Merge branch 'opperf_large_tensor_flag' of https://github.com/connorgoggins/incubator-mxnet into opperf_large_tensor_flag

Author: connorgoggins

benchmark/opperf/rules/default_params.py

@@ -369,6 +369,11 @@
 DEFAULT_DATA_SMCE = [(1024, 1024)]
 DEFAULT_LABEL_SMCE = [(1024,)]
 
+DEFAULT_LABEL_LARGE_TENSOR = [(1, 1)]
+DEFAULT_DATA_CTCLOSS = [(2**32, 1, 1)]
+DEFAULT_DATA_SMCE_LARGE_TENSOR = [(2**32 + 1, 1)]
+DEFAULT_LABEL_SMCE_LARGE_TENSOR = [(2**32 + 1,)]
+
 # For NN operators
 DEFAULT_ACT_TYPE_LR = ['leaky', 'elu', 'selu', 'gelu']
 DEFAULT_ACT_TYPE_ACTIVATION = ['relu', 'sigmoid', 'softrelu', 'softsign', 'tanh']
@@ -624,6 +629,8 @@
                                 "block_size": DEFAULT_BLOCK_SIZE_LARGE_TENSOR,
                                 "args": DEFAULT_ARGS,
                                 "index": DEFAULT_INDEX_LARGE_TENSOR,
+                                "data_smce": DEFAULT_DATA_SMCE_LARGE_TENSOR,
+                                "label_smce": DEFAULT_LABEL_SMCE_LARGE_TENSOR,
                                 "grid": DEFAULT_GRID_LARGE_TENSOR,
                                 "data_bilinearsampler": DEFAULT_DATA_BILINEAR_LARGE_TENSOR,
                                 "transform_type": DEFAULT_TRANSFORM_TYPE,
@@ -691,6 +698,7 @@
                                 "lhs_fill_element_0index": DEFAULT_LHS_FEI_LARGE_TENSOR,
                                 "rhs_fill_element_0index": DEFAULT_RHS_FEI_LARGE_TENSOR,
                                 "mhs": DEFAULT_MHS_LARGE_TENSOR,
+                                "data_softmax": DEFAULT_LABEL_SOFTMAX_LARGE_TENSOR,
                                 "data_spatialtransformer": DEFAULT_DATA_ST_LARGE_TENSOR,
                                 "loc_spatialtransformer": DEFAULT_LOC_TAR_ST_LARGE_TENSOR,
                                 "target_shape": DEFAULT_LOC_TAR_ST_LARGE_TENSOR,
@@ -700,6 +708,10 @@
                                 "output_size": DEFAULT_OUTPUT_SIZE_LARGE_TENSOR,
                                 "kernel_col2im": DEFAULT_KERNEL_LARGE_TENSOR,
                                 "stride_col2im": DEFAULT_STRIDE,
+                                "data_ctcloss": DEFAULT_DATA_CTCLOSS,
+                                "label_ctcloss": DEFAULT_LABEL_LARGE_TENSOR,
+                                "data_ctc_loss": DEFAULT_DATA_CTCLOSS,
+                                "label_ctc_loss": DEFAULT_LABEL_LARGE_TENSOR,
                                 "data_rnn": DEFAULT_DATA_RNN_LARGE_TENSOR,
                                 "p_rnn": DEFAULT_P_RNN,
                                 "parameters": DEFAULT_PARAMETERS_LARGE_TENSOR,

benchmark/opperf/utils/op_registry_utils.py

@@ -125,7 +125,7 @@ def prepare_op_inputs(op, arg_params, int64_tensor):
                    'MAERegressionOutput', 'SVMOutput', 'L2Normalization', 'LayerNorm', 'InstanceNorm',
                    'Embedding', 'Correlation', 'im2col', 'LRN', 'squeeze', 'fill_element_0index'}
 
-    custom_data_int64 = {'random_pdf_dirichlet', 'random_pdf_exponential', 'random_pdf_gamma', 'random_pdf_generalized_negative_binomial', 'random_pdf_negative_binomial', 'random_pdf_normal', 'random_pdf_poisson', 'random_pdf_uniform', 'sample_exponential', 'sample_normal', 'sample_poisson', 'sample_uniform', 'sample_gamma', 'sample_generalized_negative_binomial', 'sample_negative_binomial'}
+    custom_data_int64 = {'random_pdf_dirichlet', 'random_pdf_exponential', 'random_pdf_gamma', 'random_pdf_generalized_negative_binomial', 'random_pdf_negative_binomial', 'random_pdf_normal', 'random_pdf_poisson', 'random_pdf_uniform', 'sample_exponential', 'sample_normal', 'sample_poisson', 'sample_uniform', 'sample_gamma', 'sample_generalized_negative_binomial', 'sample_negative_binomial', 'CTCLoss', 'ctc_loss'}
 
     int_only = {'random_randint'}
     float_only = {'log_softmax', 'softmax', 'softmin'}