",
"crimes.hex");
Keyed.remove(Key.make("nary_op_30"));
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= binary_op_32 (+ (tmp= binary_op_9 (- (tmp= unary_op_8 (year nary_op_5)) #1900)) #1900)) [17] [0:9999]) 17 \"Year\")",
"crimes.hex");
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= unary_op_10 (week nary_op_5)) [24] [0:9999]) 24 \"WeekNum\")",
"crimes.hex");
Keyed.remove(Key.make("binary_op_32"));
Keyed.remove(Key.make("binary_op_31"));
Keyed.remove(Key.make("unary_op_8"));
checkSaneFrame();
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= unary_op_11 (dayOfWeek nary_op_5)) [25] [0:9999]) 25 \"WeekDay\")",
"crimes.hex");
Keyed.remove(
Key.make(
"nfs:\\C:\\Users\\cliffc\\Desktop\\h2o-3\\smalldata\\chicago\\chicagoCrimes10k.csv.zip"));
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= unary_op_12 (hour nary_op_5)) [26] [0:9999]) 26 \"HourOfDay\")",
"crimes.hex");
crimes =
exec_str(
"(colnames= (= crimes.hex (tmp= nary_op_16 (ifelse (tmp= binary_op_15 (| (tmp= binary_op_13 (== unary_op_11 \"Sun\")) (tmp= binary_op_14 (== unary_op_11 \"Sat\")))) 1 0)) [27] [0:9999]) 27 \"Weekend\")",
"crimes.hex");
// Season is incorrectly assigned in the original chicago demo; picks up the Weekend flag
crimes =
exec_str(
"(colnames= (= crimes.hex nary_op_16 [28] [0:9999]) 28 \"Season\")", "crimes.hex");
// Standard "head of 10 rows" pattern for printing
Frame subset_33 = exec_str("(rows crimes.hex [0:10])", "subset_33");
Keyed.remove(Key.make("subset_33"));
Keyed.remove(Key.make("subset_33"));
Keyed.remove(Key.make("unary_op_29"));
Keyed.remove(Key.make("nary_op_28"));
Keyed.remove(Key.make("nary_op_27"));
Keyed.remove(Key.make("nary_op_26"));
Keyed.remove(Key.make("binary_op_25"));
Keyed.remove(Key.make("binary_op_24"));
Keyed.remove(Key.make("binary_op_23"));
Keyed.remove(Key.make("binary_op_22"));
Keyed.remove(Key.make("binary_op_21"));
Keyed.remove(Key.make("binary_op_20"));
Keyed.remove(Key.make("binary_op_19"));
Keyed.remove(Key.make("binary_op_18"));
Keyed.remove(Key.make("binary_op_17"));
Keyed.remove(Key.make("nary_op_16"));
Keyed.remove(Key.make("binary_op_15"));
Keyed.remove(Key.make("binary_op_14"));
Keyed.remove(Key.make("binary_op_13"));
Keyed.remove(Key.make("unary_op_12"));
Keyed.remove(Key.make("unary_op_11"));
Keyed.remove(Key.make("unary_op_10"));
Keyed.remove(Key.make("binary_op_9"));
Keyed.remove(Key.make("unary_op_8"));
Keyed.remove(Key.make("unary_op_7"));
Keyed.remove(Key.make("unary_op_6"));
Keyed.remove(Key.make("nary_op_5"));
checkSaneFrame();
// Standard "head of 10 rows" pattern for printing
Frame subset_34 = exec_str("(rows crimes.hex [0:10])", "subset_34");
Keyed.remove(Key.make("subset_34"));
census =
exec_str(
"(colnames= census.hex [0 1 2 3 4 5 6 7 8] [\"Community.Area\" \"COMMUNITY.AREA.NAME\" \"PERCENT.OF.HOUSING.CROWDED\" \"PERCENT.HOUSEHOLDS.BELOW.POVERTY\" \"PERCENT.AGED.16..UNEMPLOYED\" \"PERCENT.AGED.25..WITHOUT.HIGH.SCHOOL.DIPLOMA\" \"PERCENT.AGED.UNDER.18.OR.OVER.64\" \"PER.CAPITA.INCOME.\" \"HARDSHIP.INDEX\"])",
"census.hex");
Keyed.remove(Key.make("subset_34"));
Frame subset_35 = exec_str("(cols crimes.hex [-3])", "subset_35");
Frame subset_36 = exec_str("(cols weather.hex [-1])", "subset_36");
subset_36 =
exec_str(
"(colnames= subset_36 [0 1 2 3 4 5] [\"Month\" \"Day\" \"Year\" \"maxTemp\" \"meanTemp\" \"minTemp\"])",
"subset_36");
crimes.remove();
weather.remove();
// nary_op_37 = merge( X Y ); Vecs in X & nary_op_37 shared
Frame nary_op_37 = exec_str("(merge subset_35 census.hex FALSE FALSE)", "nary_op_37");
// nary_op_38 = merge( nary_op_37 subset_36); Vecs in nary_op_38 and nary_pop_37 and X shared
Frame subset_41 =
exec_str(
"(rows (tmp= nary_op_38 (merge nary_op_37 subset_36 TRUE FALSE)) (tmp= binary_op_40 (<= (tmp= nary_op_39 (h2o.runif nary_op_38 30792152736.5179)) #0.8)))",
"subset_41");
// Standard "head of 10 rows" pattern for printing
Frame subset_44 = exec_str("(rows subset_41 [0:10])", "subset_44");
