dnn-conv2d.daph

/*
 * Copyright 2021 The DAPHNE Consortium
 *
 * Licensed 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.
 */

# ---- !!! ----
# NOTE: This script uses DNN operations that DAPHNE can only provide via CUDA (e.g., GPU support required)


msec_factor = as.f32(0.000001);

# num images
N = 5;

# image size (will be squared)
Isize = 32;

# num channels
C = 10;

# num filters
Fsize = 5;

# filter size
F = 64;

# seed for random input generation
seed = 42;

t_start = now();
X = rand(N, Isize*Isize*C, 0.0, 256.0, 1.0, seed);
# print("X:",0); print(X);
# print("X dims: ",0) ;print(nrow(X),0);print("x",0);print(ncol(X));

# Uncomment, to alternatively read pre generated input
# X = readMatrix("test_input.csv");
# N = as.si64(nrow(X));
# t_end_read = now();
# print("Time input read:  ",0);print(as.f32((t_end_read - t_start)) * msec_factor,0);print(" ms");

W = rand(F, C*Fsize*Fsize, 0.0, 1.0, 1.0, seed);
b = rand(1, F, 0.0, 1.0, 1.0, seed);

gamma = rand(F, 1, 1.0, 1.0, 1.0, seed);
beta = fill(0.0, F, 1);
ema_mean = fill(0.0, F, 1);
ema_var = rand(F, 1, 1.0, 1.0, 1.0, seed);
eps = 0.00001;

t_start_conv = now();
X_conv, Hout, Wout = conv2d(X, W, N, F, Isize, Isize, Fsize, Fsize, 1, 1, 1, 1, b);
t_end_conv = now();
X_bn = batch_norm2d(X_conv, gamma, beta, ema_mean, ema_var, eps);
t_end_bn = now();
X_relu = relu(X_bn);
t_end_relu = now();

# print("X_conv",0); print(X_conv[fill(1,1,1),]);
print("X_conv dims: ",0) ;print(nrow(X_conv),0);print("x",0);print(ncol(X_conv));
print("Time conv:  ",0);print(as.f32((t_end_conv - t_start_conv)) * msec_factor,0);print(" ms");
print("Time bn:    ",0);print(as.f32((t_end_bn - t_end_conv)) * msec_factor,0);print(" ms");
print("Time relu:  ",0);print(as.f32((t_end_relu - t_end_bn)) * msec_factor,0);print(" ms");
print("Time total: ",0);print(as.f32((t_end_relu - t_start)) * msec_factor,0);print(" ms");