<!-- ~ Licensed to the Apache Software Foundation (ASF) under one ~ or more contributor license agreements. See the NOTICE file ~ distributed with this work for additional information ~ regarding copyright ownership. The ASF licenses this file ~ to you under the Apache License, Version 2.0 (the ~ "License"); you may not use this file except in compliance ~ with the License. You may obtain a copy of the License at ~ ~ http://www.apache.org/licenses/LICENSE-2.0 ~ ~ 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. ~ -->

MXNet GPU Operator Benchmarks

Settings

1. MXNet - v1.5.0 with CUDA 10.1
2. Instance - P3.2x

Operator	Avg Forward Time (ms)	Avg. Backward Time (ms)	Max Mem Usage (Storage) (Bytes)	Inputs
rcbrt	0.0384	0.0393	2097.1521	{'data': (1024, 1024)}
rcbrt	0.0268	0.0252	20.0	{'data': (10000, 1)}
rcbrt	0.0378	0.039	2000.0	{'data': (10000, 100)}
min	0.0679	0.0955	0.002	{'axis': (), 'data': (1024, 1024)}
min	0.0337	0.022	0.002	{'axis': 0, 'data': (10000, 1)}
min	0.0669	0.0921	0.002	{'axis': (0, 1), 'data': (10000, 100)}
broadcast_maximum	0.0284	0.0308	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
Flatten	0.0416	---	2097.1521	{'data': (1024, 1024)}
Flatten	0.03	---	20.0	{'data': (10000, 1)}
Flatten	0.0437	---	2000.0	{'data': (10000, 100)}
stop_gradient	---	---	2097.1521	{'data': (1024, 1024)}
stop_gradient	---	---	20.0	{'data': (10000, 1)}
stop_gradient	---	---	2000.0	{'data': (10000, 100)}
broadcast_minimum	0.0284	0.0294	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
random_generalized_negative_binomial	63.299	---	2097.1521	{'shape': (1024, 1024)}
random_generalized_negative_binomial	0.6491	---	20.0	{'shape': (10000, 1)}
random_generalized_negative_binomial	60.3705	---	2000.0	{'shape': (10000, 100)}
broadcast_sub	0.0198	0.0208	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
round	0.0363	---	2097.1521	{'data': (1024, 1024)}
round	0.0248	---	20.0	{'data': (10000, 1)}
round	0.0358	---	2000.0	{'data': (10000, 100)}
cosh	0.0384	0.0389	2097.1521	{'data': (1024, 1024)}
cosh	0.0271	0.0255	20.0	{'data': (10000, 1)}
cosh	0.0378	0.0387	2000.0	{'data': (10000, 100)}
max	0.068	0.095	0.002	{'axis': (), 'data': (1024, 1024)}
max	0.0336	0.0219	0.002	{'axis': 0, 'data': (10000, 1)}
max	0.0665	0.0918	0.002	{'axis': (0, 1), 'data': (10000, 100)}
tanh	0.0391	0.038	2097.1521	{'data': (1024, 1024)}
tanh	0.0271	0.025	20.0	{'data': (10000, 1)}
tanh	0.0373	0.0376	2000.0	{'data': (10000, 100)}
relu	0.0382	0.0383	2097.1521	{'data': (1024, 1024)}
relu	0.0272	0.0251	20.0	{'data': (10000, 1)}
relu	0.0383	0.038	2000.0	{'data': (10000, 100)}
negative	0.0367	---	2097.1521	{'data': (1024, 1024)}
negative	0.025	---	20.0	{'data': (10000, 1)}
negative	0.0361	---	2000.0	{'data': (10000, 100)}
random_randint	9.0025	---	2097.1521	{'shape': (1024, 1024), 'high': 5, 'low': 0}
