2021-02-09 05:16:31 +01:00
/*
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* *
* *
* * This program and the accompanying materials are made available under the
* * terms of the Apache License , Version 2.0 which is available at
* * https : //www.apache.org/licenses/LICENSE-2.0.
* *
* * See the NOTICE file distributed with this work for additional
* * information regarding copyright ownership .
* * Unless required by applicable law or agreed to in writing , software
* * distributed under the License is distributed on an " AS IS " BASIS , WITHOUT
* * WARRANTIES OR CONDITIONS OF ANY KIND , either express or implied . See the
* * License for the specific language governing permissions and limitations
* * under the License .
* *
* * SPDX - License - Identifier : Apache - 2.0
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
*/
2019-06-06 14:21:15 +02:00
2020-03-02 10:49:41 +01:00
# include <system/op_boilerplate.h>
2019-06-06 14:21:15 +02:00
# if NOT_EXCLUDED(OP_strided_slice)
# include <array>
# include <ops/declarable/CustomOperations.h>
# include <helpers/ShapeUtils.h>
# include <helpers/BitwiseUtils.h>
2020-03-02 10:49:41 +01:00
namespace sd {
2019-06-06 14:21:15 +02:00
namespace ops {
constexpr int kShrinkAxis = - 1 , kNewAxis = - 2 ;
struct StridedSliceSparseSpec {
int dims ;
int num_add_axis_after_ellipsis ;
std : : vector < int > * begin_tensor ;
const std : : vector < int > * end_tensor ;
const std : : vector < int > * strides_tensor ;
const int begin_mask , end_mask ;
int ellipsis_mask ;
const int new_axis_mask , shrink_axis_mask ;
} ;
struct StridedSliceDenseSpec {
const int dims ;
int begin_mask ;
int end_mask ;
bool begin_valid ;
bool end_valid ;
std : : vector < int > & begin ;
std : : vector < int > & end ;
std : : vector < int > & strides ;
std : : vector < int > final_shape_gather_indices ;
int shrink_axis_mask ;
public :
bool buildDenseSpec ( StridedSliceSparseSpec & sparse_spec ) {
if ( this - > begin . size ( ) < dims )
this - > begin . resize ( dims ) ;
if ( this - > end . size ( ) < dims )
this - > end . resize ( dims ) ;
if ( this - > strides . size ( ) < dims )
this - > strides . resize ( dims ) ;
this - > begin_mask = 0 ;
this - > end_mask = 0 ;
this - > shrink_axis_mask = 0 ;
{
int full_index = 0 ;
this - > begin_valid = sparse_spec . begin_tensor ! = nullptr ;
this - > end_valid = sparse_spec . end_tensor ! = nullptr ;
for ( int e = 0 ; e < sparse_spec . dims ; e + + ) {
if ( ( 1 < < e ) & sparse_spec . ellipsis_mask ) {
2020-03-02 10:49:41 +01:00
int next_index = sd : : math : : nd4j_min < int > ( this - > dims - ( sparse_spec . dims - e ) + 1 + sparse_spec . num_add_axis_after_ellipsis , this - > dims ) ;
2019-06-06 14:21:15 +02:00
for ( ; full_index < next_index ; full_index + + ) {
// new_axis' aren't real axis so you have to skip
this - > begin [ full_index ] = this - > end [ full_index ] = 0 ;
this - > strides [ full_index ] = 1 ;
this - > begin_mask | = ( 1 < < full_index ) ;
this - > end_mask | = ( 1 < < full_index ) ;
this - > final_shape_gather_indices . push_back ( full_index ) ;
}
} else if ( ( 1 < < e ) & sparse_spec . new_axis_mask ) {
this - > final_shape_gather_indices . emplace_back ( kNewAxis ) ;
} else {
if ( full_index = = this - > begin . size ( ) ) {
nd4j_printf ( " Index out of range: %i out of %i \n " , full_index , this - > dims ) ;
return false ;
}
// Gather slicing spec into appropriate index
if ( sparse_spec . begin_tensor ! = nullptr )
this - > begin [ full_index ] = sparse_spec . begin_tensor - > at ( e ) ;
if ( sparse_spec . end_tensor ! = nullptr )
this - > end [ full_index ] = sparse_spec . end_tensor - > at ( e ) ;
this - > strides [ full_index ] = sparse_spec . strides_tensor - > at ( e ) ;
if ( sparse_spec . begin_mask & ( 1 < < e ) )
this - > begin_mask | = ( 1 < < full_index ) ;
if ( sparse_spec . end_mask & ( 1 < < e ) )
this - > end_mask | = ( 1 < < full_index ) ;
// If shrink, record where to get the dimensionality from (i.e.
// new_axis creates a fake 1 size dimension. Also remember shrink
// axis (now in dense form) so we can ignore dense->end below.
if ( sparse_spec . shrink_axis_mask & ( 1 < < e ) ) {
this - > final_shape_gather_indices . push_back ( kShrinkAxis ) ;
this - > shrink_axis_mask | = ( 1 < < full_index ) ;
} else {
this - > final_shape_gather_indices . push_back ( full_index ) ;
}
full_index + + ;
}
}
}
return true ;
}
} ;
void vectorize ( std : : vector < Nd4jLong > & input_shape ) {
if ( input_shape . size ( ) = = 2 & & input_shape [ 0 ] = = 1 ) {
int v = input_shape [ 1 ] ;
input_shape . clear ( ) ;
input_shape . emplace_back ( v ) ;
}
}
bool _preprocess_strided_slice ( std : : vector < Nd4jLong > * indicesList , std : : vector < Nd4jLong > * final_shape , std : : vector < Nd4jLong > & input_shape , std : : vector < int > & begin , std : : vector < int > & end , std : : vector < int > & strides , int begin_mask , int ellipsis_mask , int end_mask , int new_axis_mask , int shrink_axis_mask , bool * is_identity , bool * is_simple_slice , bool * slice_dim0 ) {
std : : vector < int > preshape ;
bool ellipsis_seen = false ;
StridedSliceSparseSpec sparse_spec = { ( int ) strides . size ( ) ,
0 ,
& begin ,
& end ,
& strides ,
begin_mask ,
end_mask ,
ellipsis_mask ,
new_axis_mask ,
shrink_axis_mask } ;
for ( int i = 0 ; i < sparse_spec . dims ; i + + ) {
if ( ellipsis_seen & & ( ( 1 < < i ) & new_axis_mask ) ! = 0 ) {
sparse_spec . num_add_axis_after_ellipsis + + ;
}
if ( ( 1 < < i ) & ellipsis_mask ) {
ellipsis_seen = true ;
}
}
// If no ellipsis insert one at the end
if ( ! ellipsis_seen ) {
sparse_spec . ellipsis_mask | = ( 1 < < sparse_spec . dims ) ;
sparse_spec . dims + + ; // this effects loop iteration below
}
StridedSliceDenseSpec dense_spec = { ( int ) input_shape . size ( ) , 0 , 0 , false , false , begin , end , strides } ;
if ( ! dense_spec . buildDenseSpec ( sparse_spec ) )
return false ;
//nd4j_printv("Input shape: ", input_shape);
for ( int e = 0 ; e < ( int ) input_shape . size ( ) ; e + + ) {
int begin_idx = begin [ e ] ;
int end_idx = end [ e ] ;
int stride_idx = strides [ e ] ;
int size_idx = input_shape [ e ] ;
bool shrink_i = ( dense_spec . shrink_axis_mask & ( 1 < < e ) ) ;
if ( stride_idx = = 0 ) {
nd4j_printf ( " Stride is 0 at index %i \n " , e ) ;
return false ;
}
if ( size_idx = = - 1 ) {
preshape . emplace_back ( shrink_i ? 1 : - 1 ) ;
continue ;
}
const std : : array < int , 2 > masks = { { dense_spec . begin_mask & ( 1 < < e ) , dense_spec . end_mask & ( 1 < < e ) } } ;
const std : : array < int , 2 > valid_range = { { stride_idx > 0 ? 0 : - 1 , stride_idx > 0 ? size_idx : size_idx - 1 } } ;
auto canonical = [ stride_idx , e , size_idx , masks , valid_range ] ( int x , int c ) {
if ( masks [ c ] ) {
return stride_idx > 0 ? valid_range [ c ] : valid_range [ ( c + 1 ) & 1 ] ;
} else {
int x_fwd = x < 0 ? size_idx + x : x ; // make negative indices positive
return x_fwd < valid_range [ 0 ] ? valid_range [ 0 ] : x_fwd > valid_range [ 1 ] ? valid_range [ 1 ] : x_fwd ;
}
} ;
if ( shrink_i & & stride_idx < = 0 ) {
nd4j_printf ( " StridedSlice: only stride 1 allowed on non-range indexing \n " , e ) ;
return false ;
}
( * is_simple_slice ) & = stride_idx = = 1 ;
const bool begin_and_end_masked = ( begin_mask & ( 1 < < e ) ) & & ( end_mask & ( 1 < < e ) ) ;
if ( dense_spec . begin_valid & & dense_spec . end_valid ) {
if ( shrink_i ) {
int x_fwd = begin_idx < 0 ? size_idx + begin_idx : begin_idx ;
begin_idx = x_fwd ;
end_idx = begin_idx + 1 ;
if ( x_fwd < 0 | | x_fwd > = size_idx ) {
nd4j_printf ( " slice index %i of dimension %i out of bounds. \n " , begin_idx , e ) ;
return false ;
}
} else {
begin_idx = canonical ( begin_idx , 0 ) ;
end_idx = canonical ( end_idx , 1 ) ;
}
} else {
( * is_identity ) & = stride_idx = = 1 & & begin_and_end_masked ;
( * slice_dim0 ) & = ( e = = 0 & & stride_idx = = 1 ) | | begin_and_end_masked ;
}
int interval_length = 1 ;
bool known_interval = false ;
if ( dense_spec . begin_valid & & dense_spec . end_valid ) {
interval_length = end_idx - begin_idx ;
known_interval = true ;
} else if ( shrink_i ) {
interval_length = 1 ;
known_interval = true ;
} else if ( begin_and_end_masked ) {
if ( size_idx > 0 ) {
if ( stride_idx < 0 ) {
interval_length = - size_idx ;
} else {
interval_length = size_idx ;
}
known_interval = true ;
}
}
if ( known_interval ) {
int size_i ;
if ( interval_length = = 0 | | ( ( interval_length < 0 ) ! = ( stride_idx < 0 ) ) ) {
size_i = input_shape . size ( ) = = 2 & & input_shape [ 0 ] = = 1 ? 1 : 0 ;
} else {
size_i = interval_length / stride_idx + ( interval_length % stride_idx ! = 0 ? 1 : 0 ) ;
}
if ( indicesList ! = nullptr ) {
if ( interval_length > 1 ) {
indicesList - > push_back ( begin_idx ) ;
indicesList - > push_back ( end_idx ) ;
indicesList - > push_back ( stride_idx ) ;
// (*indicesList)[3*e] = begin_idx;
// (*indicesList)[3*e+1] = end_idx;
// (*indicesList)[3*e+2] = stride_idx;
}
else if ( interval_length = = 1 ) {
indicesList - > push_back ( begin_idx ) ;
indicesList - > push_back ( begin_idx + 1 ) ;
indicesList - > push_back ( 1 ) ;
// (*indicesList)[3*e] = begin_idx;
// (*indicesList)[3*e+1] = begin_idx + 1;
