Brutex
/
cavis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix javadoc and cleanup 

Signed-off-by: brian <brian@brutex.de>
master
Brian Rosenberger 2022-10-21 16:36:56 +02:00
parent 656d367812
commit 4dfc637305
18 changed files with 317 additions and 252 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
build.gradle
View File

@ -56,6 +56,7 @@ configurations.all {
}
allprojects { Project proj ->
apply plugin: 'com.google.osdetector'
@ -162,3 +163,21 @@ allprojects { Project proj ->
}
}
}
task aggregatedJavadocs(type: Javadoc, description: 'Generate javadocs from all child projects as if it was a single project', group: 'Documentation') {
subprojects.each { proj ->
proj.tasks.withType(Javadoc).each { javadocTask ->
logger.quiet("Adding javadoc for project " + proj.name)
source += javadocTask.source
classpath += javadocTask.classpath
excludes += javadocTask.excludes
includes += javadocTask.includes
}
}
destinationDir = file("$buildDir/docs/javadoc")
title = "$project.name $version API"
options.author true
options.links 'http://docs.oracle.com/javase/8/docs/api/'
options.addStringOption('Xdoclint:none', '-quiet')
}

2
cavis-datavec/cavis-datavec-api/src/main/java/org/datavec/api/writable/Text.java
View File

@ -106,7 +106,7 @@ public class Text extends BinaryComparable implements WritableComparable<BinaryC
if (position < 0)
return -1; // duh.
ByteBuffer bb = ByteBuffer.wrap(bytes).position(position);
ByteBuffer bb = (ByteBuffer) ByteBuffer.wrap(bytes).position(position);
return bytesToCodePoint(bb.slice());
}

11
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/autodiff/samediff/SameDiff.java
View File

@ -1462,7 +1462,7 @@ public class SameDiff extends SDBaseOps {
* A special case of {@link #fit()}.
*
* @param iter The iterator to train the SameDiff instance with
* @param numEpochs The number of epochs for training. Must be > 0
* @param numEpochs The number of epochs for training. Must be &gt; 0
* @param validationIter The DataSetIterator to use for validation (null to skip validation)
* @param validationFrequency The frequency with which to run validation. 1 is every epoch, 2 is every other, etc.
* @param listeners Additional listeners to use during this operation
@ -1479,7 +1479,7 @@ public class SameDiff extends SDBaseOps {
* A special case of {@link #fit()}.
*
* @param iter The iterator to train the SameDiff instance with
* @param numEpochs The number of epochs for training. Must be > 0
* @param numEpochs The number of epochs for training. Must be &gt; 0
* @param listeners Additional listeners to use during this operation
* @return a {@link History} object containing the history information for this training operation
* (evaluations specified in the {@link TrainingConfig}, loss values, and timing information).
@ -1497,7 +1497,7 @@ public class SameDiff extends SDBaseOps {
* A special case of {@link #fit()}.
*
* @param iter The iterator to train the SameDiff instance with
* @param numEpochs The number of epochs for training. Must be > 0
* @param numEpochs The number of epochs for training. Must be &gt; 0
* @param validationIter The MultiDataSetIterator to use for validation (null to skip validation)
* @param validationFrequency The frequency with which to run validation. 1 is every epoch, 2 is every other, etc.
* @param listeners Additional listeners to use during this operation
@ -1514,7 +1514,7 @@ public class SameDiff extends SDBaseOps {
* A special case of {@link #fit()}.
*
* @param iter The iterator to train the SameDiff instance with
* @param numEpochs The number of epochs for training. Must be > 0
* @param numEpochs The number of epochs for training. Must be &gt; 0
* @param listeners Additional listeners to use during this operation
* @return a {@link History} object containing the history information for this training operation
* (evaluations specified in the {@link TrainingConfig}, loss values, and timing information).
@ -3036,7 +3036,6 @@ public class SameDiff extends SDBaseOps {
* See also: {@link VariableType}
*
* @param variables Variables to convert to constants
* @return The (now constant) SDVariables
*/
public void convertToConstants(List<SDVariable> variables) {
if (variables.size() == 0)
@ -3201,7 +3200,7 @@ public class SameDiff extends SDBaseOps {
* For example, {@code z(float) = x(float)+y(float)}, changing both x and y to double results in {@code z(double) = x(double)+y(double)}
* without doing anything to change z's datatype directly (z datatype is inferred from x + y + add op).<br>
* ARRAY type SDVariables cannot be converted directly, as their datatypes are determined by the function +
* input datatypes.<b>
* input datatypes.<br>
* Note that this method should be used with caution: incorrect datatype modifications may leave your network
* in an incorrect state. For example, {@code op(x(float),y(float)) -> op(x(double),y(float))} may not be
* supported by all ops.