random_randint	0.0976	---	20.0	{'shape': (10000, 1), 'high': 5, 'low': 0}
random_randint	8.589	---	2000.0	{'shape': (10000, 100), 'high': 5, 'low': 0}
trunc	0.0367	---	2097.1521	{'data': (1024, 1024)}
trunc	0.0251	---	20.0	{'data': (10000, 1)}
trunc	0.0359	---	2000.0	{'data': (10000, 100)}
log_softmax	0.0433	0.0387	2097.1521	{'axis': -1, 'data': (1024, 1024), 'temperature': 0.5}
log_softmax	0.055	0.0386	20.0	{'axis': -1, 'data': (10000, 1), 'temperature': 0.5}
log_softmax	0.08	0.0553	2000.0	{'axis': -1, 'data': (10000, 100), 'temperature': 0.5}
broadcast_greater_equal	0.0191	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
arctan	0.0386	0.04	2097.1521	{'data': (1024, 1024)}
arctan	0.0266	0.0248	20.0	{'data': (10000, 1)}
arctan	0.0378	0.0394	2000.0	{'data': (10000, 100)}
sqrt	0.0385	0.041	2097.1521	{'data': (1024, 1024)}
sqrt	0.0269	0.025	20.0	{'data': (10000, 1)}
sqrt	0.0378	0.0397	2000.0	{'data': (10000, 100)}
dot	0.215	0.4045	2097.1521	{'lhs': (1024, 1024), 'rhs': (1024, 1024)}
dot	0.031	0.0633	2000.0	{'lhs': (1000, 10), 'transpose_b': True, 'rhs': (1000, 10)}
dot	0.042	0.0388	0.2	{'lhs': (1000, 1), 'transpose_b': True, 'transpose_a': True, 'rhs': (100, 1000)}
floor	0.0366	---	2097.1521	{'data': (1024, 1024)}
floor	0.0249	---	20.0	{'data': (10000, 1)}
floor	0.0366	---	2000.0	{'data': (10000, 100)}
broadcast_logical_xor	0.0264	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
shuffle	0.1016	---	2097.1521	{'data': (1024, 1024)}
shuffle	0.2845	---	60.0	{'data': (10000, 1)}
shuffle	0.2798	---	2000.0	{'data': (10000, 100)}
nansum	0.0741	0.0948	0.002	{'axis': (), 'data': (1024, 1024)}
nansum	0.0368	0.0221	0.002	{'axis': 0, 'data': (10000, 1)}
nansum	0.0731	0.0918	0.002	{'axis': (0, 1), 'data': (10000, 100)}
nanprod	0.0717	0.0955	0.002	{'axis': (), 'data': (1024, 1024)}
nanprod	0.0344	0.0223	0.002	{'axis': 0, 'data': (10000, 1)}
nanprod	0.0711	0.0921	0.002	{'axis': (0, 1), 'data': (10000, 100)}
broadcast_div	0.02	0.0227	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
LeakyReLU	0.0326	0.0333	2097.1521	{'slope': 0.1, 'data': (1024, 1024), 'act_type': 'leaky'}
LeakyReLU	0.0347	0.0199	20.0	{'slope': 0.1, 'data': (10000, 1), 'act_type': 'leaky'}
LeakyReLU	0.0316	0.0327	2000.0	{'slope': 0.1, 'data': (10000, 100), 'act_type': 'leaky'}
LeakyReLU	0.0328	0.0334	2097.1521	{'slope': 0.1, 'data': (1024, 1024), 'act_type': 'elu'}
LeakyReLU	0.0208	0.0194	20.0	{'slope': 0.1, 'data': (10000, 1), 'act_type': 'elu'}
LeakyReLU	0.0315	0.0326	2000.0	{'slope': 0.1, 'data': (10000, 100), 'act_type': 'elu'}
LeakyReLU	0.0323	0.033	2097.1521	{'data': (1024, 1024), 'act_type': 'selu'}
LeakyReLU	0.0209	0.0194	20.0	{'data': (10000, 1), 'act_type': 'selu'}
LeakyReLU	0.0315	0.0325	2000.0	{'data': (10000, 100), 'act_type': 'selu'}