// (*indicesList)[3*e+2] = 1;
}
}
preshape . emplace_back ( size_i ) ;
} else {
preshape . emplace_back ( - 1 ) ;
}
}
std : : vector < int > postshape ;
//nd4j_printv("Preshape: ", preshape);
final_shape - > clear ( ) ;
for ( auto gather_index : dense_spec . final_shape_gather_indices ) {
if ( gather_index > = 0 ) {
if ( preshape . size ( ) > gather_index )
final_shape - > emplace_back ( preshape . at ( gather_index ) ) ;
else
final_shape - > emplace_back ( 1 ) ;
} else if ( gather_index = = kNewAxis ) {
final_shape - > emplace_back ( 1 ) ;
}
}
//nd4j_printv("Preshape: ", preshape);
//nd4j_printv("Postshape: ", *final_shape);
return true ;
}
CUSTOM_OP_IMPL ( strided_slice , 1 , 1 , false , 0 , 5 ) {
auto x = INPUT_VARIABLE ( 0 ) ;
auto z = OUTPUT_VARIABLE ( 0 ) ;
2021-02-01 06:31:20 +01:00
if ( z - > isEmpty ( ) | | z - > lengthOf ( ) = = 0 ) {
2019-06-06 14:21:15 +02:00
return ND4J_STATUS_OK ;
}
int begin_mask = INT_ARG ( 0 ) ;
int ellipsis_mask = INT_ARG ( 1 ) ;
int end_mask = INT_ARG ( 2 ) ;
int new_axis_mask = INT_ARG ( 3 ) ;
int shrink_axis_mask = INT_ARG ( 4 ) ;
int dim_values = 0 ; //block.getIArguments()->size() - 5;
int delta = 0 ; //dim_values % 3;
int elements = 0 ; //dim_values / 3;
std : : vector < int > begin ;
std : : vector < int > end ;
std : : vector < int > strides ;
bool isLive = false ;
std : : vector < int > args ;
// statically evaluated
if ( block . getIArguments ( ) - > size ( ) > 5 ) {
dim_values = block . getIArguments ( ) - > size ( ) - 5 ;
delta = dim_values % 3 ;
elements = dim_values / 3 ;
for ( int e = 5 ; e < block . getIArguments ( ) - > size ( ) ; e + + )
args . emplace_back ( INT_ARG ( e ) ) ;
REQUIRE_TRUE ( delta = = 0 , 0 , " StridedSlice: Number of Integer arguments should be equal to input rank x 3 = %i, but got %i instead " , ( x - > rankOf ( ) * 3 ) , dim_values ) ;
ShapeUtils : : copyVectorPart ( begin , args , elements , 0 ) ;
ShapeUtils : : copyVectorPart ( end , args , elements , elements ) ;
ShapeUtils : : copyVectorPart ( strides , args , elements , elements * 2 ) ;
} else if ( block . width ( ) > 1 ) {
isLive = true ;
auto v_begin = INPUT_VARIABLE ( 1 ) ;
auto v_end = INPUT_VARIABLE ( 2 ) ;
elements = v_begin - > lengthOf ( ) ;
REQUIRE_TRUE ( v_begin - > lengthOf ( ) = = v_end - > lengthOf ( ) , 0 , " StridedSlice: Length of begin/end should match, but got %i vs %i instead " , ( int ) v_begin - > lengthOf ( ) , ( int ) v_end - > lengthOf ( ) ) ;
REQUIRE_TRUE ( ( v_begin - > rankOf ( ) = = 1 ) & & ( v_begin - > rankOf ( ) = = v_end - > rankOf ( ) ) , 0 , " StridedSlice: Rank of begin and ends should be 1, but %i given instead " , ( int ) v_end - > rankOf ( ) ) ;
for ( int e = 0 ; e < v_begin - > lengthOf ( ) ; e + + )
begin . emplace_back ( v_begin - > e < int > ( e ) ) ;
for ( int e = 0 ; e < v_end - > lengthOf ( ) ; e + + )
end . emplace_back ( v_end - > e < int > ( e ) ) ;
if ( block . width ( ) > 3 ) {
auto v_stride = INPUT_VARIABLE ( 3 ) ;
REQUIRE_TRUE ( v_stride - > lengthOf ( ) = = v_begin - > lengthOf ( ) , 0 , " StridedSlice: Length of begin/end/stride should match, but got %i vs %i vs %i instead " , ( int ) v_begin - > lengthOf ( ) , ( int ) v_end - > lengthOf ( ) , ( int ) v_stride - > lengthOf ( ) ) ;
REQUIRE_TRUE ( ( v_begin - > rankOf ( ) = = v_stride - > rankOf ( ) ) , 0 , " StridedSlice: Rank of begin and ends should be %i, but %i given instead " , ( int ) v_begin - > rankOf ( ) , v_stride - > rankOf ( ) ) ;
for ( int e = 0 ; e < v_stride - > lengthOf ( ) ; e + + )
strides . emplace_back ( v_stride - > e < int > ( e ) ) ;
} else {
for ( int e = 0 ; e < v_begin - > lengthOf ( ) ; e + + )
strides . emplace_back ( 1 ) ;
}
} else {
REQUIRE_TRUE ( false , 0 , " StridedSlice: Can't find begin/end/stride information neither in IArguments or in input arrays " ) ;
}
// validation of begin and start
std : : vector < int > ignoreBegin = BitwiseUtils : : valueBits ( begin_mask ) ;
std : : vector < int > ignoreEnd = BitwiseUtils : : valueBits ( end_mask ) ;
std : : vector < int > addAxes = BitwiseUtils : : valueBits ( new_axis_mask ) ;
std : : vector < int > moveAxes = BitwiseUtils : : valueBits ( shrink_axis_mask ) ;
if ( shrink_axis_mask = = 0 )
for ( int dim = 0 , b = 0 , e = 0 ; dim < x - > rankOf ( ) ; + + dim ) {
if ( moveAxes [ dim ] )
continue ;
if ( b < begin . size ( ) & & ! ignoreBegin [ b ] & & ! addAxes [ dim ] ) {
int first = strides [ b ] > 0 ? begin [ b ] : math : : nd4j_abs < int > ( begin [ b ] ) - 1 ;
REQUIRE_TRUE ( first < = x - > sizeAt ( dim ) , 0 , " StridedSlice: begin index should be <= corresponding dimension of input array, but got end_index = %i for dimension %i! " , begin [ b ] , dim ) ;
}
if ( e < end . size ( ) & & ! ignoreEnd [ e ] & & ! addAxes [ dim ] ) {
int last = strides [ e ] > 0 ? end [ e ] : math : : nd4j_abs < int > ( end [ e ] ) - 1 ;
REQUIRE_TRUE ( last < = x - > sizeAt ( dim ) , 0 , " StridedSlice: end index should be <= corresponding dimension of input array, but got end_index = %i for dimension %i! " , end [ e ] , dim ) ;
}
+ + b ;
+ + e ;
}
std : : vector < Nd4jLong > indices ;
auto input_shape = x - > getShapeAsVector ( ) ;
std : : vector < Nd4jLong > final_shape ;
bool is_identity ;
bool is_simple_slice ;
bool is_dim0 ;
// FIXME: remove this method once we get 1D vectors supported
//vectorize(input_shape);
REQUIRE_TRUE ( _preprocess_strided_slice ( & indices , & final_shape , input_shape , begin , end , strides , begin_mask , ellipsis_mask , end_mask , new_axis_mask , shrink_axis_mask , & is_identity , & is_simple_slice , & is_dim0 ) , 0 , " StridedSlice: shape calculation failed " ) ;
// if(z->lengthOf() == 1 && !z->isEmpty() && (input_shape.size() == 2 && input_shape[0] == 1)) { //(indices.size() == 6) && (indices[2] - indices[0] == 1)) {
// z->assign(x->e<float>(indices[0]));
// }
// else {
2019-06-15 13:34:34 +02:00
if ( indices . size ( ) ) {
2020-03-05 07:56:52 +01:00
Nd4jLong * subArrShapeInfo = nullptr ;
ALLOCATE ( subArrShapeInfo , block . getWorkspace ( ) , shape : : shapeInfoLength ( x - > rankOf ( ) ) , Nd4jLong ) ;
Nd4jLong offset ;
2020-05-09 07:06:14 +02:00
shape : : calcSubArrShapeInfoAndOffset ( indices . data ( ) , x - > shapeInfo ( ) , subArrShapeInfo , offset , true , true ) ;
2020-06-06 14:26:55 +02:00
auto subArrShapeInfoPack = ConstantShapeHelper : : getInstance ( ) . bufferForShapeInfo ( subArrShapeInfo ) ;
2020-03-05 07:56:52 +01:00
NDArray : : prepareSpecialUse ( { z } , { x } ) ;
NativeOpExecutioner : : execTransformAny ( block . launchContext ( ) , sd : : transform : : Assign ,
2020-06-06 14:26:55 +02:00
x - > bufferWithOffset ( offset ) , subArrShapeInfoPack . primary ( ) ,
x - > specialBufferWithOffset ( offset ) , subArrShapeInfoPack . special ( ) ,
2020-03-05 07:56:52 +01:00
z - > buffer ( ) , z - > shapeInfo ( ) , z - > specialBuffer ( ) , z - > specialShapeInfo ( ) ,
nullptr , nullptr , nullptr , true ) ;
NDArray : : registerSpecialUse ( { z } , { x } ) ;
RELEASE ( subArrShapeInfo , block . getWorkspace ( ) ) ;
2019-06-15 13:34:34 +02:00
}
2021-02-01 06:31:20 +01:00
else if ( ! z - > isEmpty ( ) ) {
2019-06-15 13:34:34 +02:00
z - > assign ( x - > e ( 0 ) ) ;
}
2019-06-06 14:21:15 +02:00
return Status : : OK ( ) ;
}
DECLARE_SYN ( stridedslice , strided_slice ) ;
DECLARE_SHAPE_FN ( strided_slice ) {
auto inShape = inputShape - > at ( 0 ) ;
int begin_mask = INT_ARG ( 0 ) ;
int ellipsis_mask = INT_ARG ( 1 ) ;
int end_mask = INT_ARG ( 2 ) ;
int new_axis_mask = INT_ARG ( 3 ) ;
int shrink_axis_mask = INT_ARG ( 4 ) ;
int x_rank = shape : : rank ( inShape ) ;
int dim_values = block . getIArguments ( ) - > size ( ) - 5 ;
int delta = dim_values % 3 ;
int elements = dim_values / 3 ;
std : : vector < int > begin ;
std : : vector < int > end ;
std : : vector < int > strides ;
// if that's live - shape will be resolved in runtime
if ( block . width ( ) > 1 ) {
begin = INPUT_VARIABLE ( 1 ) - > template asVectorT < int > ( ) ;
end = INPUT_VARIABLE ( 2 ) - > template asVectorT < int > ( ) ;
strides = INPUT_VARIABLE ( 3 ) - > template asVectorT < int > ( ) ;
} else if ( dim_values > 0 ) {
int delta2 = dim_values / x_rank ;
std : : vector < int > args ;
for ( int e = 5 ; e < block . getIArguments ( ) - > size ( ) ; e + + )
args . emplace_back ( INT_ARG ( e ) ) ;
2021-02-01 06:31:20 +01:00
// FIXME: probably template required here
2019-06-06 14:21:15 +02:00
ShapeUtils : : copyVectorPart ( begin , args , elements , 0 ) ;
ShapeUtils : : copyVectorPart ( end , args , elements , elements ) ;
ShapeUtils : : copyVectorPart ( strides , args , elements , elements * 2 ) ;
}
REQUIRE_TRUE ( begin . size ( ) > 0 & & end . size ( ) > 0 & & strides . size ( ) > 0 , 0 , " Strided_Slice: empty arguments " ) ;
// validation of begin and start
std : : vector < int > ignoreBegin = BitwiseUtils : : valueBits ( begin_mask ) ;
std : : vector < int > ignoreEnd = BitwiseUtils : : valueBits ( end_mask ) ;
std : : vector < int > addAxes = BitwiseUtils : : valueBits ( new_axis_mask ) ;
std : : vector < int > moveAxes = BitwiseUtils : : valueBits ( shrink_axis_mask ) ;
Merge master to upstream (#7945)
* Shugeo strided slice zeros (#14)
* Modified strided_slice op to properly work with empty-like shapes.