60
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/autodiff/samediff/ops/SDBaseOps.java
View File

@ -382,7 +382,7 @@ public class SDBaseOps {
}
/**
* Cast the array to a new datatype - for example, Integer -> Float<br>
* Cast the array to a new datatype - for example, Integer -&gt; Float<br>
*
* @param arg Input variable to cast (NDARRAY type)
* @param datatype Datatype to cast to
@ -393,7 +393,7 @@ public class SDBaseOps {
}
/**
* Cast the array to a new datatype - for example, Integer -> Float<br>
* Cast the array to a new datatype - for example, Integer -&gt; Float<br>
*
* @param name name May be null. Name for the output variable
* @param arg Input variable to cast (NDARRAY type)
@ -654,7 +654,7 @@ public class SDBaseOps {
*
* @param x Input variable (NUMERIC type)
* @param partitions 1D input with values 0 to numPartitions-1 (INT type)
* @param numPartitions Number of partitions, >= 1
* @param numPartitions Number of partitions, &gt;= 1
*/
public SDVariable[] dynamicPartition(SDVariable x, SDVariable partitions, int numPartitions) {
SDValidation.validateNumerical("dynamicPartition", "x", x);
@ -676,7 +676,7 @@ public class SDBaseOps {
* @param names names May be null. Arrays of names for the output variables.
* @param x Input variable (NUMERIC type)
* @param partitions 1D input with values 0 to numPartitions-1 (INT type)
* @param numPartitions Number of partitions, >= 1
* @param numPartitions Number of partitions, &gt;= 1
*/
public SDVariable[] dynamicPartition(String[] names, SDVariable x, SDVariable partitions,
int numPartitions) {
@ -689,7 +689,7 @@ public class SDBaseOps {
/**
* Dynamically merge the specified input arrays into a single array, using the specified indices<br>
*
* @param indices Indices to use when merging. Must be >= 1, same length as input variables (INT type)
* @param indices Indices to use when merging. Must be &gt;= 1, same length as input variables (INT type)
* @param x Input variables. (NUMERIC type)
* @return output Merged output variable (NUMERIC type)
*/
@ -705,7 +705,7 @@ public class SDBaseOps {
* Dynamically merge the specified input arrays into a single array, using the specified indices<br>
*
* @param name name May be null. Name for the output variable
* @param indices Indices to use when merging. Must be >= 1, same length as input variables (INT type)
* @param indices Indices to use when merging. Must be &gt;= 1, same length as input variables (INT type)
* @param x Input variables. (NUMERIC type)
* @return output Merged output variable (NUMERIC type)
*/
@ -943,7 +943,7 @@ public class SDBaseOps {
}
/**
* Greater than operation: elementwise x > y<br>
* Greater than operation: elementwise x &gt; y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -957,7 +957,7 @@ public class SDBaseOps {
}
/**
* Greater than operation: elementwise x > y<br>
* Greater than operation: elementwise x &gt; y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -973,7 +973,7 @@ public class SDBaseOps {
}
/**
* Greater than operation: elementwise x > y<br>
* Greater than operation: elementwise x &gt; y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -993,7 +993,7 @@ public class SDBaseOps {
}
/**
* Greater than operation: elementwise x > y<br>
* Greater than operation: elementwise x &gt; y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -1015,7 +1015,7 @@ public class SDBaseOps {
}
/**
* Greater than or equals operation: elementwise x >= y<br>
* Greater than or equals operation: elementwise x &gt;= y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -1029,7 +1029,7 @@ public class SDBaseOps {
}
/**
* Greater than or equals operation: elementwise x >= y<br>
* Greater than or equals operation: elementwise x &gt;= y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -1045,7 +1045,7 @@ public class SDBaseOps {
}
/**
* Greater than or equal to operation: elementwise x >= y<br>
* Greater than or equal to operation: elementwise x &gt;= y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -1065,7 +1065,7 @@ public class SDBaseOps {
}
/**
* Greater than or equal to operation: elementwise x >= y<br>
* Greater than or equal to operation: elementwise x &gt;= y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -1232,7 +1232,7 @@ public class SDBaseOps {
}
/**
* Less than operation: elementwise x < y<br>
* Less than operation: elementwise x &lt; y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -1246,7 +1246,7 @@ public class SDBaseOps {
}
/**
* Less than operation: elementwise x < y<br>
* Less than operation: elementwise x &lt; y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -1262,7 +1262,7 @@ public class SDBaseOps {
}
/**
* Less than operation: elementwise x < y<br>
* Less than operation: elementwise x &lt; y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -1282,7 +1282,7 @@ public class SDBaseOps {
}
/**
* Less than operation: elementwise x < y<br>
* Less than operation: elementwise x &lt; y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -1304,7 +1304,7 @@ public class SDBaseOps {
}
/**
* Less than or equals operation: elementwise x <= y<br>
* Less than or equals operation: elementwise x &lt;= y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -1318,7 +1318,7 @@ public class SDBaseOps {
}
/**
* Less than or equals operation: elementwise x <= y<br>
* Less than or equals operation: elementwise x &lt;= y<br>
*
* Return boolean array with values true where satisfied, or false otherwise.<br>
*
@ -1334,7 +1334,7 @@ public class SDBaseOps {
}
/**
* Less than or equal to operation: elementwise x <= y<br>
* Less than or equal to operation: elementwise x &lt;= y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -1354,7 +1354,7 @@ public class SDBaseOps {
}
/**
* Less than or equal to operation: elementwise x <= y<br>
* Less than or equal to operation: elementwise x &lt;= y<br>
* If x and y arrays have equal shape, the output shape is the same as these inputs.<br>
*
* Note: supports broadcasting if x and y have different shapes and are broadcastable.<br>
@ -3590,7 +3590,7 @@ public class SDBaseOps {
/**
* Generate a sequence mask (with values 0 or 1) based on the specified lengths <br>
* Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)<br>
* {@code Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)}<br>
*
* @param lengths Lengths of the sequences (NUMERIC type)
* @param maxLen Maximum sequence length
@ -3604,7 +3604,7 @@ public class SDBaseOps {
/**
* Generate a sequence mask (with values 0 or 1) based on the specified lengths <br>
* Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)<br>
* {@code Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)}<br>
*
* @param name name May be null. Name for the output variable
* @param lengths Lengths of the sequences (NUMERIC type)
@ -3620,7 +3620,7 @@ public class SDBaseOps {
/**
* Generate a sequence mask (with values 0 or 1) based on the specified lengths <br>
* Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)<br>
* {@code Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)}<br>
*
* @param lengths Lengths of the sequences (NUMERIC type)
* @param maxLen Maximum sequence length (INT type)
@ -3635,7 +3635,7 @@ public class SDBaseOps {
/**
* Generate a sequence mask (with values 0 or 1) based on the specified lengths <br>
* Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)<br>
* {@code Specifically, out[i, ..., k, j] = (j < lengths[i, ..., k] ? 1.0 : 0.0)}<br>
*
* @param name name May be null. Name for the output variable
* @param lengths Lengths of the sequences (NUMERIC type)
@ -3761,7 +3761,7 @@ public class SDBaseOps {
* then slice(input, begin=[0,1], size=[2,1] will return:<br>
* [b]<br>
* [e]<br>
* Note that for each dimension i, begin[i] + size[i] <= input.size(i)<br>
* Note that for each dimension i, begin[i] + size[i] &lt;= input.size(i)<br>
*
* @param input input Variable to get subset of (NUMERIC type)
* @param begin Beginning index. Must be same length as rank of input array (Size: AtLeast(min=1))
@ -3783,7 +3783,7 @@ public class SDBaseOps {
* then slice(input, begin=[0,1], size=[2,1] will return:<br>
* [b]<br>
* [e]<br>
* Note that for each dimension i, begin[i] + size[i] <= input.size(i)<br>
* Note that for each dimension i, begin[i] + size[i] &lt;= input.size(i)<br>
*
* @param name name May be null. Name for the output variable
* @param input input Variable to get subset of (NUMERIC type)
@ -3807,7 +3807,7 @@ public class SDBaseOps {
* then slice(input, begin=[0,1], size=[2,1] will return:<br>
* [b]<br>
* [e]<br>
* Note that for each dimension i, begin[i] + size[i] <= input.size(i)<br>
* Note that for each dimension i, begin[i] + size[i] &lt;= input.size(i)<br>
*
* @param input input Variable to get subset of (NUMERIC type)
* @param begin Beginning index. Must be same length as rank of input array (INT type)
@ -3829,7 +3829,7 @@ public class SDBaseOps {
* then slice(input, begin=[0,1], size=[2,1] will return:<br>
* [b]<br>
* [e]<br>
* Note that for each dimension i, begin[i] + size[i] <= input.size(i)<br>
* Note that for each dimension i, begin[i] + size[i] &lt;= input.size(i)<br>
*
* @param name name May be null. Name for the output variable
* @param input input Variable to get subset of (NUMERIC type)