LeakyReLU	0.0349	0.0917	2097.1521	{'gamma': (1, 1024), 'data': (1024, 1024), 'act_type': 'prelu'}
LeakyReLU	0.0244	0.0465	20.0	{'gamma': (1, 1), 'data': (10000, 1), 'act_type': 'prelu'}
LeakyReLU	0.034	0.1003	2000.0	{'gamma': (1, 100), 'data': (10000, 100), 'act_type': 'prelu'}
FullyConnected	0.3141	0.3486	4.096	{'weight': (64, 196608), 'num_hidden': 64, 'data': (32, 3, 256, 256), 'bias': (64,), 'flatten': True}
FullyConnected	0.1381	0.5142	3145.728	{'weight': (64, 256), 'num_hidden': 64, 'data': (32, 3, 256, 256), 'bias': (64,), 'flatten': False}
square	0.038	0.038	2097.1521	{'data': (1024, 1024)}
square	0.0281	0.0262	20.0	{'data': (10000, 1)}
square	0.0424	0.0422	2000.0	{'data': (10000, 100)}
elemwise_mul	0.0186	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
broadcast_not_equal	0.0262	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
broadcast_logical_or	0.019	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
logical_not	0.0366	---	2097.1521	{'data': (1024, 1024)}
logical_not	0.0247	---	20.0	{'data': (10000, 1)}
logical_not	0.0357	---	2000.0	{'data': (10000, 100)}
broadcast_add	0.0284	0.0289	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
broadcast_hypot	0.0202	0.0233	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
sign	0.038	0.0333	2097.1521	{'data': (1024, 1024)}
sign	0.0269	0.0252	20.0	{'data': (10000, 1)}
sign	0.0379	0.0326	2000.0	{'data': (10000, 100)}
arccos	0.0385	0.0424	2097.1521	{'data': (1024, 1024)}
arccos	0.0268	0.0253	20.0	{'data': (10000, 1)}
arccos	0.0374	0.0413	2000.0	{'data': (10000, 100)}
erf	0.0383	0.0384	2097.1521	{'data': (1024, 1024)}
erf	0.0266	0.0247	20.0	{'data': (10000, 1)}
erf	0.0374	0.0384	2000.0	{'data': (10000, 100)}
degrees	0.0375	0.0326	2097.1521	{'data': (1024, 1024)}
degrees	0.0265	0.0247	20.0	{'data': (10000, 1)}
degrees	0.0376	0.0317	2000.0	{'data': (10000, 100)}
fix	0.0366	---	2097.1521	{'data': (1024, 1024)}
fix	0.0251	---	20.0	{'data': (10000, 1)}
fix	0.0362	---	2000.0	{'data': (10000, 100)}
cos	0.0386	0.0389	2097.1521	{'data': (1024, 1024)}
cos	0.0276	0.0255	20.0	{'data': (10000, 1)}
cos	0.0385	0.0388	2000.0	{'data': (10000, 100)}
broadcast_plus	---	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
sample_exponential	0.1359	---	4194.3042	{'lam': [1.0, 8.5], 'shape': (1024, 1024)}
sample_exponential	0.0734	---	40.0	{'lam': [1.0, 8.5], 'shape': (10000, 1)}
sample_exponential	0.1323	---	4000.0	{'lam': [1.0, 8.5], 'shape': (10000, 100)}
make_loss	0.0415	---	2097.1521	{'data': (1024, 1024)}
make_loss	0.0301	---	20.0	{'data': (10000, 1)}
make_loss	0.0432	---	2000.0	{'data': (10000, 100)}
argmax_channel	0.288	---	2.048	{'data': (1024, 1024)}
argmax_channel	0.0264	---	20.0	{'data': (10000, 1)}
argmax_channel	0.0439	---	20.0	{'data': (10000, 100)}