* Fixed test for reduce_mean with empty-like input.
* [WIP] Last merge (#15)
* correct logsoftmax looss (#2)
* Small SameDiff listener fix (#4)
* Various fixes (#6)
* #7839 Fix for asXMatrix and tests
* #7866 EmbeddingSequenceLayer dtype fix + test
* #7856 SameDiff save/load stream methods
* #7859 RegressionEvaluation rank 4 fix + tests + axis configuration
* EvaluationBinary 3d/4d
* More evaluation 3d/4d tests
* #7847 Evaluation empty checks
* Small test ifx
* #7848 Fix median edge case
* Improve DL4J samediff layer tests
* [WIP] FastText wrapper implemented (#8)
* FastText implemented
* Some fixes
* Fix shapes for wordsNearest
* Validation of input vectors
* Fixes
* Fixed test
* Thread tagged
* Some tweaks
* setContextClassLoader for DeallocatorServiceThread
* Numpy format tests (#1)
* Various fixes (#11)
* #7852 SameDiff gather fix
* #7892 SameDiff placeholder to constant conversion
* #7890 validate input rank for MLN/CG init methods
* Fix broken permute shape calculation
* Permute and gather fixes
* Tests
* #7850 LogSumExp fix + test
* Handful of test fixes
* Empty arrays with non-scalar shapes (#10)
* minor rearrangements for lambdas
* empty tensors with non-scalar shapes
* numpy empty tensors with non-scalar shapes
* few more empty tweaks
* Small fixes
* conv3d signature update
* micro fix in batchnorm mkldnn
* Import fixes
* Fix
* MKL-DNN update
* Small fill fix
* fill with empty input + test
* Fixes
* Small error improvement
* Fix
* one special test
* couple of fixes for lstm
* Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone
* Fixes
* FP16
* Unsigned
* BFloat16
* Fill op - empty tweaks
* - couple of fixes for empty arrays construction
- stack updated
* strided slice fix
* one transform test
* provide method for reducing shapeInfo in case of input array is empty
* Fixed reduceAlongDimensions to use empty input properly.
* couple of broadcast tests
* couple of tests broadcast tests + tweak to make them pass
* add check of non-empty to methods producing sub-arrays
* Fixed reshapeC with zeros in shape.
* complete empty check in reduce_... legacy ops
* Concat and cumsum/prod
* Tweak to empty shape inference on import
* add empty check to the rest of reduce legacy ops
* one more test
* correct typo in evalReduceShapeInfoEmpty
* Added tests for reduce_* ops to tests with zero shapes.
* few more tests for empty reductions
* Fixed strided_slice op with empty case and tests.
* one more empty reduction test
* Fixed strided_slice test.
* add empty check to NDArray::reshapei
* infOrMax
* empty min/max with infinity tests
* made unstack working correctly with empty arrays
* few IndexReduce tests + tweaks for empty shapes
* add test for empty concat
* few tests fixed
* Validation fix for reductions on empty shapes
* Reverse fix
* Reduction shape calc fixes
* SameDiff.generateOutputVariable: don't use shape function to determine number of outputs
* Range fix
* - NDArray constructor updated for scalars/empty arrays
- few tests fixed
* More fixes
* Empty creator fixes
* concat fix
* concat fix
* TF import tests: allow 'both all NaN' and 'both all inf' to pass
* Slice, zero fraction, and reshape fixes
* transpose, gather
* Zero fraction
* scalar cast fix
* Empty reduction axis support
* few more tests fixed
* Fixed input checks conforming with TF for concat op and tests.
* few tests fixed
* matmul scalar shape fix
* Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats.
* broadcast bool fix
* few more tests
* few more tests
* correct evalReduceShapeInfoEmpty
* argmax/argmin + tests
* one more empty edge case + one more test
* argmax/argmin/realdiv_bp tweaks
* empty reshape test + fix
* Helper fixes
* Small fixes
* Gather test fix
* Gather test fix
* Small fixes
* reduce scalar zero values
* scalar mean workaround
* Remove debug code
* along dim mean workaround
* one more test
* - equalsTo() tweak for empty arrays
- one more test
* broadcast tweaks
* [WIP] Fixing outstanding issues for NLP (#9)
* Avoid using not-inited objects
* Test fixed.
* Redundant method avoided for models like FastText
* KMeans++ implementation
* KMeans++ implementation
* Disable parallel execution
* KMeans++
* Tests
* Dev branch merge (#16)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Fix some issues on master (#17)
* Fix DataVec test issue
* Fix issue with dl4j SameDiff output layer
* Dtype fix for lambda layers
* #7912 BertIterator dtype fix (use float32 not global default)
* [WIP] Next set of CUDA stuff (#7)
New CUDA implementations and improvements
* bad file
* Dev branch master merge (#23)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* SameDiff ops, TF import and fixes (#24)
* CheckNumerics tests + fixes + misc fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fake quant
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* FakeQuantWithMinMaxArgs
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* CheckNumerics fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix libnd4j ALL_INTS and ALL_FLOATS declaration (uint and bfloat types)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Small fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Javadoc
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Exception tweak
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix for out of scope stack allocated var use
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignores
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignore for known failing test (already logged issue)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Merge upstream to fork (#25)
* Add thousand-separator commas to TotalParams (#7915)
* Add thousand-separator commas to TotalParams
The number of parameters can be quite large, and it would help the reading of the summary printout to have the TotalParams column & values at the bottom have thousand-separator-commas in them.
* Add thousand-separator commas to MultiLayerNetwork
Corresponding change to MultiLayerNetwork
Signed-off-by: Jxtps Jxtps <jxtps435@gmail.com>
* Update contributing and issue/PR templates (#7934)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix link to AdaDelta paper (#7942)
Fix link to AdaDelta paper hosted on matthewzeiler.com
Signed-off-by: Jxtps
* Fixes, and ignores for known/logged failing issues (#7943)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* SameDiff + DL4J/SameDiff: Multiple fixes (#28)
* #7919 HDF5 attribute buffer length fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7909 Arbiter constructor exception ux improvements
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7925 RNN output layer length checks
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Add listener for validating inputs are not incorrectly modified
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Integrate NonInplaceValidationListener into tests
* #7844 DL4J SameDiff fixes for variable minibatch size
* DL4J SameDiff fixes - ensure gradient for input placeholder is available
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Tweaks to ExternalErrorsFunction - use placeholders, make more robust
* Another fix
* More fixes
* More SameDiff/DL4J fixes
* Scope out scalar array creation in BaseScalarOp
* Remove debug code
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] Final dev branch merge (#29)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* [WIP] Multiple dataset iterators (#27)
* Splitting dataset into arbitrary number
* Fixes
* Multiple split of iterator
* Test
* Test
* Some fixes
* signature change
* one more tweak
Signed-off-by: raver119 <raver119@gmail.com>
* one more test for sequential use of DataSetIteratorSplitter
Signed-off-by: raver119 <raver119@gmail.com>
* Fixes
* Fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* minor test fix
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* couple of assertions tweaked
Signed-off-by: raver119 <raver119@gmail.com>
* MDS splitter test :/
Signed-off-by: raver119 <raver119@gmail.com>
* Minor refactoring
* Multi dataset
* Some fixes
* More tests
* Small number of test fixes/improvements (failures on CI) (#31)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] More CUDA stuff (#26)
* initial commit
Signed-off-by: raver119 <raver119@gmail.com>
* LRN BP CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* less memory
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed bug with crop_and_resize op helper.
* get rid of unnecessary index-calculation dunction
Signed-off-by: Yurii <yurii@skymind.io>
* Fixed sort with nth_element cuda-based helper.
* Refactored nth_element.
* Refactored nth_element op and tests.
* Modified usage of dim array with sortTad routine.
* Refactored main routine of helper for non_max_image_suppression op.