4
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/evaluation/classification/Evaluation.java
View File

@ -176,7 +176,7 @@ public class Evaluation extends BaseEvaluation<Evaluation> {
* Constructor to use for top N accuracy
*
* @param labels Labels for the classes (may be null)
* @param topN Value to use for top N accuracy calculation (<=1: standard accuracy). Note that with top N
* @param topN Value to use for top N accuracy calculation (&lt;=1: standard accuracy). Note that with top N
* accuracy, an example is considered 'correct' if the probability for the true class is one of the
* highest N values
*/
@ -1173,7 +1173,7 @@ public class Evaluation extends BaseEvaluation<Evaluation> {
/**
* False Alarm Rate (FAR) reflects rate of misclassified to classified records
* <a href="http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=1058&context=isw">http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=1058&context=isw</a><br>
* {@link }http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=1058&context=isw}<br>
* Note: value returned will differ depending on number of classes and settings.<br>
* 1. For binary classification, if the positive class is set (via default value of 1, via constructor,
* or via {@link #setBinaryPositiveClass(Integer)}), the returned value will be for the specified positive class

2
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/api/blas/BlasException.java
View File

@ -34,7 +34,7 @@ public class BlasException extends Error {
}
/**
* Principal constructor - error message & error code
* Principal constructor - error message &amp; error code
* @param message the error message to put into the Exception
* @param errorCode the library error number
*/

21
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/api/blas/Lapack.java
View File