BlockGrad	0.042	---	2097.1521	{'data': (1024, 1024)}
BlockGrad	0.0295	---	20.0	{'data': (10000, 1)}
BlockGrad	0.0438	---	2000.0	{'data': (10000, 100)}
BatchNorm	1.1379	2.3158	12582.9238	{'gamma': (3,), 'moving_var': (3,), 'data': (32, 3, 256, 256), 'beta': (3,), 'moving_mean': (3,)}
BatchNorm	1.7465	3.5398	19200.0117	{'gamma': (3,), 'moving_var': (3,), 'data': (32, 3, 10000, 10), 'beta': (3,), 'moving_mean': (3,)}
sample_negative_binomial	1.5065	---	4194.3042	{'shape': (1024, 1024), 'p': [0.4, 0.77], 'k': [20, 49]}
sample_negative_binomial	0.8542	---	40.0	{'shape': (10000, 1), 'p': [0.4, 0.77], 'k': [20, 49]}
sample_negative_binomial	1.476	---	4000.0	{'shape': (10000, 100), 'p': [0.4, 0.77], 'k': [20, 49]}
batch_dot	4.7282	9.3181	67108.8672	{'lhs': (32, 1024, 1024), 'rhs': (32, 1024, 1024)}
batch_dot	0.3021	1.229	64000.0	{'lhs': (32, 1000, 10), 'transpose_b': True, 'rhs': (32, 1000, 10)}
batch_dot	0.0523	0.0636	6.4	{'lhs': (32, 1000, 1), 'transpose_b': True, 'transpose_a': True, 'rhs': (32, 100, 1000)}
rsqrt	0.0386	0.0422	2097.1521	{'data': (1024, 1024)}
rsqrt	0.0268	0.0252	20.0	{'data': (10000, 1)}
rsqrt	0.0377	0.0408	2000.0	{'data': (10000, 100)}
broadcast_logical_and	0.0191	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
mean	0.0818	0.043	0.002	{'axis': (), 'data': (1024, 1024)}
mean	0.0384	0.026	0.002	{'axis': 0, 'data': (10000, 1)}
mean	0.0801	0.0421	0.002	{'axis': (0, 1), 'data': (10000, 100)}
ceil	0.0363	---	2097.1521	{'data': (1024, 1024)}
ceil	0.0249	---	20.0	{'data': (10000, 1)}
ceil	0.0357	---	2000.0	{'data': (10000, 100)}
min_axis	0.0569	---	0.002	{'axis': (), 'data': (1024, 1024)}
min_axis	0.0319	---	0.002	{'axis': 0, 'data': (10000, 1)}
min_axis	0.0563	---	0.002	{'axis': (0, 1), 'data': (10000, 100)}
sigmoid	0.0395	0.0392	2097.1521	{'data': (1024, 1024)}
sigmoid	0.0412	0.0256	20.0	{'data': (10000, 1)}
sigmoid	0.0378	0.0384	2000.0	{'data': (10000, 100)}
broadcast_power	0.0209	0.023	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
gamma	0.0415	0.0761	2097.1521	{'data': (1024, 1024)}
gamma	0.0279	0.0279	20.0	{'data': (10000, 1)}
gamma	0.041	0.0736	2000.0	{'data': (10000, 100)}
radians	0.0378	0.0327	2097.1521	{'data': (1024, 1024)}
radians	0.0268	0.0248	20.0	{'data': (10000, 1)}
radians	0.0373	0.032	2000.0	{'data': (10000, 100)}
prod	0.0671	0.0949	0.002	{'axis': (), 'data': (1024, 1024)}
prod	0.0334	0.0218	0.002	{'axis': 0, 'data': (10000, 1)}
prod	0.0663	0.091	0.002	{'axis': (0, 1), 'data': (10000, 100)}
abs	0.0382	0.039	2097.1521	{'data': (1024, 1024)}
abs	0.0271	0.0254	20.0	{'data': (10000, 1)}
abs	0.0369	0.0381	2000.0	{'data': (10000, 100)}
reciprocal	0.0378	0.0399	2097.1521	{'data': (1024, 1024)}
reciprocal	0.0267	0.0248	20.0	{'data': (10000, 1)}
reciprocal	0.0374	0.0391	2000.0	{'data': (10000, 100)}