* non_max_image_suppression op helper with cuda kernel implementation. Initial revision.
* fix vol2col cuda kernel
* meh
Signed-off-by: raver119 <raver119@gmail.com>
* topK concept
Signed-off-by: raver119 <raver119@gmail.com>
* unsorted topK with scanWitdh of 1
Signed-off-by: raver119 <raver119@gmail.com>
* correct vol2col tests
* sorted/unsorted topK
Signed-off-by: raver119 <raver119@gmail.com>
* implementation and fixing col2im/col2vol
* Corrected usage flags with input/output with reverse op.
* dup is const now
Signed-off-by: raver119 <raver119@gmail.com>
* percentile op
Signed-off-by: raver119 <raver119@gmail.com>
* group tests for mapool2d
Signed-off-by: Yurii <yurii@skymind.io>
* special test for george
Signed-off-by: raver119 <raver119@gmail.com>
* less threads for sortTad
Signed-off-by: raver119 <raver119@gmail.com>
* provide conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* remove auther in sort tad kernel code
Signed-off-by: Yurii <yurii@skymind.io>
* provide depthwise_conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* - max_pooling_with_argmax
- null check for special use
Signed-off-by: raver119 <raver119@gmail.com>
* dts cuda
Signed-off-by: raver119 <raver119@gmail.com>
* provide sconv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* std cuda
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op to conform TF implementation.
* Improved suppression helper.
* provide pooling3d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* more of minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* (bi)dynamic_rnn
Signed-off-by: raver119 <raver119@gmail.com>
* templates init order
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op.
* Added cuda kernel for non_max_suppression.
* CPU sort by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value tests
Signed-off-by: raver119 <raver119@gmail.com>
* Eliminate compiler error with cuda implementation.
* - repaired gradCheck in cuda
- provide conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* missed signature
Signed-off-by: raver119 <raver119@gmail.com>
* provide depthwise_conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* Implementation of lup helper with cuda kernel. Initial commit.
* further work on backprops for convolutions
Signed-off-by: Yurii <yurii@skymind.io>
* CUDA linear sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* CUDA tad sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* start providing of backprop for pooling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* Added atomicAdd for bool datatype.
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition scalar CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* important comment
Signed-off-by: raver119 <raver119@gmail.com>
* fix pooling2d/3d backprop helpers
Signed-off-by: Yurii <yurii@skymind.io>
* Added non-linear test with dynamic_partition.
* Improved test for dynamic_partition.
* dynamic_partition TAD concept
Signed-off-by: raver119 <raver119@gmail.com>
* - dynamic_partition TAD CUDA impl
- dynamic_partition TAD CPU fix
Signed-off-by: raver119 <raver119@gmail.com>
* - rewrite cpu code for usampling2d/3d
- write cuda code for usampling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* dynamic_stitch CUDA vector case
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case impl
Signed-off-by: raver119 <raver119@gmail.com>
* Added tests for dynamic_stitch 3D-4D cases.
* minor tests tweaks
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed type check for dynamic stitch.
* min/max bp
Signed-off-by: raver119 <raver119@gmail.com>
* rewrite code for upsampling2d/3d cpu
Signed-off-by: Yurii <yurii@skymind.io>
* reduce min/max/norm_max bp
Signed-off-by: raver119 <raver119@gmail.com>
* lup implementation. Additional enhancements.
* provide code for upsamling2d/3d backprop
Signed-off-by: Yurii <yurii@skymind.io>
* weightedCrossEntropyWithLogits
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed template math atomicMul for 64bit ints.
* Refactored dynamic_partition_bp op.
* inverseBroadcast fix
Signed-off-by: raver119 <raver119@gmail.com>
* DynamicPartitionBP test datatype fixed.
* - nd4j_atomicMul Windows fix
- cpu/NDArrayLambda.hpp excluded from CUDA
Signed-off-by: raver119 <raver119@gmail.com>
2019-06-27 17:37:04 +02:00
//if (0 == shrink_axis_mask)
if ( false )
2019-06-06 14:21:15 +02:00
for ( int dim = 0 , b = 0 , e = 0 ; dim < x_rank ; + + dim ) {
if ( moveAxes [ dim ] )
continue ;
if ( b < begin . size ( ) & & ! ignoreBegin [ b ] & & ! addAxes [ dim ] ) {
int first = strides [ b ] > 0 ? begin [ b ] : math : : nd4j_abs < int > ( begin [ b ] ) - 1 ;
REQUIRE_TRUE ( first < = inShape [ dim + 1 ] , 0 , " StridedSlice: begin index should be <= corresponding dimension of input array, but got end_index = %i for dimension %i! " , begin [ b ] , dim ) ;
}
if ( e < end . size ( ) & & ! ignoreEnd [ e ] & & ! addAxes [ dim ] ) {
int last = strides [ e ] > 0 ? end [ e ] : math : : nd4j_abs < int > ( end [ e ] ) - 1 ;
REQUIRE_TRUE ( last < = inShape [ dim + 1 ] , 0 , " StridedSlice: end index should be <= corresponding dimension of input array, but got end_index = %i for dimension %i! " , end [ e ] , dim ) ;
}
+ + b ;
+ + e ;
}
std : : vector < Nd4jLong > input_shape ; //(shape::rank(inShape));
auto inputLen = shape : : length ( inShape ) ;
std : : vector < Nd4jLong > shape ;
auto rank = shape : : rank ( inShape ) ;
auto shortShape = shape : : shapeOf ( inShape ) ;
for ( auto e = 0 ; e < rank ; e + + )
input_shape . emplace_back ( shortShape [ e ] ) ;
bool is_identity ;
bool is_simple_slice ;
bool is_dim0 ;
2019-06-15 13:34:34 +02:00
std : : vector < Nd4jLong > indices ;
bool result = _preprocess_strided_slice ( & indices , & shape , input_shape , begin , end , strides , begin_mask , ellipsis_mask , end_mask , new_axis_mask , shrink_axis_mask , & is_identity , & is_simple_slice , & is_dim0 ) ;
if ( indices . size ( ) ) {
2020-06-06 14:26:55 +02:00
auto newShape = ConstantShapeHelper : : getInstance ( ) . createShapeInfo ( ArrayOptions : : dataType ( inShape ) , ' c ' ,
2019-06-15 13:34:34 +02:00
shape ) ;
Merge master to upstream (#7945)
* Shugeo strided slice zeros (#14)
* Modified strided_slice op to properly work with empty-like shapes.
* Fixed test for reduce_mean with empty-like input.
* [WIP] Last merge (#15)
* correct logsoftmax looss (#2)
* Small SameDiff listener fix (#4)
* Various fixes (#6)
* #7839 Fix for asXMatrix and tests
* #7866 EmbeddingSequenceLayer dtype fix + test
* #7856 SameDiff save/load stream methods
* #7859 RegressionEvaluation rank 4 fix + tests + axis configuration
* EvaluationBinary 3d/4d
* More evaluation 3d/4d tests
* #7847 Evaluation empty checks
* Small test ifx
* #7848 Fix median edge case
* Improve DL4J samediff layer tests
* [WIP] FastText wrapper implemented (#8)
* FastText implemented
* Some fixes
* Fix shapes for wordsNearest
* Validation of input vectors
* Fixes
* Fixed test
* Thread tagged
* Some tweaks
* setContextClassLoader for DeallocatorServiceThread
* Numpy format tests (#1)
* Various fixes (#11)
* #7852 SameDiff gather fix
* #7892 SameDiff placeholder to constant conversion
* #7890 validate input rank for MLN/CG init methods
* Fix broken permute shape calculation
* Permute and gather fixes
* Tests
* #7850 LogSumExp fix + test
* Handful of test fixes
* Empty arrays with non-scalar shapes (#10)
* minor rearrangements for lambdas
* empty tensors with non-scalar shapes
* numpy empty tensors with non-scalar shapes
* few more empty tweaks
* Small fixes
* conv3d signature update
* micro fix in batchnorm mkldnn
* Import fixes
* Fix
* MKL-DNN update
* Small fill fix
* fill with empty input + test
* Fixes
* Small error improvement
* Fix
* one special test
* couple of fixes for lstm
* Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone
* Fixes
* FP16
* Unsigned
* BFloat16
* Fill op - empty tweaks
* - couple of fixes for empty arrays construction
- stack updated
* strided slice fix
* one transform test
* provide method for reducing shapeInfo in case of input array is empty
* Fixed reduceAlongDimensions to use empty input properly.
* couple of broadcast tests
* couple of tests broadcast tests + tweak to make them pass
* add check of non-empty to methods producing sub-arrays
* Fixed reshapeC with zeros in shape.
* complete empty check in reduce_... legacy ops
* Concat and cumsum/prod
* Tweak to empty shape inference on import
* add empty check to the rest of reduce legacy ops
* one more test
* correct typo in evalReduceShapeInfoEmpty
* Added tests for reduce_* ops to tests with zero shapes.
* few more tests for empty reductions
* Fixed strided_slice op with empty case and tests.
* one more empty reduction test
* Fixed strided_slice test.
* add empty check to NDArray::reshapei
* infOrMax
* empty min/max with infinity tests
* made unstack working correctly with empty arrays
* few IndexReduce tests + tweaks for empty shapes
* add test for empty concat
* few tests fixed
* Validation fix for reductions on empty shapes
* Reverse fix
* Reduction shape calc fixes
* SameDiff.generateOutputVariable: don't use shape function to determine number of outputs
* Range fix
* - NDArray constructor updated for scalars/empty arrays
- few tests fixed
* More fixes
* Empty creator fixes
* concat fix
* concat fix
* TF import tests: allow 'both all NaN' and 'both all inf' to pass
* Slice, zero fraction, and reshape fixes
* transpose, gather
* Zero fraction
* scalar cast fix
* Empty reduction axis support
* few more tests fixed
* Fixed input checks conforming with TF for concat op and tests.
* few tests fixed
* matmul scalar shape fix
* Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats.
* broadcast bool fix
* few more tests
* few more tests
* correct evalReduceShapeInfoEmpty
* argmax/argmin + tests
* one more empty edge case + one more test
* argmax/argmin/realdiv_bp tweaks
* empty reshape test + fix
* Helper fixes
* Small fixes
* Gather test fix
* Gather test fix
* Small fixes
* reduce scalar zero values
* scalar mean workaround
* Remove debug code
* along dim mean workaround
* one more test
* - equalsTo() tweak for empty arrays
- one more test
* broadcast tweaks
* [WIP] Fixing outstanding issues for NLP (#9)
* Avoid using not-inited objects
* Test fixed.