@ -28,15 +28,15 @@ public interface Lapack {
* LU decomposiiton of a matrix
* Factorize a matrix A
*
* The matrix A is overridden by the L & U combined.
* The matrix A is overridden by the L &amp; U combined.
* The permutation results are returned directly as a vector. To
* create the permutation matrix use getPFactor method
* To split out the L & U matrix use getLFactor and getUFactor methods
* To split out the L &amp; U matrix use getLFactor and getUFactor methods
*
* getrf = triangular factorization (TRF) of a general matrix (GE)
*
* @param A the input matrix, it will be overwritten with the factors
* @returns Permutation array
* @return Permutation array
* @throws Error - with a message to indicate failure (usu. bad params)
*/
INDArray getrf(INDArray A);
@ -53,7 +53,7 @@ public interface Lapack {
* matrix Q and an upper triangular R matrix
*
* @param A the input matrix, it will be overwritten with the factors
* @param The R array if null R is not returned
* @param R The R array if null R is not returned
* @throws Error - with a message to indicate failure (usu. bad params)
*/
void geqrf(INDArray A, INDArray R);
@ -71,8 +71,7 @@ public interface Lapack {
* lower L ( or upper U ) triangular matrix
*
* @param A the input matrix, it will be overwritten with the factors
* @param whether to return the upper (false) or lower factor
* @returns Permutation array
* @param lower whether to return the upper (false) or lower factor
* @throws Error - with a message to indicate failure (usu. bad params)
*/
void potrf(INDArray A, boolean lower);
@ -122,7 +121,7 @@ public interface Lapack {
*
* @param M - the size of the permutation matrix ( usu. the # rows in factored matrix )
* @param ipiv - the vector returned from a refactoring
* @returned the square permutation matrix - size is the M x M
* @return the square permutation matrix - size is the M x M
*/
INDArray getPFactor(int M, INDArray ipiv);
@ -131,8 +130,8 @@ public interface Lapack {
* extracts the L (lower triangular) matrix from the LU factor result
* L will be the same dimensions as A
*
* @param A - the combined L & U matrices returned from factorization
* @returned the lower triangular with unit diagonal
* @param A - the combined L &amp; U matrices returned from factorization
* @return the lower triangular with unit diagonal
*/
INDArray getLFactor(INDArray A);
@ -141,8 +140,8 @@ public interface Lapack {
* extracts the U (upper triangular) matrix from the LU factor result
* U will be n x n matrix where n = num cols in A
*
* @param A - the combined L & U matrices returned from factorization
* @returned the upper triangular matrix
* @param A - the combined L &amp; U matrices returned from factorization
* @return the upper triangular matrix
*/
INDArray getUFactor(INDArray A);

8
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/api/blas/Level1.java
View File

@ -26,7 +26,7 @@ import org.nd4j.linalg.api.ndarray.INDArray;
public interface Level1 {
/**
* computes a vector-vector dot product.
* @param n
* @param N
* @param alpha
* @param X
* @param Y
@ -65,7 +65,7 @@ public interface Level1 {
/**
* finds the element of a
* vector that has the largest absolute value.
* @param n the length to iterate for
* @param N the length to iterate for
* @param arr the array to get the max
* index for
* @param stride the stride for the array
@ -105,7 +105,7 @@ public interface Level1 {
/**
* computes a vector-scalar product and adds the result to a vector.
* @param n
* @param N
* @param alpha
* @param x
* @param y
@ -115,7 +115,7 @@ public interface Level1 {
/**
* computes a vector-scalar product and adds the result to a vector.
* y = a*x + y
* @param n number of operations
* @param N number of operations
* @param alpha
* @param x X
* @param offsetX offset of first element of X in buffer

9
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/api/buffer/BaseDataBuffer.java
View File

@ -1297,7 +1297,7 @@ public abstract class BaseDataBuffer implements DataBuffer {
if (offset() == 0) {
return wrappedBuffer().asIntBuffer();
} else
return wrappedBuffer().asIntBuffer().position((int) offset());
return (IntBuffer) wrappedBuffer().asIntBuffer().position((int) offset());
}
@Override
@ -1308,7 +1308,7 @@ public abstract class BaseDataBuffer implements DataBuffer {
if (offset() == 0) {
return wrappedBuffer().asLongBuffer();
} else
return wrappedBuffer().asLongBuffer().position((int) offset());
return (LongBuffer) wrappedBuffer().asLongBuffer().position((int) offset());
}
@Override
@ -1319,7 +1319,7 @@ public abstract class BaseDataBuffer implements DataBuffer {
if (offset() == 0) {
return wrappedBuffer().asDoubleBuffer();
} else {
return wrappedBuffer().asDoubleBuffer().position((int) (offset()));
return (DoubleBuffer) wrappedBuffer().asDoubleBuffer().position((int) (offset()));
}
}
@ -1331,7 +1331,8 @@ public abstract class BaseDataBuffer implements DataBuffer {
if (offset() == 0) {
return wrappedBuffer().asFloatBuffer();
} else {
return wrappedBuffer().asFloatBuffer().position((int) (offset()));
return (FloatBuffer) wrappedBuffer().asFloatBuffer()
.position((int) (offset()));
}
}

2
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/checkutil/NDArrayCreationUtil.java
View File

@ -34,7 +34,7 @@ import java.util.*;
public class NDArrayCreationUtil {
private NDArrayCreationUtil() {}
/** Get an array of INDArrays (2d) all with the specified shape. Pair<INDArray,String> returned to aid
/** Get an array of INDArrays (2d) all with the specified shape. {@code Pair<INDArray,String>} returned to aid
* debugging: String contains information on how to reproduce the matrix (i.e., which function, and arguments)
* Each NDArray in the returned array has been obtained by applying an operation such as transpose, tensorAlongDimension,
* etc to an original array.

2
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/dataset/api/DataSet.java
View File

@ -302,7 +302,7 @@ public interface DataSet extends Iterable<org.nd4j.linalg.dataset.DataSet>, Seri
* Get the example metadata, or null if no metadata has been set
*
* @return List of metadata instances
* @see {@link #getExampleMetaData(Class)} for convenience method for types
* {@link #getExampleMetaData(Class)} for convenience method for types
*/
List<Serializable> getExampleMetaData();

2
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/dataset/api/MultiDataSet.java
View File

@ -180,7 +180,7 @@ public interface MultiDataSet extends Serializable {
* Get the example metadata, or null if no metadata has been set
*
* @return List of metadata instances
* @see {@link #getExampleMetaData(Class)} for convenience method for types
* {@link #getExampleMetaData(Class)} for convenience method for types
*/
List<Serializable> getExampleMetaData();