Keyed.remove(Key.make("subset_44"));
Keyed.remove(Key.make("subset_44"));
Keyed.remove(Key.make("binary_op_40"));
Keyed.remove(Key.make("nary_op_37"));
Frame subset_43 =
exec_str("(rows nary_op_38 (tmp= binary_op_42 (> nary_op_39 #0.8)))", "subset_43");
// Chicago demo continues on past, but this is all I've captured for now
checkSaneFrame();
} finally {
Exec.exec(
"(setTimeZone \""
+ oldtz
+ "\")"); // Restore time zone (which is global, and will affect following tests)
if (weather != null) weather.remove();
if (crimes != null) crimes.remove();
if (census != null) census.remove();
for (String s :
new String[] {
"nary_op_5",
"unary_op_6",
"unary_op_7",
"unary_op_8",
"binary_op_9",
"unary_op_10",
"unary_op_11",
"unary_op_12",
"binary_op_13",
"binary_op_14",
"binary_op_15",
"nary_op_16",
"binary_op_17",
"binary_op_18",
"binary_op_19",
"binary_op_20",
"binary_op_21",
"binary_op_22",
"binary_op_23",
"binary_op_24",
"binary_op_25",
"nary_op_26",
"nary_op_27",
"nary_op_28",
"unary_op_29",
"binary_op_30",
"binary_op_31",
"binary_op_32",
"subset_33",
"subset_34",
"subset_35",
"subset_36",
"nary_op_37",
"nary_op_38",
"nary_op_39",
"binary_op_40",
"subset_41",
"binary_op_42",
"subset_43",
"subset_44",
}) Keyed.remove(Key.make(s));
}
}
public static Frame parse_test_file(Key outputKey, String fname) {
File f = find_test_file_static(fname);
assert f != null && f.exists() : " file not found: " + fname;
NFSFileVec nfs = NFSFileVec.make(f);
return ParseDataset.parse(outputKey, nfs._key);
}
@Test
@Ignore
public void run() {
Scope.enter();
try {
File file = find_test_file("bigdata/laptop/mnist/train.csv.gz");
File valid = find_test_file("bigdata/laptop/mnist/test.csv.gz");
if (file != null) {
NFSFileVec trainfv = NFSFileVec.make(file);
Frame frame = ParseDataset.parse(Key.make(), trainfv._key);
NFSFileVec validfv = NFSFileVec.make(valid);
Frame vframe = ParseDataset.parse(Key.make(), validfv._key);
DeepLearningParameters p = new DeepLearningParameters();
// populate model parameters
p._model_id = Key.make("dl_mnist_model");
p._train = frame._key;
// p._valid = vframe._key;
p._response_column = "C785"; // last column is the response
p._activation = DeepLearningParameters.Activation.RectifierWithDropout;
// p._activation = DeepLearningParameters.Activation.MaxoutWithDropout;
p._hidden = new int[] {800, 800};
p._input_dropout_ratio = 0.2;
p._mini_batch_size = 1;
p._train_samples_per_iteration = 50000;
p._score_duty_cycle = 0;
// p._shuffle_training_data = true;
// p._l1= 1e-5;
// p._max_w2= 10;
p._epochs = 10 * 5. / 6;
// Convert response 'C785' to categorical (digits 1 to 10)
int ci = frame.find("C785");
Scope.track(frame.replace(ci, frame.vecs()[ci].toEnum())._key);
Scope.track(vframe.replace(ci, vframe.vecs()[ci].toEnum())._key);
DKV.put(frame);
DKV.put(vframe);
// speed up training
p._adaptive_rate =
true; // disable adaptive per-weight learning rate -> default settings for learning rate
// and momentum are probably not ideal (slow convergence)
p._replicate_training_data =
true; // avoid extra communication cost upfront, got enough data on each node for load
// balancing
p._overwrite_with_best_model = true; // no need to keep the best model around
p._classification_stop = -1;
p._score_interval = 60; // score and print progress report (only) every 20 seconds
p._score_training_samples =
10000; // only score on a small sample of the training set -> don't want to spend too
// much time scoring (note: there will be at least 1 row per chunk)
DeepLearning dl = new DeepLearning(p);
DeepLearningModel model = null;
try {
model = dl.trainModel().get();
} catch (Throwable t) {
t.printStackTrace();
throw new RuntimeException(t);
} finally {
dl.remove();
if (model != null) {
model.delete();
}
}
} else {
Log.info("Please run ./gradlew syncBigDataLaptop in the top-level directory of h2o-3.");
}
} catch (Throwable t) {
t.printStackTrace();
throw new RuntimeException(t);
} finally {
Scope.exit();
}
}