* Redundant method avoided for models like FastText
* KMeans++ implementation
* KMeans++ implementation
* Disable parallel execution
* KMeans++
* Tests
* Dev branch merge (#16)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Fix some issues on master (#17)
* Fix DataVec test issue
* Fix issue with dl4j SameDiff output layer
* Dtype fix for lambda layers
* #7912 BertIterator dtype fix (use float32 not global default)
* [WIP] Next set of CUDA stuff (#7)
New CUDA implementations and improvements
* bad file
* Dev branch master merge (#23)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* SameDiff ops, TF import and fixes (#24)
* CheckNumerics tests + fixes + misc fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fake quant
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* FakeQuantWithMinMaxArgs
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* CheckNumerics fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix libnd4j ALL_INTS and ALL_FLOATS declaration (uint and bfloat types)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Small fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Javadoc
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Exception tweak
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix for out of scope stack allocated var use
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignores
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignore for known failing test (already logged issue)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Merge upstream to fork (#25)
* Add thousand-separator commas to TotalParams (#7915)
* Add thousand-separator commas to TotalParams
The number of parameters can be quite large, and it would help the reading of the summary printout to have the TotalParams column & values at the bottom have thousand-separator-commas in them.
* Add thousand-separator commas to MultiLayerNetwork
Corresponding change to MultiLayerNetwork
Signed-off-by: Jxtps Jxtps <jxtps435@gmail.com>
* Update contributing and issue/PR templates (#7934)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix link to AdaDelta paper (#7942)
Fix link to AdaDelta paper hosted on matthewzeiler.com
Signed-off-by: Jxtps
* Fixes, and ignores for known/logged failing issues (#7943)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* SameDiff + DL4J/SameDiff: Multiple fixes (#28)
* #7919 HDF5 attribute buffer length fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7909 Arbiter constructor exception ux improvements
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7925 RNN output layer length checks
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Add listener for validating inputs are not incorrectly modified
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Integrate NonInplaceValidationListener into tests
* #7844 DL4J SameDiff fixes for variable minibatch size
* DL4J SameDiff fixes - ensure gradient for input placeholder is available
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Tweaks to ExternalErrorsFunction - use placeholders, make more robust
* Another fix
* More fixes
* More SameDiff/DL4J fixes
* Scope out scalar array creation in BaseScalarOp
* Remove debug code
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] Final dev branch merge (#29)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* [WIP] Multiple dataset iterators (#27)
* Splitting dataset into arbitrary number
* Fixes
* Multiple split of iterator
* Test
* Test
* Some fixes
* signature change
* one more tweak
Signed-off-by: raver119 <raver119@gmail.com>
* one more test for sequential use of DataSetIteratorSplitter
Signed-off-by: raver119 <raver119@gmail.com>
* Fixes
* Fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* minor test fix
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* couple of assertions tweaked
Signed-off-by: raver119 <raver119@gmail.com>
* MDS splitter test :/
Signed-off-by: raver119 <raver119@gmail.com>
* Minor refactoring
* Multi dataset
* Some fixes
* More tests
* Small number of test fixes/improvements (failures on CI) (#31)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] More CUDA stuff (#26)
* initial commit
Signed-off-by: raver119 <raver119@gmail.com>
* LRN BP CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* less memory
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed bug with crop_and_resize op helper.
* get rid of unnecessary index-calculation dunction
Signed-off-by: Yurii <yurii@skymind.io>
* Fixed sort with nth_element cuda-based helper.
* Refactored nth_element.
* Refactored nth_element op and tests.
* Modified usage of dim array with sortTad routine.
* Refactored main routine of helper for non_max_image_suppression op.
* non_max_image_suppression op helper with cuda kernel implementation. Initial revision.
* fix vol2col cuda kernel
* meh
Signed-off-by: raver119 <raver119@gmail.com>
* topK concept
Signed-off-by: raver119 <raver119@gmail.com>
* unsorted topK with scanWitdh of 1
Signed-off-by: raver119 <raver119@gmail.com>
* correct vol2col tests
* sorted/unsorted topK
Signed-off-by: raver119 <raver119@gmail.com>
* implementation and fixing col2im/col2vol
* Corrected usage flags with input/output with reverse op.
* dup is const now
Signed-off-by: raver119 <raver119@gmail.com>
* percentile op
Signed-off-by: raver119 <raver119@gmail.com>
* group tests for mapool2d
Signed-off-by: Yurii <yurii@skymind.io>
* special test for george
Signed-off-by: raver119 <raver119@gmail.com>
* less threads for sortTad
Signed-off-by: raver119 <raver119@gmail.com>
* provide conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* remove auther in sort tad kernel code
Signed-off-by: Yurii <yurii@skymind.io>
* provide depthwise_conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* - max_pooling_with_argmax
- null check for special use
Signed-off-by: raver119 <raver119@gmail.com>
* dts cuda
Signed-off-by: raver119 <raver119@gmail.com>
* provide sconv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* std cuda
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op to conform TF implementation.
* Improved suppression helper.
* provide pooling3d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* more of minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* (bi)dynamic_rnn
Signed-off-by: raver119 <raver119@gmail.com>
* templates init order
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op.
* Added cuda kernel for non_max_suppression.
* CPU sort by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value tests
Signed-off-by: raver119 <raver119@gmail.com>
* Eliminate compiler error with cuda implementation.
* - repaired gradCheck in cuda
- provide conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* missed signature
Signed-off-by: raver119 <raver119@gmail.com>
* provide depthwise_conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* Implementation of lup helper with cuda kernel. Initial commit.
* further work on backprops for convolutions
Signed-off-by: Yurii <yurii@skymind.io>
* CUDA linear sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* CUDA tad sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* start providing of backprop for pooling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* Added atomicAdd for bool datatype.
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition scalar CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* important comment
Signed-off-by: raver119 <raver119@gmail.com>
* fix pooling2d/3d backprop helpers
Signed-off-by: Yurii <yurii@skymind.io>
* Added non-linear test with dynamic_partition.
* Improved test for dynamic_partition.
* dynamic_partition TAD concept
Signed-off-by: raver119 <raver119@gmail.com>
* - dynamic_partition TAD CUDA impl
- dynamic_partition TAD CPU fix
Signed-off-by: raver119 <raver119@gmail.com>
* - rewrite cpu code for usampling2d/3d
- write cuda code for usampling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* dynamic_stitch CUDA vector case
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case impl
Signed-off-by: raver119 <raver119@gmail.com>
* Added tests for dynamic_stitch 3D-4D cases.
* minor tests tweaks
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed type check for dynamic stitch.
* min/max bp
Signed-off-by: raver119 <raver119@gmail.com>
* rewrite code for upsampling2d/3d cpu
Signed-off-by: Yurii <yurii@skymind.io>
* reduce min/max/norm_max bp
Signed-off-by: raver119 <raver119@gmail.com>
* lup implementation. Additional enhancements.
* provide code for upsamling2d/3d backprop
Signed-off-by: Yurii <yurii@skymind.io>
* weightedCrossEntropyWithLogits
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed template math atomicMul for 64bit ints.
* Refactored dynamic_partition_bp op.
* inverseBroadcast fix
Signed-off-by: raver119 <raver119@gmail.com>
* DynamicPartitionBP test datatype fixed.
* - nd4j_atomicMul Windows fix
- cpu/NDArrayLambda.hpp excluded from CUDA
Signed-off-by: raver119 <raver119@gmail.com>
2019-06-27 17:37:04 +02:00
// if (inputLen > 1) {
2020-06-06 14:26:55 +02:00
// newShape = ConstantShapeHelper::getInstance().createShapeInfo(ArrayOptions::dataType(inShape), 'c',
Merge master to upstream (#7945)
* Shugeo strided slice zeros (#14)
* Modified strided_slice op to properly work with empty-like shapes.
* Fixed test for reduce_mean with empty-like input.
* [WIP] Last merge (#15)
* correct logsoftmax looss (#2)
* Small SameDiff listener fix (#4)
* Various fixes (#6)
* #7839 Fix for asXMatrix and tests
* #7866 EmbeddingSequenceLayer dtype fix + test
* #7856 SameDiff save/load stream methods
* #7859 RegressionEvaluation rank 4 fix + tests + axis configuration
* EvaluationBinary 3d/4d
* More evaluation 3d/4d tests
* #7847 Evaluation empty checks
* Small test ifx
* #7848 Fix median edge case
* Improve DL4J samediff layer tests
* [WIP] FastText wrapper implemented (#8)
* FastText implemented
* Some fixes
* Fix shapes for wordsNearest
* Validation of input vectors
* Fixes
* Fixed test
* Thread tagged
* Some tweaks
* setContextClassLoader for DeallocatorServiceThread
* Numpy format tests (#1)
* Various fixes (#11)
* #7852 SameDiff gather fix
* #7892 SameDiff placeholder to constant conversion
* #7890 validate input rank for MLN/CG init methods
* Fix broken permute shape calculation
* Permute and gather fixes
* Tests
* #7850 LogSumExp fix + test
* Handful of test fixes
* Empty arrays with non-scalar shapes (#10)
* minor rearrangements for lambdas
* empty tensors with non-scalar shapes
* numpy empty tensors with non-scalar shapes
* few more empty tweaks
* Small fixes
* conv3d signature update
* micro fix in batchnorm mkldnn
* Import fixes
* Fix
* MKL-DNN update
* Small fill fix
* fill with empty input + test
* Fixes
* Small error improvement
* Fix
* one special test
* couple of fixes for lstm
* Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone
* Fixes
* FP16
* Unsigned
* BFloat16
* Fill op - empty tweaks
* - couple of fixes for empty arrays construction
- stack updated
* strided slice fix
* one transform test
* provide method for reducing shapeInfo in case of input array is empty
* Fixed reduceAlongDimensions to use empty input properly.
* couple of broadcast tests
* couple of tests broadcast tests + tweak to make them pass
* add check of non-empty to methods producing sub-arrays
* Fixed reshapeC with zeros in shape.
* complete empty check in reduce_... legacy ops
* Concat and cumsum/prod
* Tweak to empty shape inference on import
* add empty check to the rest of reduce legacy ops
* one more test
* correct typo in evalReduceShapeInfoEmpty
* Added tests for reduce_* ops to tests with zero shapes.
* few more tests for empty reductions
* Fixed strided_slice op with empty case and tests.
* one more empty reduction test
* Fixed strided_slice test.