256
cavis-dnn/cavis-dnn-api/src/main/java/org/nd4j/linalg/util/ND4JTestUtils.java
View File

@ -35,140 +35,154 @@ import java.util.*;
public class ND4JTestUtils {
private ND4JTestUtils(){ }
private ND4JTestUtils() {
}
@AllArgsConstructor
@Data
public static class ComparisonResult {
List<Triple<File,File,Boolean>> allResults;
List<Triple<File,File,Boolean>> passed;
List<Triple<File,File,Boolean>> failed;
List<File> skippedDir1;
List<File> skippedDir2;
@AllArgsConstructor
@Data
public static class ComparisonResult {
List<Triple<File, File, Boolean>> allResults;
List<Triple<File, File, Boolean>> passed;
List<Triple<File, File, Boolean>> failed;
List<File> skippedDir1;
List<File> skippedDir2;
}
/**
* A function for use with {@link #validateSerializedArrays(File, File, boolean, BiFunction)}
* using {@code INDArray#equals(Object)}
*/
public static class EqualsFn implements BiFunction<INDArray, INDArray, Boolean> {
@Override
public Boolean apply(INDArray i1, INDArray i2) {
return i1.equals(i2);
}
}
/**
* A function for use with {@link #validateSerializedArrays(File, File, boolean, BiFunction)}
* using {@link INDArray#equalsWithEps(Object, double)}
*/
@AllArgsConstructor
public static class EqualsWithEpsFn implements BiFunction<INDArray, INDArray, Boolean> {
private final double eps;
@Override
public Boolean apply(INDArray i1, INDArray i2) {
return i1.equalsWithEps(i2, eps);
}
}
/**
* Scan the specified directories for matching files (i.e., same path relative to their respective
* root directories) and compare the contents using INDArray.equals (via {@link EqualsFn} Assumes
* the saved files represent INDArrays saved with {@link Nd4j#saveBinary(INDArray, File)}
*
* @param dir1 First directory
* @param dir2 Second directory
* @param recursive Whether to search recursively (i.e., include files in subdirectories
* @return Comparison results
*/
public static ComparisonResult validateSerializedArrays(File dir1, File dir2, boolean recursive)
throws Exception {
return validateSerializedArrays(dir1, dir2, recursive, new EqualsFn());
}
/**
* Scan the specified directories for matching files (i.e., same path relative to their respective
* root directories) and compare the contents using a provided function.<br> Assumes the saved
* files represent INDArrays saved with {@link Nd4j#saveBinary(INDArray, File)}
*
* @param dir1 First directory
* @param dir2 Second directory
* @param recursive Whether to search recursively (i.e., include files in subdirectories
* @return Comparison results
*/
public static ComparisonResult validateSerializedArrays(File dir1, File dir2, boolean recursive,
BiFunction<INDArray, INDArray, Boolean> evalFn) throws Exception {
File[] f1 = FileUtils.listFiles(dir1, null, recursive).toArray(new File[0]);
File[] f2 = FileUtils.listFiles(dir2, null, recursive).toArray(new File[0]);
Preconditions.checkState(f1.length > 0, "No files found for directory 1: %s",
dir1.getAbsolutePath());
Preconditions.checkState(f2.length > 0, "No files found for directory 2: %s",
dir2.getAbsolutePath());
Map<String, File> relativized1 = new HashMap<>();
Map<String, File> relativized2 = new HashMap<>();
URI u = dir1.toURI();
for (File f : f1) {
if (!f.isFile()) {
continue;
}
String relative = u.relativize(f.toURI()).getPath();
relativized1.put(relative, f);
}
/**
* A function for use with {@link #validateSerializedArrays(File, File, boolean, BiFunction)} using {@code INDArray#equals(Object)}
*/
public static class EqualsFn implements BiFunction<INDArray,INDArray,Boolean> {
@Override
public Boolean apply(INDArray i1, INDArray i2) {
return i1.equals(i2);
u = dir2.toURI();
for (File f : f2) {
if (!f.isFile()) {
continue;
}
String relative = u.relativize(f.toURI()).getPath();
relativized2.put(relative, f);
}
/**
* A function for use with {@link #validateSerializedArrays(File, File, boolean, BiFunction)} using {@link INDArray#equalsWithEps(Object, double)}
*/
@AllArgsConstructor
public static class EqualsWithEpsFn implements BiFunction<INDArray,INDArray,Boolean> {
private final double eps;
@Override
public Boolean apply(INDArray i1, INDArray i2) {
return i1.equalsWithEps(i2, eps);
}
List<File> skipped1 = new ArrayList<>();
for (String s : relativized1.keySet()) {
if (!relativized2.containsKey(s)) {
skipped1.add(relativized1.get(s));
}
}
/**
* Scan the specified directories for matching files (i.e., same path relative to their respective root directories)
* and compare the contents using INDArray.equals (via {@link EqualsFn}