sample_generalized_negative_binomial	1.0903	---	4194.3042	{'mu': [2.0, 2.5], 'alpha': [0.0, 2.5], 'shape': (1024, 1024)}
sample_generalized_negative_binomial	0.6051	---	40.0	{'mu': [2.0, 2.5], 'alpha': [0.0, 2.5], 'shape': (10000, 1)}
sample_generalized_negative_binomial	1.0655	---	4000.0	{'mu': [2.0, 2.5], 'alpha': [0.0, 2.5], 'shape': (10000, 100)}
rint	0.0369	---	2097.1521	{'data': (1024, 1024)}
rint	0.025	---	20.0	{'data': (10000, 1)}
rint	0.0357	---	2000.0	{'data': (10000, 100)}
arcsin	0.0383	0.0412	2097.1521	{'data': (1024, 1024)}
arcsin	0.0269	0.0253	20.0	{'data': (10000, 1)}
arcsin	0.038	0.0402	2000.0	{'data': (10000, 100)}
sample_poisson	0.4035	---	4194.3042	{'lam': [1.0, 8.5], 'shape': (1024, 1024)}
sample_poisson	0.3288	---	40.0	{'lam': [1.0, 8.5], 'shape': (10000, 1)}
sample_poisson	0.4029	---	4000.0	{'lam': [1.0, 8.5], 'shape': (10000, 100)}
Pooling	0.0225	0.0262	49.152	{'stride': 1, 'pool_type': 'avg', 'data': (32, 3, 256), 'layout': 'NCW', 'global_pool': 0, 'kernel': 3, 'pad': 1}
sample_uniform	0.1386	---	4194.3042	{'shape': (1024, 1024), 'high': [1.0, 3.7], 'low': [0.0, 2.5]}
sample_uniform	0.0776	---	40.0	{'shape': (10000, 1), 'high': [1.0, 3.7], 'low': [0.0, 2.5]}
sample_uniform	0.1358	---	4000.0	{'shape': (10000, 100), 'high': [1.0, 3.7], 'low': [0.0, 2.5]}
ones_like	0.0326	---	2097.1521	{'data': (1024, 1024)}
ones_like	0.0251	---	40.0	{'data': (10000, 1)}
ones_like	0.032	---	2000.0	{'data': (10000, 100)}
identity	---	---	2097.1521	{'data': (1024, 1024)}
identity	---	---	40.0	{'data': (10000, 1)}
identity	---	---	2000.0	{'data': (10000, 100)}
broadcast_lesser_equal	0.0189	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
broadcast_lesser	0.0189	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
zeros_like	0.0326	---	2097.1521	{'data': (1024, 1024)}
zeros_like	0.0252	---	40.0	{'data': (10000, 1)}
zeros_like	0.0321	---	2000.0	{'data': (10000, 100)}
random_uniform	3.1027	---	2097.1521	{'shape': (1024, 1024), 'high': 5, 'low': 0}
random_uniform	0.0588	---	20.0	{'shape': (10000, 1), 'high': 5, 'low': 0}
random_uniform	2.9599	---	2000.0	{'shape': (10000, 100), 'high': 5, 'low': 0}
broadcast_mod	0.0196	0.0227	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
cbrt	0.0383	0.0405	2097.1521	{'data': (1024, 1024)}
cbrt	0.0272	0.0251	20.0	{'data': (10000, 1)}
cbrt	0.0377	0.0394	2000.0	{'data': (10000, 100)}
broadcast_minus	---	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
random_negative_binomial	52.4089	---	2097.1521	{'shape': (1024, 1024), 'p': 1, 'k': 1}
random_negative_binomial	0.5224	---	20.0	{'shape': (10000, 1), 'p': 1, 'k': 1}
random_negative_binomial	49.9987	---	2000.0	{'shape': (10000, 100), 'p': 1, 'k': 1}
log2	0.0382	0.0405	2097.1521	{'data': (1024, 1024)}
log2	0.0269	0.0253	20.0	{'data': (10000, 1)}