* add empty check to NDArray::reshapei
* infOrMax
* empty min/max with infinity tests
* made unstack working correctly with empty arrays
* few IndexReduce tests + tweaks for empty shapes
* add test for empty concat
* few tests fixed
* Validation fix for reductions on empty shapes
* Reverse fix
* Reduction shape calc fixes
* SameDiff.generateOutputVariable: don't use shape function to determine number of outputs
* Range fix
* - NDArray constructor updated for scalars/empty arrays
- few tests fixed
* More fixes
* Empty creator fixes
* concat fix
* concat fix
* TF import tests: allow 'both all NaN' and 'both all inf' to pass
* Slice, zero fraction, and reshape fixes
* transpose, gather
* Zero fraction
* scalar cast fix
* Empty reduction axis support
* few more tests fixed
* Fixed input checks conforming with TF for concat op and tests.
* few tests fixed
* matmul scalar shape fix
* Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats.
* broadcast bool fix
* few more tests
* few more tests
* correct evalReduceShapeInfoEmpty
* argmax/argmin + tests
* one more empty edge case + one more test
* argmax/argmin/realdiv_bp tweaks
* empty reshape test + fix
* Helper fixes
* Small fixes
* Gather test fix
* Gather test fix
* Small fixes
* reduce scalar zero values
* scalar mean workaround
* Remove debug code
* along dim mean workaround
* one more test
* - equalsTo() tweak for empty arrays
- one more test
* broadcast tweaks
* [WIP] Fixing outstanding issues for NLP (#9)
* Avoid using not-inited objects
* Test fixed.
* Redundant method avoided for models like FastText
* KMeans++ implementation
* KMeans++ implementation
* Disable parallel execution
* KMeans++
* Tests
* Dev branch merge (#16)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Fix some issues on master (#17)
* Fix DataVec test issue
* Fix issue with dl4j SameDiff output layer
* Dtype fix for lambda layers
* #7912 BertIterator dtype fix (use float32 not global default)
* [WIP] Next set of CUDA stuff (#7)
New CUDA implementations and improvements
* bad file
* Dev branch master merge (#23)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* SameDiff ops, TF import and fixes (#24)
* CheckNumerics tests + fixes + misc fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fake quant
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* FakeQuantWithMinMaxArgs
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* CheckNumerics fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix libnd4j ALL_INTS and ALL_FLOATS declaration (uint and bfloat types)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Small fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Javadoc
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Exception tweak
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix for out of scope stack allocated var use
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignores
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignore for known failing test (already logged issue)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Merge upstream to fork (#25)
* Add thousand-separator commas to TotalParams (#7915)
* Add thousand-separator commas to TotalParams
The number of parameters can be quite large, and it would help the reading of the summary printout to have the TotalParams column & values at the bottom have thousand-separator-commas in them.
* Add thousand-separator commas to MultiLayerNetwork
Corresponding change to MultiLayerNetwork
Signed-off-by: Jxtps Jxtps <jxtps435@gmail.com>
* Update contributing and issue/PR templates (#7934)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix link to AdaDelta paper (#7942)
Fix link to AdaDelta paper hosted on matthewzeiler.com
Signed-off-by: Jxtps
* Fixes, and ignores for known/logged failing issues (#7943)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* SameDiff + DL4J/SameDiff: Multiple fixes (#28)
* #7919 HDF5 attribute buffer length fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7909 Arbiter constructor exception ux improvements
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7925 RNN output layer length checks
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Add listener for validating inputs are not incorrectly modified
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Integrate NonInplaceValidationListener into tests
* #7844 DL4J SameDiff fixes for variable minibatch size
* DL4J SameDiff fixes - ensure gradient for input placeholder is available
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Tweaks to ExternalErrorsFunction - use placeholders, make more robust
* Another fix
* More fixes
* More SameDiff/DL4J fixes
* Scope out scalar array creation in BaseScalarOp
* Remove debug code
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] Final dev branch merge (#29)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* [WIP] Multiple dataset iterators (#27)
* Splitting dataset into arbitrary number
* Fixes
* Multiple split of iterator
* Test
* Test
* Some fixes
* signature change
* one more tweak
Signed-off-by: raver119 <raver119@gmail.com>
* one more test for sequential use of DataSetIteratorSplitter
Signed-off-by: raver119 <raver119@gmail.com>
* Fixes
* Fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* minor test fix
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* couple of assertions tweaked
Signed-off-by: raver119 <raver119@gmail.com>
* MDS splitter test :/
Signed-off-by: raver119 <raver119@gmail.com>
* Minor refactoring
* Multi dataset
* Some fixes
* More tests
* Small number of test fixes/improvements (failures on CI) (#31)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] More CUDA stuff (#26)
* initial commit
Signed-off-by: raver119 <raver119@gmail.com>
* LRN BP CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* less memory
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed bug with crop_and_resize op helper.
* get rid of unnecessary index-calculation dunction
Signed-off-by: Yurii <yurii@skymind.io>
* Fixed sort with nth_element cuda-based helper.
* Refactored nth_element.
* Refactored nth_element op and tests.
* Modified usage of dim array with sortTad routine.
* Refactored main routine of helper for non_max_image_suppression op.
* non_max_image_suppression op helper with cuda kernel implementation. Initial revision.
* fix vol2col cuda kernel
* meh
Signed-off-by: raver119 <raver119@gmail.com>
* topK concept
Signed-off-by: raver119 <raver119@gmail.com>
* unsorted topK with scanWitdh of 1
Signed-off-by: raver119 <raver119@gmail.com>
* correct vol2col tests
* sorted/unsorted topK
Signed-off-by: raver119 <raver119@gmail.com>
* implementation and fixing col2im/col2vol
* Corrected usage flags with input/output with reverse op.
* dup is const now
Signed-off-by: raver119 <raver119@gmail.com>
* percentile op
Signed-off-by: raver119 <raver119@gmail.com>
* group tests for mapool2d
Signed-off-by: Yurii <yurii@skymind.io>
* special test for george
Signed-off-by: raver119 <raver119@gmail.com>
* less threads for sortTad
Signed-off-by: raver119 <raver119@gmail.com>
* provide conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* remove auther in sort tad kernel code
Signed-off-by: Yurii <yurii@skymind.io>
* provide depthwise_conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* - max_pooling_with_argmax
- null check for special use
Signed-off-by: raver119 <raver119@gmail.com>
* dts cuda
Signed-off-by: raver119 <raver119@gmail.com>
* provide sconv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* std cuda
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op to conform TF implementation.
* Improved suppression helper.
* provide pooling3d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* more of minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* (bi)dynamic_rnn
Signed-off-by: raver119 <raver119@gmail.com>
* templates init order
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op.
* Added cuda kernel for non_max_suppression.
* CPU sort by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value tests
Signed-off-by: raver119 <raver119@gmail.com>
* Eliminate compiler error with cuda implementation.
* - repaired gradCheck in cuda
- provide conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* missed signature
Signed-off-by: raver119 <raver119@gmail.com>
* provide depthwise_conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* Implementation of lup helper with cuda kernel. Initial commit.
* further work on backprops for convolutions
Signed-off-by: Yurii <yurii@skymind.io>
* CUDA linear sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* CUDA tad sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* start providing of backprop for pooling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* Added atomicAdd for bool datatype.
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition scalar CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* important comment
Signed-off-by: raver119 <raver119@gmail.com>
* fix pooling2d/3d backprop helpers
Signed-off-by: Yurii <yurii@skymind.io>
* Added non-linear test with dynamic_partition.
* Improved test for dynamic_partition.
* dynamic_partition TAD concept
Signed-off-by: raver119 <raver119@gmail.com>
* - dynamic_partition TAD CUDA impl
- dynamic_partition TAD CPU fix
Signed-off-by: raver119 <raver119@gmail.com>
* - rewrite cpu code for usampling2d/3d
- write cuda code for usampling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* dynamic_stitch CUDA vector case
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case impl
Signed-off-by: raver119 <raver119@gmail.com>
* Added tests for dynamic_stitch 3D-4D cases.
* minor tests tweaks
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed type check for dynamic stitch.
* min/max bp
Signed-off-by: raver119 <raver119@gmail.com>
* rewrite code for upsampling2d/3d cpu
Signed-off-by: Yurii <yurii@skymind.io>
* reduce min/max/norm_max bp
Signed-off-by: raver119 <raver119@gmail.com>
* lup implementation. Additional enhancements.
* provide code for upsamling2d/3d backprop
Signed-off-by: Yurii <yurii@skymind.io>
* weightedCrossEntropyWithLogits
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed template math atomicMul for 64bit ints.
* Refactored dynamic_partition_bp op.
* inverseBroadcast fix
Signed-off-by: raver119 <raver119@gmail.com>
* DynamicPartitionBP test datatype fixed.
* - nd4j_atomicMul Windows fix
- cpu/NDArrayLambda.hpp excluded from CUDA
Signed-off-by: raver119 <raver119@gmail.com>
2019-06-27 17:37:04 +02:00
// shape);
// } else {
2020-06-06 14:26:55 +02:00
// newShape = ConstantShapeHelper::getInstance().scalarShapeInfo(ArrayOptions::dataType(inShape));
Merge master to upstream (#7945)
* Shugeo strided slice zeros (#14)
* Modified strided_slice op to properly work with empty-like shapes.
* Fixed test for reduce_mean with empty-like input.
* [WIP] Last merge (#15)
* correct logsoftmax looss (#2)
* Small SameDiff listener fix (#4)
* Various fixes (#6)
* #7839 Fix for asXMatrix and tests
* #7866 EmbeddingSequenceLayer dtype fix + test
* #7856 SameDiff save/load stream methods
* #7859 RegressionEvaluation rank 4 fix + tests + axis configuration
* EvaluationBinary 3d/4d
* More evaluation 3d/4d tests
* #7847 Evaluation empty checks
* Small test ifx
* #7848 Fix median edge case
* Improve DL4J samediff layer tests
* [WIP] FastText wrapper implemented (#8)
* FastText implemented
* Some fixes
* Fix shapes for wordsNearest
* Validation of input vectors
* Fixes
* Fixed test
* Thread tagged
* Some tweaks
* setContextClassLoader for DeallocatorServiceThread
* Numpy format tests (#1)
* Various fixes (#11)
* #7852 SameDiff gather fix
* #7892 SameDiff placeholder to constant conversion
* #7890 validate input rank for MLN/CG init methods
* Fix broken permute shape calculation
* Permute and gather fixes
* Tests
* #7850 LogSumExp fix + test
* Handful of test fixes
* Empty arrays with non-scalar shapes (#10)
* minor rearrangements for lambdas
* empty tensors with non-scalar shapes
* numpy empty tensors with non-scalar shapes
* few more empty tweaks
* Small fixes
* conv3d signature update
* micro fix in batchnorm mkldnn
* Import fixes
* Fix
* MKL-DNN update
* Small fill fix
* fill with empty input + test
* Fixes
* Small error improvement
* Fix
* one special test
* couple of fixes for lstm
* Rewrite TFGraphMapper.getNDArrayFromTensor to be maintainable and less error prone
* Fixes
* FP16
* Unsigned
* BFloat16
* Fill op - empty tweaks
* - couple of fixes for empty arrays construction
- stack updated
* strided slice fix
* one transform test
* provide method for reducing shapeInfo in case of input array is empty
* Fixed reduceAlongDimensions to use empty input properly.