* Assumes the saved files represent INDArrays saved with {@link Nd4j#saveBinary(INDArray, File)}
* @param dir1 First directory
* @param dir2 Second directory
* @param recursive Whether to search recursively (i.e., include files in subdirectories
* @return Comparison results
*/
public static ComparisonResult validateSerializedArrays(File dir1, File dir2, boolean recursive) throws Exception {
return validateSerializedArrays(dir1, dir2, recursive, new EqualsFn());
List<File> skipped2 = new ArrayList<>();
for (String s : relativized2.keySet()) {
if (!relativized1.containsKey(s)) {
skipped2.add(relativized1.get(s));
}
}
/**
* Scan the specified directories for matching files (i.e., same path relative to their respective root directories)
* and compare the contents using a provided function.<br>
* Assumes the saved files represent INDArrays saved with {@link Nd4j#saveBinary(INDArray, File)}
* @param dir1 First directory
* @param dir2 Second directory
* @param recursive Whether to search recursively (i.e., include files in subdirectories
* @return Comparison results
*/
public static ComparisonResult validateSerializedArrays(File dir1, File dir2, boolean recursive, BiFunction<INDArray,INDArray,Boolean> evalFn) throws Exception {
File[] f1 = FileUtils.listFiles(dir1, null, recursive).toArray(new File[0]);
File[] f2 = FileUtils.listFiles(dir2, null, recursive).toArray(new File[0]);
List<Triple<File, File, Boolean>> allResults = new ArrayList<>();
List<Triple<File, File, Boolean>> passed = new ArrayList<>();
List<Triple<File, File, Boolean>> failed = new ArrayList<>();
for (Map.Entry<String, File> e : relativized1.entrySet()) {
File file1 = e.getValue();
File file2 = relativized2.get(e.getKey());
Preconditions.checkState(f1.length > 0, "No files found for directory 1: %s", dir1.getAbsolutePath() );
Preconditions.checkState(f2.length > 0, "No files found for directory 2: %s", dir2.getAbsolutePath() );
Map<String,File> relativized1 = new HashMap<>();
Map<String,File> relativized2 = new HashMap<>();
URI u = dir1.toURI();
for(File f : f1){
if(!f.isFile())
continue;
String relative = u.relativize(f.toURI()).getPath();
relativized1.put(relative, f);
if (file2 == null) {
continue;
}
u = dir2.toURI();
for(File f : f2){
if(!f.isFile())
continue;
String relative = u.relativize(f.toURI()).getPath();
relativized2.put(relative, f);
}
List<File> skipped1 = new ArrayList<>();
for(String s : relativized1.keySet()){
if(!relativized2.containsKey(s)){
skipped1.add(relativized1.get(s));
}
}
List<File> skipped2 = new ArrayList<>();
for(String s : relativized2.keySet()){
if(!relativized1.containsKey(s)){
skipped2.add(relativized1.get(s));
}
}
List<Triple<File,File,Boolean>> allResults = new ArrayList<>();
List<Triple<File,File,Boolean>> passed = new ArrayList<>();
List<Triple<File,File,Boolean>> failed = new ArrayList<>();
for(Map.Entry<String,File> e : relativized1.entrySet()){
File file1 = e.getValue();
File file2 = relativized2.get(e.getKey());
if(file2 == null)
continue;
INDArray i1 = Nd4j.readBinary(file1);
INDArray i2 = Nd4j.readBinary(file2);
boolean b = evalFn.apply(i1, i2);
Triple<File,File,Boolean> t = new Triple<>(file1, file2, b);
allResults.add(t);
if(b){
passed.add(t);
} else {
failed.add(t);
}
}
Comparator<Triple<File,File,Boolean>> c = new Comparator<Triple<File, File, Boolean>>() {
@Override
public int compare(Triple<File, File, Boolean> o1, Triple<File, File, Boolean> o2) {
return o1.getFirst().compareTo(o2.getFirst());
}
};
Collections.sort(allResults, c);
Collections.sort(passed, c);
Collections.sort(failed, c);
Collections.sort(skipped1);
Collections.sort(skipped2);
return new ComparisonResult(allResults, passed, failed, skipped1, skipped2);
INDArray i1 = Nd4j.readBinary(file1);
INDArray i2 = Nd4j.readBinary(file2);
boolean b = evalFn.apply(i1, i2);
Triple<File, File, Boolean> t = new Triple<>(file1, file2, b);
allResults.add(t);
if (b) {
passed.add(t);
} else {
failed.add(t);
}
}
Comparator<Triple<File, File, Boolean>> c = new Comparator<Triple<File, File, Boolean>>() {
@Override
public int compare(Triple<File, File, Boolean> o1, Triple<File, File, Boolean> o2) {
return o1.getFirst().compareTo(o2.getFirst());
}
};
Collections.sort(allResults, c);
Collections.sort(passed, c);
Collections.sort(failed, c);
Collections.sort(skipped1);
Collections.sort(skipped2);
return new ComparisonResult(allResults, passed, failed, skipped1, skipped2);
}
}