log2	0.0376	0.0403	2000.0	{'data': (10000, 100)}
broadcast_greater	0.0191	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
max_axis	0.0564	---	0.002	{'axis': (), 'data': (1024, 1024)}
max_axis	0.0315	---	0.002	{'axis': 0, 'data': (10000, 1)}
max_axis	0.0561	---	0.002	{'axis': (0, 1), 'data': (10000, 100)}
sample_gamma	0.3435	---	4194.3042	{'alpha': [0.0, 2.5], 'beta': [1.0, 0.7], 'shape': (1024, 1024)}
sample_gamma	0.2938	---	40.0	{'alpha': [0.0, 2.5], 'beta': [1.0, 0.7], 'shape': (10000, 1)}
sample_gamma	0.341	---	4000.0	{'alpha': [0.0, 2.5], 'beta': [1.0, 0.7], 'shape': (10000, 100)}
sin	0.0382	0.0386	2097.1521	{'data': (1024, 1024)}
sin	0.0268	0.0244	20.0	{'data': (10000, 1)}
sin	0.0374	0.0386	2000.0	{'data': (10000, 100)}
sum	0.0797	0.0341	0.002	{'axis': (), 'data': (1024, 1024)}
sum	0.0362	0.021	0.002	{'axis': 0, 'data': (10000, 1)}
sum	0.0771	0.0335	0.002	{'axis': (0, 1), 'data': (10000, 100)}
erfinv	0.1069	0.1112	2097.1521	{'data': (1024, 1024)}
erfinv	0.029	0.0271	20.0	{'data': (10000, 1)}
erfinv	0.1043	0.1055	2000.0	{'data': (10000, 100)}
random_gamma	46.4529	---	2097.1521	{'shape': (1024, 1024)}
random_gamma	0.466	---	20.0	{'shape': (10000, 1)}
random_gamma	44.4208	---	2000.0	{'shape': (10000, 100)}
broadcast_mul	0.019	0.0222	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
arccosh	0.0405	0.0414	2097.1521	{'data': (1024, 1024)}
arccosh	0.0269	0.0253	20.0	{'data': (10000, 1)}
arccosh	0.0396	0.0402	2000.0	{'data': (10000, 100)}
log1p	0.0383	0.0399	2097.1521	{'data': (1024, 1024)}
log1p	0.0269	0.0248	20.0	{'data': (10000, 1)}
log1p	0.0376	0.0394	2000.0	{'data': (10000, 100)}
size_array	0.0229	---	0.004	{'data': (1024, 1024)}
size_array	0.0367	---	0.004	{'data': (10000, 1)}
size_array	0.022	---	0.004	{'data': (10000, 100)}
arcsinh	0.0396	0.0406	2097.1521	{'data': (1024, 1024)}
arcsinh	0.0269	0.0249	20.0	{'data': (10000, 1)}
arcsinh	0.0388	0.0396	2000.0	{'data': (10000, 100)}
elemwise_div	0.0188	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
elemwise_add	0.019	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
elemwise_sub	0.0186	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
arctanh	0.0385	0.0397	2097.1521	{'data': (1024, 1024)}
arctanh	0.0266	0.0245	20.0	{'data': (10000, 1)}
arctanh	0.0376	0.0391	2000.0	{'data': (10000, 100)}
log	0.0381	0.0399	2097.1521	{'data': (1024, 1024)}
log	0.0268	0.025	20.0	{'data': (10000, 1)}
log	0.0377	0.0394	2000.0	{'data': (10000, 100)}
gammaln	0.0461	0.0664	2097.1521	{'data': (1024, 1024)}
gammaln	0.0273	0.0266	20.0	{'data': (10000, 1)}
gammaln	0.0455	0.0645	2000.0	{'data': (10000, 100)}
Dropout	0.122	0.0957	25165.8242	{'mode': 'always', 'data': (32, 3, 256, 256), 'p': 0.5}
Dropout	0.0739	0.0241	400.0	{'mode': 'always', 'data': (10000, 10), 'p': 0.5}