* couple of broadcast tests
* couple of tests broadcast tests + tweak to make them pass
* add check of non-empty to methods producing sub-arrays
* Fixed reshapeC with zeros in shape.
* complete empty check in reduce_... legacy ops
* Concat and cumsum/prod
* Tweak to empty shape inference on import
* add empty check to the rest of reduce legacy ops
* one more test
* correct typo in evalReduceShapeInfoEmpty
* Added tests for reduce_* ops to tests with zero shapes.
* few more tests for empty reductions
* Fixed strided_slice op with empty case and tests.
* one more empty reduction test
* Fixed strided_slice test.
* add empty check to NDArray::reshapei
* infOrMax
* empty min/max with infinity tests
* made unstack working correctly with empty arrays
* few IndexReduce tests + tweaks for empty shapes
* add test for empty concat
* few tests fixed
* Validation fix for reductions on empty shapes
* Reverse fix
* Reduction shape calc fixes
* SameDiff.generateOutputVariable: don't use shape function to determine number of outputs
* Range fix
* - NDArray constructor updated for scalars/empty arrays
- few tests fixed
* More fixes
* Empty creator fixes
* concat fix
* concat fix
* TF import tests: allow 'both all NaN' and 'both all inf' to pass
* Slice, zero fraction, and reshape fixes
* transpose, gather
* Zero fraction
* scalar cast fix
* Empty reduction axis support
* few more tests fixed
* Fixed input checks conforming with TF for concat op and tests.
* few tests fixed
* matmul scalar shape fix
* Fixed checkout for data type and scalarity with concat to allow non-empty scalars with vector concats.
* broadcast bool fix
* few more tests
* few more tests
* correct evalReduceShapeInfoEmpty
* argmax/argmin + tests
* one more empty edge case + one more test
* argmax/argmin/realdiv_bp tweaks
* empty reshape test + fix
* Helper fixes
* Small fixes
* Gather test fix
* Gather test fix
* Small fixes
* reduce scalar zero values
* scalar mean workaround
* Remove debug code
* along dim mean workaround
* one more test
* - equalsTo() tweak for empty arrays
- one more test
* broadcast tweaks
* [WIP] Fixing outstanding issues for NLP (#9)
* Avoid using not-inited objects
* Test fixed.
* Redundant method avoided for models like FastText
* KMeans++ implementation
* KMeans++ implementation
* Disable parallel execution
* KMeans++
* Tests
* Dev branch merge (#16)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Fix some issues on master (#17)
* Fix DataVec test issue
* Fix issue with dl4j SameDiff output layer
* Dtype fix for lambda layers
* #7912 BertIterator dtype fix (use float32 not global default)
* [WIP] Next set of CUDA stuff (#7)
New CUDA implementations and improvements
* bad file
* Dev branch master merge (#23)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* SameDiff ops, TF import and fixes (#24)
* CheckNumerics tests + fixes + misc fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fake quant
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fixes
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* FakeQuantWithMinMaxArgs
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* CheckNumerics fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix libnd4j ALL_INTS and ALL_FLOATS declaration (uint and bfloat types)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Small fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Javadoc
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Exception tweak
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix for out of scope stack allocated var use
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignores
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Ignore for known failing test (already logged issue)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Merge upstream to fork (#25)
* Add thousand-separator commas to TotalParams (#7915)
* Add thousand-separator commas to TotalParams
The number of parameters can be quite large, and it would help the reading of the summary printout to have the TotalParams column & values at the bottom have thousand-separator-commas in them.
* Add thousand-separator commas to MultiLayerNetwork
Corresponding change to MultiLayerNetwork
Signed-off-by: Jxtps Jxtps <jxtps435@gmail.com>
* Update contributing and issue/PR templates (#7934)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Fix link to AdaDelta paper (#7942)
Fix link to AdaDelta paper hosted on matthewzeiler.com
Signed-off-by: Jxtps
* Fixes, and ignores for known/logged failing issues (#7943)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* SameDiff + DL4J/SameDiff: Multiple fixes (#28)
* #7919 HDF5 attribute buffer length fix
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7909 Arbiter constructor exception ux improvements
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7925 RNN output layer length checks
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Add listener for validating inputs are not incorrectly modified
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* #7939 Integrate NonInplaceValidationListener into tests
* #7844 DL4J SameDiff fixes for variable minibatch size
* DL4J SameDiff fixes - ensure gradient for input placeholder is available
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* Tweaks to ExternalErrorsFunction - use placeholders, make more robust
* Another fix
* More fixes
* More SameDiff/DL4J fixes
* Scope out scalar array creation in BaseScalarOp
* Remove debug code
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] Final dev branch merge (#29)
* SameDiff: convertDataType and gradient check util improvements (#12)
* GradCheck util improvements
* StopGradient constructor + test
* SameDiff: Add datatype conversion
* Javadoc and add DataType.isNumerical()
* Small fix
* Fix SameDiff TF import test cases intermediate naming (workaround for bad default)
* TFGraphTestAllHelper: check intermediates in execution order
* Add missing debug listener
* [WIP] lstmBlock fix + other changes (#13)
- fixes lstmBlock issue
- changes NDArray method reshape(), permute(), transpose() by making them return instance instead of pointer
- CheckNumerics op
- fixes for ReduceBool IsInfOrNan & IsFinite
* Small test fix
* CheckNumerics op wrapper
* Compatibility of deserialization (#18)
Signed-off-by: Alexander Stoyakin <alexander.stoyakin@gmail.com>
* SameDiff: add activation gradient checking support for debugging (#19)
* SameDiff gradient checker: first pass on activation gradient checks
* Fixes + tests for activation gradient checking
* Javadoc
* [WIP] Some nd4j data type corrections (#20)
* Adjust data type
* Set correct Data type.
* Size of proper data type.
* fix averaged cpu load (#22)
* [WIP] Multiple dataset iterators (#27)
* Splitting dataset into arbitrary number
* Fixes
* Multiple split of iterator
* Test
* Test
* Some fixes
* signature change
* one more tweak
Signed-off-by: raver119 <raver119@gmail.com>
* one more test for sequential use of DataSetIteratorSplitter
Signed-off-by: raver119 <raver119@gmail.com>
* Fixes
* Fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* one more test for Alexander
Signed-off-by: raver119 <raver119@gmail.com>
* minor test fix
Signed-off-by: raver119 <raver119@gmail.com>
* Some fixes
* Some fixes
* couple of assertions tweaked
Signed-off-by: raver119 <raver119@gmail.com>
* MDS splitter test :/
Signed-off-by: raver119 <raver119@gmail.com>
* Minor refactoring
* Multi dataset
* Some fixes
* More tests
* Small number of test fixes/improvements (failures on CI) (#31)
Signed-off-by: AlexDBlack <blacka101@gmail.com>
* [WIP] More CUDA stuff (#26)
* initial commit
Signed-off-by: raver119 <raver119@gmail.com>
* LRN BP CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* less memory
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed bug with crop_and_resize op helper.
* get rid of unnecessary index-calculation dunction
Signed-off-by: Yurii <yurii@skymind.io>
* Fixed sort with nth_element cuda-based helper.
* Refactored nth_element.
* Refactored nth_element op and tests.
* Modified usage of dim array with sortTad routine.
* Refactored main routine of helper for non_max_image_suppression op.
* non_max_image_suppression op helper with cuda kernel implementation. Initial revision.
* fix vol2col cuda kernel
* meh
Signed-off-by: raver119 <raver119@gmail.com>
* topK concept
Signed-off-by: raver119 <raver119@gmail.com>
* unsorted topK with scanWitdh of 1
Signed-off-by: raver119 <raver119@gmail.com>
* correct vol2col tests
* sorted/unsorted topK
Signed-off-by: raver119 <raver119@gmail.com>
* implementation and fixing col2im/col2vol
* Corrected usage flags with input/output with reverse op.
* dup is const now
Signed-off-by: raver119 <raver119@gmail.com>
* percentile op
Signed-off-by: raver119 <raver119@gmail.com>
* group tests for mapool2d
Signed-off-by: Yurii <yurii@skymind.io>
* special test for george
Signed-off-by: raver119 <raver119@gmail.com>
* less threads for sortTad
Signed-off-by: raver119 <raver119@gmail.com>
* provide conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* remove auther in sort tad kernel code
Signed-off-by: Yurii <yurii@skymind.io>
* provide depthwise_conv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* - max_pooling_with_argmax
- null check for special use
Signed-off-by: raver119 <raver119@gmail.com>
* dts cuda
Signed-off-by: raver119 <raver119@gmail.com>
* provide sconv2d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* std cuda
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op to conform TF implementation.
* Improved suppression helper.
* provide pooling3d for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* more of minor lstm rearrangements
Signed-off-by: raver119 <raver119@gmail.com>
* (bi)dynamic_rnn
Signed-off-by: raver119 <raver119@gmail.com>
* templates init order
Signed-off-by: raver119 <raver119@gmail.com>
* Refactored non_max_suppression op.
* Added cuda kernel for non_max_suppression.
* CPU sort by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value
Signed-off-by: raver119 <raver119@gmail.com>
* CPU sort TAD by key/value tests
Signed-off-by: raver119 <raver119@gmail.com>
* Eliminate compiler error with cuda implementation.
* - repaired gradCheck in cuda
- provide conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* missed signature
Signed-off-by: raver119 <raver119@gmail.com>
* provide depthwise_conv2d_bp for cuda
Signed-off-by: Yurii <yurii@skymind.io>
* Implementation of lup helper with cuda kernel. Initial commit.