4
cavis-dnn/cavis-dnn-data/cavis-dnn-data-datasets/src/main/java/org/deeplearning4j/datasets/iterator/impl/EmnistDataSetIterator.java
View File

@ -211,7 +211,7 @@ public class EmnistDataSetIterator extends BaseDatasetIterator {
}
/**
* Get the labels as a List<String>
* Get the labels as a {@code List<String>}
*
* @return Labels
*/
@ -244,7 +244,7 @@ public class EmnistDataSetIterator extends BaseDatasetIterator {
}
/**
* Get the label assignments for the given set as a List<String>
* Get the label assignments for the given set as a {@code List<String>}
*
* @param dataSet DataSet to get the label assignment for
* @return Label assignment and given dataset

159
cavis-dnn/cavis-dnn-data/cavis-dnn-data-datasets/src/main/java/org/deeplearning4j/datasets/iterator/impl/LFWDataSetIterator.java
View File

@ -31,76 +31,109 @@ import java.util.Random;
public class LFWDataSetIterator extends RecordReaderDataSetIterator {
/** Loads subset of images with given imgDim returned by the generator. */
public LFWDataSetIterator(int[] imgDim) {
this(LFWLoader.SUB_NUM_IMAGES, LFWLoader.SUB_NUM_IMAGES, imgDim, LFWLoader.SUB_NUM_LABELS, false,
new ParentPathLabelGenerator(), true, 1, null, new Random(System.currentTimeMillis()));
}
/**
* Loads subset of images with given imgDim returned by the generator.
*/
public LFWDataSetIterator(int[] imgDim) {
this(LFWLoader.SUB_NUM_IMAGES, LFWLoader.SUB_NUM_IMAGES, imgDim, LFWLoader.SUB_NUM_LABELS,
false,
new ParentPathLabelGenerator(), true, 1, null, new Random(System.currentTimeMillis()));
}
/** Loads images with given batchSize, numExamples returned by the generator. */
public LFWDataSetIterator(int batchSize, int numExamples) {
this(batchSize, numExamples, new int[] {LFWLoader.HEIGHT, LFWLoader.WIDTH, LFWLoader.CHANNELS},
LFWLoader.NUM_LABELS, false, LFWLoader.LABEL_PATTERN, true, 1, null,
new Random(System.currentTimeMillis()));
}
/**
* Loads images with given batchSize, numExamples returned by the generator.
*/
public LFWDataSetIterator(int batchSize, int numExamples) {
this(batchSize, numExamples, new int[]{LFWLoader.HEIGHT, LFWLoader.WIDTH, LFWLoader.CHANNELS},
LFWLoader.NUM_LABELS, false, LFWLoader.LABEL_PATTERN, true, 1, null,
new Random(System.currentTimeMillis()));
}
/** Loads images with given batchSize, numExamples, imgDim returned by the generator. */
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim) {
this(batchSize, numExamples, imgDim, LFWLoader.NUM_LABELS, false, LFWLoader.LABEL_PATTERN, true, 1, null,
new Random(System.currentTimeMillis()));
}
/**
* Loads images with given batchSize, numExamples, imgDim returned by the generator.
*/
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim) {
this(batchSize, numExamples, imgDim, LFWLoader.NUM_LABELS, false, LFWLoader.LABEL_PATTERN, true,
1, null,
new Random(System.currentTimeMillis()));
}
/** Loads images with given batchSize, imgDim, useSubset, returned by the generator. */
public LFWDataSetIterator(int batchSize, int[] imgDim, boolean useSubset) {
this(batchSize, useSubset ? LFWLoader.SUB_NUM_IMAGES : LFWLoader.NUM_IMAGES, imgDim,
useSubset ? LFWLoader.SUB_NUM_LABELS : LFWLoader.NUM_LABELS, useSubset, LFWLoader.LABEL_PATTERN,
true, 1, null, new Random(System.currentTimeMillis()));
}
/**
* Loads images with given batchSize, imgDim, useSubset, returned by the generator.
*/
public LFWDataSetIterator(int batchSize, int[] imgDim, boolean useSubset) {
this(batchSize, useSubset ? LFWLoader.SUB_NUM_IMAGES : LFWLoader.NUM_IMAGES, imgDim,
useSubset ? LFWLoader.SUB_NUM_LABELS : LFWLoader.NUM_LABELS, useSubset,
LFWLoader.LABEL_PATTERN,
true, 1, null, new Random(System.currentTimeMillis()));
}
/** Loads images with given batchSize, numExamples, imgDim, train, & splitTrainTest returned by the generator. */
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, boolean train, double splitTrainTest) {
this(batchSize, numExamples, imgDim, LFWLoader.NUM_LABELS, false, LFWLoader.LABEL_PATTERN, train,
splitTrainTest, null, new Random(System.currentTimeMillis()));
}
/**
* Loads images with given batchSize, numExamples, imgDim, train, &amp; splitTrainTest returned
* by the generator.
*/
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, boolean train,
double splitTrainTest) {
this(batchSize, numExamples, imgDim, LFWLoader.NUM_LABELS, false, LFWLoader.LABEL_PATTERN,
train,
splitTrainTest, null, new Random(System.currentTimeMillis()));
}
/** Loads images with given batchSize, numExamples, numLabels, train, & splitTrainTest returned by the generator. */
public LFWDataSetIterator(int batchSize, int numExamples, int numLabels, boolean train, double splitTrainTest) {
this(batchSize, numExamples, new int[] {LFWLoader.HEIGHT, LFWLoader.WIDTH, LFWLoader.CHANNELS}, numLabels,
false, null, train, splitTrainTest, null, new Random(System.currentTimeMillis()));
}
/**
* Loads images with given batchSize, numExamples, numLabels, train, &amp; splitTrainTest
* returned by the generator.
*/
public LFWDataSetIterator(int batchSize, int numExamples, int numLabels, boolean train,
double splitTrainTest) {
this(batchSize, numExamples, new int[]{LFWLoader.HEIGHT, LFWLoader.WIDTH, LFWLoader.CHANNELS},
numLabels,
false, null, train, splitTrainTest, null, new Random(System.currentTimeMillis()));
}
/** Loads images with given batchSize, numExamples, imgDim, numLabels, useSubset, train, splitTrainTest & Random returned by the generator. */
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, int numLabels, boolean useSubset,
boolean train, double splitTrainTest, Random rng) {
this(batchSize, numExamples, imgDim, numLabels, useSubset, LFWLoader.LABEL_PATTERN, train, splitTrainTest, null,
rng);
}
/**
* Loads images with given batchSize, numExamples, imgDim, numLabels, useSubset, train,
* splitTrainTest &amp; Random returned by the generator.
*/
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, int numLabels,
boolean useSubset,
boolean train, double splitTrainTest, Random rng) {
this(batchSize, numExamples, imgDim, numLabels, useSubset, LFWLoader.LABEL_PATTERN, train,
splitTrainTest, null,
rng);
}
/** Loads images with given batchSize, numExamples, imgDim, numLabels, useSubset, train, splitTrainTest & Random returned by the generator. */
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, int numLabels, boolean useSubset,
PathLabelGenerator labelGenerator, boolean train, double splitTrainTest, Random rng) {
this(batchSize, numExamples, imgDim, numLabels, useSubset, labelGenerator, train, splitTrainTest, null, rng);
}
/**
* Loads images with given batchSize, numExamples, imgDim, numLabels, useSubset, train,
* splitTrainTest &amp; Random returned by the generator.
*/
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, int numLabels,
boolean useSubset,
PathLabelGenerator labelGenerator, boolean train, double splitTrainTest, Random rng) {
this(batchSize, numExamples, imgDim, numLabels, useSubset, labelGenerator, train,
splitTrainTest, null, rng);
}
/**
* Create LFW data specific iterator
* @param batchSize the batch size of the examples
* @param numExamples the overall number of examples
* @param imgDim an array of height, width and channels
* @param numLabels the overall number of examples
* @param useSubset use a subset of the LFWDataSet
* @param labelGenerator path label generator to use
* @param train true if use train value
* @param splitTrainTest the percentage to split data for train and remainder goes to test
* @param imageTransform how to transform the image
* @param rng random number to lock in batch shuffling
* */
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, int numLabels, boolean useSubset,
PathLabelGenerator labelGenerator, boolean train, double splitTrainTest,
ImageTransform imageTransform, Random rng) {
super(new LFWLoader(imgDim, imageTransform, useSubset).getRecordReader(batchSize, numExamples, imgDim,
numLabels, labelGenerator, train, splitTrainTest, rng), batchSize, 1, numLabels);
}
/**
* Create LFW data specific iterator
*
* @param batchSize the batch size of the examples
* @param numExamples the overall number of examples
* @param imgDim an array of height, width and channels
* @param numLabels the overall number of examples
* @param useSubset use a subset of the LFWDataSet
* @param labelGenerator path label generator to use
* @param train true if use train value
* @param splitTrainTest the percentage to split data for train and remainder goes to test
* @param imageTransform how to transform the image
* @param rng random number to lock in batch shuffling
*/
public LFWDataSetIterator(int batchSize, int numExamples, int[] imgDim, int numLabels,
boolean useSubset,
PathLabelGenerator labelGenerator, boolean train, double splitTrainTest,
ImageTransform imageTransform, Random rng) {
super(new LFWLoader(imgDim, imageTransform, useSubset).getRecordReader(batchSize, numExamples,
imgDim,
numLabels, labelGenerator, train, splitTrainTest, rng), batchSize, 1, numLabels);
}
}

