saymrwulf/pytorch
1
0
Fork 0
mirror of https://github.com/saymrwulf/pytorch.git synced 2026-05-15 21:00:47 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
pytorch/caffe2/python/operator_test
History
Alexander Sidorov 64b0039ef9 rnn_cell_test: make it determinitistic and speed up 
Summary:
In this PR I do the following:

1. split lstm_test_main into several tests for LSTM, MiLSTM and various Norm based versions
2. instead of looping over various gradient / optimization parameters now they are random inputs through hypothesis.
3.  These change make the test faster and we can avoid limiting number of examples
4. Fix a minor bug with gradient checker in RNN unroll test running twice
5. Generate seed for numpy in hypothesis. This make hypothesis avoid having fluky tests

Also note that Norm tests sometimes fail. I haven't looked into it much, it could be just precision issues. New test split should help identify these issues.
Closes https://github.com/caffe2/caffe2/pull/1678

Reviewed By: pietern

Differential Revision: D6657076

Pulled By: salexspb

fbshipit-source-id: 9f59c71ccd2c818156e9d2424c3423d450b8c8e2
2018-01-04 15:00:42 -08:00
..
activation_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
adagrad_test.py Resuppress adagrad health checks 2017-12-14 16:34:41 -08:00
adam_test.py Ensure indices list in sparse optimizer tests is unique 2018-01-03 12:19:14 -08:00
apmeter_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
assert_test.py Implement AssertOp 2017-12-06 17:18:52 -08:00
atomic_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
batch_box_cox_test.py Faster BatchBoxCox Operator using MKL 2017-11-03 12:04:03 -07:00
batch_sparse_to_dense_op_test.py Add batch-based/row-based sparse from/to dense operator 2017-11-16 00:59:21 -08:00
blobs_queue_db_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
boolean_mask_test.py repeat sequence mask for data dims 2017-10-18 18:08:08 -07:00
boolean_unmask_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
cast_op_test.py cast op for empty batch 2017-11-16 12:20:20 -08:00
channel_shuffle_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
checkpoint_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
clip_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
concat_split_op_test.py Allow negative indices in Concat/Split ops 2017-11-13 18:32:24 -08:00
conditional_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
conv_test.py Always check cuDNN support in test_convolution_gradients 2017-12-04 08:50:39 -08:00
conv_transpose_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
copy_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
cosine_embedding_criterion_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
counter_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
crf_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
cross_entropy_ops_test.py WeightedSigmoidCrossEntropyWithLogits 2017-10-16 17:34:38 -07:00
cudnn_recurrent_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
dataset_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
deform_conv_test.py Deformable convolution operator in Caffe2 2017-09-26 16:20:31 -07:00
distance_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
dropout_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
duplicate_operands_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
elementwise_linear_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
elementwise_logical_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
elementwise_op_broadcast_test.py Add support for doing broadcast with single elem dimensions at both ends 2017-11-01 18:33:11 -07:00
elementwise_ops_test.py Don't run CUDA tests for ops without CUDA implementation 2017-11-08 10:28:02 -08:00
emptysample_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
extend_tensor_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
fc_operator_test.py Fix gpu test for FCTransposed 2017-12-13 15:48:18 -08:00
filler_ops_test.py GPU fallback for LengthsRangeFill Op 2017-12-19 15:42:13 -08:00
find_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
flatten_op_test.py Enable Flatten operator to take an arbitrary axis arguemnt 2017-10-13 12:28:22 -07:00
flexible_top_k_test.py add flexible topK op 2017-11-09 16:48:14 -08:00
gather_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
gather_ranges_op_test.py Don't run CUDA tests for ops without CUDA implementation 2017-11-08 10:28:02 -08:00
given_tensor_fill_op_test.py Add "dtype" parameter for GivenTensorOp 2017-10-23 16:06:37 -07:00
glu_op_test.py Supress hypothesis health check in glu_op_test.py 2017-12-03 22:51:46 -08:00
group_conv_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
gru_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
hsm_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
im2col_col2im_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
image_input_op_test.py Py3 test fixes 2017-12-05 10:34:41 -08:00
index_hash_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
index_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
instance_norm_test.py Fix instance_norm_test.test_instance_norm_model_helper 2017-11-22 15:05:29 -08:00
