public final void set_params(DeepLearningParameters p) {
parameters = (DeepLearningParameters) p.clone();
}
/**
* Main constructor
*
* @param params Model parameters
* @param dinfo Data Info
* @param nClasses number of classes (1 for regression, 0 for autoencoder)
* @param train User-given training data frame, prepared by AdaptTestTrain
* @param valid User-specified validation data frame, prepared by AdaptTestTrain
*/
public DeepLearningModelInfo(
final DeepLearningParameters params,
final DataInfo dinfo,
int nClasses,
Frame train,
Frame valid) {
_classification = nClasses > 1;
_train = train;
_valid = valid;
data_info = dinfo;
parameters =
(DeepLearningParameters) params.clone(); // make a copy, don't change model's parameters
DeepLearningParameters.Sanity.modifyParms(
parameters, parameters, nClasses); // sanitize the model_info's parameters
final int num_input = dinfo.fullN();
final int num_output =
get_params()._autoencoder
? num_input
: (_classification ? train.lastVec().cardinality() : 1);
if (!get_params()._autoencoder) assert (num_output == nClasses);
_saw_missing_cats = dinfo._cats > 0 ? new boolean[data_info._cats] : null;
assert (num_input > 0);
assert (num_output > 0);
if (has_momenta() && adaDelta())
throw new IllegalArgumentException(
"Cannot have non-zero momentum and adaptive rate at the same time.");
final int layers = get_params()._hidden.length;
// units (# neurons for each layer)
units = new int[layers + 2];
if (get_params()._max_categorical_features <= Integer.MAX_VALUE - dinfo._nums)
units[0] = Math.min(dinfo._nums + get_params()._max_categorical_features, num_input);
else units[0] = num_input;
System.arraycopy(get_params()._hidden, 0, units, 1, layers);
units[layers + 1] = num_output;
boolean printLevels = units[0] > 1000L;
boolean warn = units[0] > 100000L;
if (printLevels) {
final String[][] domains = dinfo._adaptedFrame.domains();
int[] levels = new int[domains.length];
for (int i = 0; i < levels.length; ++i) {
levels[i] = domains[i] != null ? domains[i].length : 0;
}
Arrays.sort(levels);
if (warn) {
Log.warn(
"===================================================================================================================================");
Log.warn(
num_input
+ " input features"
+ (dinfo._cats > 0 ? " (after categorical one-hot encoding)" : "")
+ ". Can be slow and require a lot of memory.");
}
if (levels[levels.length - 1] > 0) {
int levelcutoff = levels[levels.length - 1 - Math.min(10, levels.length - 1)];
int count = 0;
for (int i = 0;
i < dinfo._adaptedFrame.numCols() - (get_params()._autoencoder ? 0 : 1) && count < 10;
++i) {
if (dinfo._adaptedFrame.domains()[i] != null
&& dinfo._adaptedFrame.domains()[i].length >= levelcutoff) {
if (warn) {
Log.warn(
"Categorical feature '"
+ dinfo._adaptedFrame._names[i]
+ "' has cardinality "
+ dinfo._adaptedFrame.domains()[i].length
+ ".");
} else {
Log.info(
"Categorical feature '"
+ dinfo._adaptedFrame._names[i]
+ "' has cardinality "
+ dinfo._adaptedFrame.domains()[i].length
+ ".");
}
}
count++;
}
}
if (warn) {
Log.warn("Suggestions:");
Log.warn(" *) Limit the size of the first hidden layer");
if (dinfo._cats > 0) {
Log.warn(
" *) Limit the total number of one-hot encoded features with the parameter 'max_categorical_features'");
Log.warn(
" *) Run h2o.interaction(...,pairwise=F) on high-cardinality categorical columns to limit the factor count, see http://learn.h2o.ai");
}
Log.warn(
"===================================================================================================================================");
}
}
// weights (to connect layers)
dense_row_weights = new Storage.DenseRowMatrix[layers + 1];
dense_col_weights = new Storage.DenseColMatrix[layers + 1];
// decide format of weight matrices row-major or col-major
if (get_params()._col_major)
dense_col_weights[0] = new Storage.DenseColMatrix(units[1], units[0]);
else dense_row_weights[0] = new Storage.DenseRowMatrix(units[1], units[0]);
for (int i = 1; i <= layers; ++i)
dense_row_weights[i] = new Storage.DenseRowMatrix(units[i + 1] /*rows*/, units[i] /*cols*/);
// biases (only for hidden layers and output layer)
biases = new Storage.DenseVector[layers + 1];
for (int i = 0; i <= layers; ++i) biases[i] = new Storage.DenseVector(units[i + 1]);
// average activation (only for hidden layers)
if (get_params()._autoencoder && get_params()._sparsity_beta > 0) {
avg_activations = new Storage.DenseVector[layers];
mean_a = new float[layers];
for (int i = 0; i < layers; ++i) avg_activations[i] = new Storage.DenseVector(units[i + 1]);
}
allocateHelperArrays();
// for diagnostics
mean_rate = new float[units.length];
rms_rate = new float[units.length];
mean_bias = new float[units.length];
rms_bias = new float[units.length];
mean_weight = new float[units.length];
rms_weight = new float[units.length];
}
@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();
}
}