softmax	0.0454	0.0438	2097.1521	{'axis': -1, 'data': (1024, 1024), 'temperature': 0.5}
softmax	0.056	0.0377	20.0	{'axis': -1, 'data': (10000, 1), 'temperature': 0.5}
softmax	0.0805	0.0609	2000.0	{'axis': -1, 'data': (10000, 100), 'temperature': 0.5}
expm1	0.0386	0.0387	2097.1521	{'data': (1024, 1024)}
expm1	0.0275	0.025	20.0	{'data': (10000, 1)}
expm1	0.0375	0.0381	2000.0	{'data': (10000, 100)}
log10	0.0382	0.0403	2097.1521	{'data': (1024, 1024)}
log10	0.0274	0.0251	20.0	{'data': (10000, 1)}
log10	0.0369	0.0386	2000.0	{'data': (10000, 100)}
random_poisson	15.4201	---	2097.1521	{'shape': (1024, 1024)}
random_poisson	0.176	---	20.0	{'shape': (10000, 1)}
random_poisson	14.9428	---	2000.0	{'shape': (10000, 100)}
sinh	0.0381	0.0381	2097.1521	{'data': (1024, 1024)}
sinh	0.0269	0.0246	20.0	{'data': (10000, 1)}
sinh	0.0376	0.038	2000.0	{'data': (10000, 100)}
random_normal	16.4453	---	2097.1521	{'shape': (1024, 1024)}
random_normal	0.1864	---	20.0	{'shape': (10000, 1)}
random_normal	15.553	---	2000.0	{'shape': (10000, 100)}
hard_sigmoid	0.031	0.0336	2097.1521	{'alpha': 0.25, 'data': (1024, 1024), 'beta': 0.5}
hard_sigmoid	0.0197	0.0182	20.0	{'alpha': 0.25, 'data': (10000, 1), 'beta': 0.5}
hard_sigmoid	0.0303	0.0322	2000.0	{'alpha': 0.25, 'data': (10000, 100), 'beta': 0.5}
flatten	---	---	2097.1521	{'data': (1024, 1024)}
flatten	---	---	20.0	{'data': (10000, 1)}
flatten	---	---	2000.0	{'data': (10000, 100)}
random_exponential	15.9188	---	2097.1521	{'shape': (1024, 1024)}
random_exponential	0.1663	---	40.0	{'shape': (10000, 1)}
random_exponential	15.1982	---	2000.0	{'shape': (10000, 100)}
tan	0.0395	0.0377	2097.1521	{'data': (1024, 1024)}
tan	0.027	0.025	20.0	{'data': (10000, 1)}
tan	0.0384	0.037	2000.0	{'data': (10000, 100)}
broadcast_equal	0.019	---	0.012	{'lhs': [(1024, 1024), (10000, 10), (10000, 1)], 'rhs': [(1024, 1024), (10000, 10), (10000, 1)]}
softsign	0.0384	0.0402	2097.1521	{'data': (1024, 1024)}
softsign	0.0269	0.025	20.0	{'data': (10000, 1)}
softsign	0.0375	0.04	2000.0	{'data': (10000, 100)}
Convolution	0.0578	0.1121	1040.384	{'stride': (1,), 'data': (32, 3, 256), 'bias': (64,), 'layout': 'NCW', 'dilate': (1,), 'kernel': (3,), 'weight': (64, 3, 3), 'pad': (0,), 'num_filter': 64}
sample_normal	0.1549	---	4194.3042	{'mu': [2.0, 2.5], 'sigma': [1.0, 3.7], 'shape': (1024, 1024)}
sample_normal	0.1027	---	40.0	{'mu': [2.0, 2.5], 'sigma': [1.0, 3.7], 'shape': (10000, 1)}
sample_normal	0.1522	---	4000.0	{'mu': [2.0, 2.5], 'sigma': [1.0, 3.7], 'shape': (10000, 100)}
exp	0.0365	---	2097.1521	{'data': (1024, 1024)}
exp	0.0246	---	20.0	{'data': (10000, 1)}
exp	0.0356	---	2000.0	{'data': (10000, 100)}
sum_axis	0.0637	---	0.004	{'axis': (), 'data': (1024, 1024)}
sum_axis	0.0351	---	0.002	{'axis': 0, 'data': (10000, 1)}
sum_axis	0.0631	---	0.002	{'axis': (0, 1), 'data': (10000, 100)}