* further work on backprops for convolutions
Signed-off-by: Yurii <yurii@skymind.io>
* CUDA linear sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* CUDA tad sort by key/val
Signed-off-by: raver119 <raver119@gmail.com>
* start providing of backprop for pooling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* Added atomicAdd for bool datatype.
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic partition scalar CUDA
Signed-off-by: raver119 <raver119@gmail.com>
* important comment
Signed-off-by: raver119 <raver119@gmail.com>
* fix pooling2d/3d backprop helpers
Signed-off-by: Yurii <yurii@skymind.io>
* Added non-linear test with dynamic_partition.
* Improved test for dynamic_partition.
* dynamic_partition TAD concept
Signed-off-by: raver119 <raver119@gmail.com>
* - dynamic_partition TAD CUDA impl
- dynamic_partition TAD CPU fix
Signed-off-by: raver119 <raver119@gmail.com>
* - rewrite cpu code for usampling2d/3d
- write cuda code for usampling2d/3d
Signed-off-by: Yurii <yurii@skymind.io>
* dynamic_stitch CUDA vector case
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case concept
Signed-off-by: raver119 <raver119@gmail.com>
* dynamic_stitch CUDA TAD case impl
Signed-off-by: raver119 <raver119@gmail.com>
* Added tests for dynamic_stitch 3D-4D cases.
* minor tests tweaks
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed type check for dynamic stitch.
* min/max bp
Signed-off-by: raver119 <raver119@gmail.com>
* rewrite code for upsampling2d/3d cpu
Signed-off-by: Yurii <yurii@skymind.io>
* reduce min/max/norm_max bp
Signed-off-by: raver119 <raver119@gmail.com>
* lup implementation. Additional enhancements.
* provide code for upsamling2d/3d backprop
Signed-off-by: Yurii <yurii@skymind.io>
* weightedCrossEntropyWithLogits
Signed-off-by: raver119 <raver119@gmail.com>
* Fixed template math atomicMul for 64bit ints.
* Refactored dynamic_partition_bp op.
* inverseBroadcast fix
Signed-off-by: raver119 <raver119@gmail.com>
* DynamicPartitionBP test datatype fixed.
* - nd4j_atomicMul Windows fix
- cpu/NDArrayLambda.hpp excluded from CUDA
Signed-off-by: raver119 <raver119@gmail.com>
2019-06-27 17:37:04 +02:00
// }
2020-05-09 07:06:14 +02:00
return SHAPELIST ( newShape ) ;
}
2019-06-06 14:21:15 +02:00
2020-06-06 14:26:55 +02:00
return SHAPELIST ( ConstantShapeHelper : : getInstance ( ) . emptyShapeInfo ( ArrayOptions : : dataType ( inShape ) ) ) ;
2019-06-06 14:21:15 +02:00
}
CUSTOM_OP_IMPL ( strided_slice_bp , 2 , 1 , false , 0 , 5 ) {
auto x = INPUT_VARIABLE ( 0 ) ;
auto epsNext = INPUT_VARIABLE ( 1 ) ;
auto output = OUTPUT_VARIABLE ( 0 ) ;
int begin_mask = INT_ARG ( 0 ) ;
int ellipsis_mask = INT_ARG ( 1 ) ;
int end_mask = INT_ARG ( 2 ) ;
int new_axis_mask = INT_ARG ( 3 ) ;
int shrink_axis_mask = INT_ARG ( 4 ) ;
int dim_values = 0 ; //block.getIArguments()->size() - 5;
int delta = 0 ; //dim_values % 3;
int elements = 0 ; //dim_values / 3;
std : : vector < int > begin ;
std : : vector < int > end ;
std : : vector < int > strides ;
bool isLive = false ;
std : : vector < int > args ;
// statically evaluated
if ( block . getIArguments ( ) - > size ( ) > 5 ) {
dim_values = block . getIArguments ( ) - > size ( ) - 5 ;
delta = dim_values % 3 ;
elements = dim_values / 3 ;
for ( int e = 5 ; e < block . getIArguments ( ) - > size ( ) ; e + + )
args . emplace_back ( INT_ARG ( e ) ) ;
REQUIRE_TRUE ( delta = = 0 , 0 , " StridedSliceBP: Number of Integer arguments should be equal to input rank x 3 = %i, but got %i instead " , ( x - > rankOf ( ) * 3 ) , dim_values ) ;
ShapeUtils : : copyVectorPart ( begin , args , elements , 0 ) ;
ShapeUtils : : copyVectorPart ( end , args , elements , elements ) ;
ShapeUtils : : copyVectorPart ( strides , args , elements , elements * 2 ) ;
} else if ( block . width ( ) > = 3 ) {
isLive = true ;
auto v_begin = INPUT_VARIABLE ( 2 ) ;
auto v_end = INPUT_VARIABLE ( 3 ) ;
elements = v_begin - > lengthOf ( ) ;
REQUIRE_TRUE ( v_begin - > lengthOf ( ) = = v_end - > lengthOf ( ) , 0 , " StridedSliceBP: Length of begin/end should match, but got %i vs %i instead " , ( int ) v_begin - > lengthOf ( ) , ( int ) v_end - > lengthOf ( ) ) ;
for ( int e = 0 ; e < v_begin - > lengthOf ( ) ; e + + )
begin . emplace_back ( v_begin - > e < int > ( e ) ) ;
for ( int e = 0 ; e < v_end - > lengthOf ( ) ; e + + )
end . emplace_back ( v_end - > e < int > ( e ) ) ;
if ( block . width ( ) > = 4 ) {
auto v_stride = INPUT_VARIABLE ( 4 ) ;
REQUIRE_TRUE ( v_stride - > lengthOf ( ) = = v_begin - > lengthOf ( ) , 0 , " StridedSliceBP: Length of begin/end/stride should match, but got %i vs %i vs %i instead " , ( int ) v_begin - > lengthOf ( ) , ( int ) v_end - > lengthOf ( ) , ( int ) v_stride - > lengthOf ( ) ) ;
for ( int e = 0 ; e < v_stride - > lengthOf ( ) ; e + + )
strides . emplace_back ( v_stride - > e < int > ( e ) ) ;
} else {
for ( int e = 0 ; e < v_begin - > lengthOf ( ) ; e + + )
strides . emplace_back ( 1 ) ;
}
} else {
REQUIRE_TRUE ( false , 0 , " StridedSliceBP: Can't find begin/end/stride information neither in IArguments or in input arrays " ) ;
}
// validation of begin and start
std : : vector < int > ignoreBegin = BitwiseUtils : : valueBits ( begin_mask ) ;
std : : vector < int > ignoreEnd = BitwiseUtils : : valueBits ( end_mask ) ;
std : : vector < int > addAxes = BitwiseUtils : : valueBits ( new_axis_mask ) ;
std : : vector < int > moveAxes = BitwiseUtils : : valueBits ( shrink_axis_mask ) ;
for ( int dim = 0 , b = 0 , e = 0 ; dim < x - > rankOf ( ) ; + + dim ) {
if ( moveAxes [ dim ] )
continue ;
if ( b < begin . size ( ) & & ! ignoreBegin [ b ] & & ! addAxes [ dim ] ) {
int first = strides [ b ] > 0 ? begin [ b ] : math : : nd4j_abs < int > ( begin [ b ] ) - 1 ;
REQUIRE_TRUE ( first < = x - > sizeAt ( dim ) , 0 , " StridedSlice: begin index should be <= corresponding dimension of input array, but got end_index = %i for dimension %i! " , begin [ b ] , dim ) ;
}
if ( e < end . size ( ) & & ! ignoreEnd [ e ] & & ! addAxes [ dim ] ) {
int last = strides [ e ] > 0 ? end [ e ] : math : : nd4j_abs < int > ( end [ e ] ) - 1 ;
REQUIRE_TRUE ( last < = x - > sizeAt ( dim ) , 0 , " StridedSlice: end index should be <= corresponding dimension of input array, but got end_index = %i for dimension %i! " , end [ e ] , dim ) ;
}
+ + b ;
+ + e ;
}
auto input_shape = x - > getShapeAsVector ( ) ;
std : : vector < Nd4jLong > indices ;
std : : vector < Nd4jLong > final_shape ;
bool is_identity ;
bool is_simple_slice ;
bool is_dim0 ;
// FIXME: remove this method once we get 1D vectors supported
vectorize ( input_shape ) ;
REQUIRE_TRUE ( _preprocess_strided_slice ( & indices , & final_shape , input_shape , begin , end , strides , begin_mask , ellipsis_mask , end_mask , new_axis_mask , shrink_axis_mask , & is_identity , & is_simple_slice , & is_dim0 ) , 0 , " StridedSliceBP: shape calculation failed " ) ;
2019-11-07 11:44:02 +01:00
//REQUIRE_TRUE(epsNext->isSameShape(final_shape), 0, "StridedSlice_bp: gradOut shape should be equals to output from strided_slice op.");
//Zero output array, so unused elements have 0 gradient
output - > nullify ( ) ;
//
// the first case: only for scalar gradient step
if ( epsNext - > lengthOf ( ) = = 1 & & ( indices . size ( ) = = 3 & & ( indices [ 1 ] - indices [ 0 ] ) = = 1 | | ( indices [ 2 ] - indices [ 0 ] = = 1 ) ) ) {
2019-06-06 14:21:15 +02:00
output - > p ( indices [ 0 ] , * epsNext ) ;
}
2019-11-07 11:44:02 +01:00
else { // else for other cases
2019-06-06 14:21:15 +02:00
auto sub = ( * output ) ( indices , true , true ) ;
sub . assign ( epsNext ) ;
}
return Status : : OK ( ) ;
}
DECLARE_SHAPE_FN ( strided_slice_bp ) {
auto inShape = inputShape - > at ( 0 ) ;
Nd4jLong * newShape ;
COPY_SHAPE ( inShape , newShape ) ;
2019-07-24 16:11:29 +02:00
return SHAPELIST ( CONSTANT ( newShape ) ) ;
2019-06-06 14:21:15 +02:00
}
DECLARE_TYPES ( strided_slice ) {
getOpDescriptor ( )
2020-03-02 10:49:41 +01:00
- > setAllowedInputTypes ( sd : : DataType : : ANY )
2019-06-06 14:21:15 +02:00
- > setSameMode ( true ) ;
}
DECLARE_TYPES ( strided_slice_bp ) {
getOpDescriptor ( )
2020-03-02 10:49:41 +01:00
- > setAllowedInputTypes ( sd : : DataType : : ANY )
2019-06-06 14:21:15 +02:00
- > setAllowedOutputTypes ( { ALL_FLOATS } ) ;
}
}
}
# endif