layer_norm_op_test.py update transformer code for layer_norm() API change 2017-12-01 20:19:31 -08:00
leaky_relu_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
learning_rate_op_test.py add hill learning rate scheduling 2017-12-19 23:35:44 -08:00
lengths_tile_op_test.py Add GPU support for LengthsTile 2017-10-12 11:17:34 -07:00
lengths_top_k_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
listwise_l2r_operator_test.py Listwise Ranking -- LambdaNDCG 2017-11-16 00:05:48 -08:00
load_save_test.py Move load_save_test.py from caffe2/python/ to caffe2/python/operator_test/ 2018-01-03 17:42:55 -08:00
loss_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
lpnorm_op_test.py Don't run CUDA tests for ops without CUDA implementation 2017-11-08 10:28:02 -08:00
map_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
margin_ranking_criterion_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
math_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
matmul_op_test.py Implement BatchMatMul with Numpy-style batch broadcast semantics 2017-12-05 10:34:35 -08:00
merge_id_lists_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
mkl_conv_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
mkl_packed_fc_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
mkl_speed_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
mod_op_test.py add modulo operator 2017-11-06 16:44:16 -08:00
momentum_sgd_test.py Ensure indices list in sparse optimizer tests is unique 2018-01-03 12:19:14 -08:00
mpi_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
negate_gradient_op_test.py NegateGradientOp and test 2017-11-13 18:05:14 -08:00
normalize_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
one_hot_ops_test.py Add BatchBucketOneHotOp 2017-10-06 13:25:12 -07:00
pack_ops_test.py Reducing array sizes used in pack_ops_test to prevent time outs during Travis CI builds 2017-12-04 12:48:53 -08:00
pack_rnn_sequence_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
pad_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
partition_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
piecewise_linear_transform_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
pooling_test.py Manually applying cudnn5 pull request. 2018-01-02 15:31:33 -08:00
prepend_dim_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
python_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
rank_loss_operator_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
rebatching_queue_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
record_queue_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
recurrent_net_executor_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
recurrent_network_test.py Switch RNNOp to use NetDef argument for step represenetation. 2017-10-10 22:01:51 -07:00
reduce_ops_test.py Refactor Reduce{Front,Back}{Sum,Mean} Operators 2017-12-14 20:02:36 -08:00
reduction_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
relu_op_test.py Kick fbsync 2017-09-29 16:31:11 -07:00
reshape_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
resize_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
rmac_regions_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
rnn_cell_test.py rnn_cell_test: make it determinitistic and speed up 2018-01-04 15:00:42 -08:00
segment_ops_test.py SortedSegmentMean/SortedSegmentLogMeanExp Gradients CUDA implementation. 2017-12-14 13:05:19 -08:00
selu_op_test.py selu op 2017-12-15 15:38:44 -08:00
sequence_ops_test.py CUDA implementation for GatherPadddingOp 2017-12-15 16:05:31 -08:00
shape_inference_test.py Refactor Reduce{Front,Back}{Sum,Mean} Operators 2017-12-14 20:02:36 -08:00
sinusoid_position_encoding_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
softmax_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
softplus_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
sparse_gradient_checker_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
sparse_lengths_sum_benchmark.py Re-license to Apache 2017-09-28 16:22:00 -07:00
sparse_ops_test.py Add CUDA version of ScatterAssign 2017-10-24 10:20:03 -07:00
sparse_to_dense_mask_op_test.py Add int64 support to sparse_to_dense_mask_op 2017-10-10 17:19:44 -07:00
spatial_bn_op_test.py Updated brew SpatialBN to use initializers 2017-11-02 11:25:45 -07:00
specialized_segment_ops_test.py Fix boundary checking in 8-bit sparselengthssum ops 2017-11-03 05:19:57 -07:00
square_root_divide_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
stats_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
string_ops_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
text_file_reader_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
tile_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
top_k_test.py Reduce caffe2 GPU topk test sizes 2017-11-10 07:47:00 -08:00
unique_uniform_fill_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
utility_ops_test.py Manually applying cudnn5 pull request. 2018-01-02 15:31:33 -08:00
video_input_op_test.py Re-license to Apache 2017-09-28 16:22:00 -07:00
weighted_sample_test.py support non-normalized weights 2017-10-27 23:18:25 -07:00
weighted_sum_test.py Don't run CUDA tests for ops without CUDA implementation 2017-11-08 10:28:02 -08:00