2
cavis-full/build.gradle
View File

@ -19,7 +19,7 @@ dependencies {
//TODO for the two below.. either platform specific uber jars or a single big one with all platforms
api group: "org.bytedeco", name: "javacpp", version: "1.5.7", classifier: "linux-x86_64"
//api group: "org.bytedeco", name: "javacpp", version: "1.5.7"
api group: 'net.brutex.cavis-native', name: 'cavis-native-lib', version: '1.0.0-SNAPSHOT', classifier: "linux-x86_64-avx2-cpu"
// api group: 'net.brutex.cavis-native', name: 'cavis-native-lib', version: '1.0.0-SNAPSHOT', classifier: "linux-x86_64-avx2-cpu"
//api group: 'net.brutex.cavis-native', name: 'cavis-native-lib', version: '1.0.0-SNAPSHOT'
rootProject.getAllprojects().each { Project sproj ->
if(!sproj.name.equals(name) && !sproj.name.equals("cavis-common-platform")

2
cavis-nd4j/cavis-nd4j-aeron/src/main/java/org/nd4j/aeron/ipc/NDArrayMessage.java
View File

@ -256,7 +256,7 @@ public class NDArrayMessage implements Serializable {
String messageId = UUID.randomUUID().toString();
for (int i = 0; i < ret.length; i++) {
//data: only grab a chunk of the data
ByteBuffer view = wholeBuffer.byteBuffer().asReadOnlyBuffer().position(i * chunkSize);
ByteBuffer view = (ByteBuffer) wholeBuffer.byteBuffer().asReadOnlyBuffer().position(i * chunkSize);
view.limit(Math.min(i * chunkSize + chunkSize, wholeBuffer.capacity()));
view.order(ByteOrder.nativeOrder());
view = view.slice();

4
cavis-nd4j/cavis-nd4j-parameter-server/cavis-nd4j-parameter-server-node/src/main/java/org/nd4j/parameterserver/distributed/transport/RoutedTransport.java
View File

@ -664,8 +664,8 @@ public class RoutedTransport extends BaseTransport {
public static class RemoteConnectionBuilder {
private final Object locker = new Object();
private final AtomicBoolean activated = new AtomicBoolean();
private Object locker = new Object();
private AtomicBoolean activated = new AtomicBoolean();
}
}