Skip to content

N
ngraph
Unverified Commit 2b0a5489 authored by Adam Procter's avatar Adam Procter Committed by GitHub
Browse files
Options

Change convolution reference to work with f32 (#409)

parent d87b0065
Hide whitespace changes
Inline Side-by-side
Showing with 93 additions and 74 deletions
test/convolution_test.in.cpp
View file @ 2b0a5489
This source diff could not be displayed because it is too large. You can view the blob instead.
test/ref_generators/generate_convolution_ref.py
View file @ 2b0a5489
......@@ -16,27 +16,22 @@
import sys
import numpy as np
import math
import random
from operator import mul
# Imposes the shape on the given 1-D array to produce a C-style-indexed n-D array.
def shaped_from_flat(shape,flat):
total_elems = reduce(mul,shape)
# Generates an array of random floating point literals of the given length, from a fixed seed.
def random_array_float_literals(length,seed=8086):
literals = []
assert(len(flat) == total_elems)
random.seed(seed)
arr = np.array(flat)
arr.shape = shape
for i in range(0,length):
literal_n = random.randint(0,99)
literal_sign = random.randint(0,1)
literal_str = ('-' if literal_sign==1 else '') + '.' + ('%02d' % literal_n)
literals.append(literal_str)
return arr
# Creates a linspaced array from 1 to n where n is the number of elements in the shape, then
# imposes the shape on the array to produce a C-style-indexed n-D array.
def shaped_linspace(shape):
total_elems = reduce(mul,shape)
flat = np.linspace(1,total_elems,total_elems)
return shaped_from_flat(shape,flat)
return literals
# Elementwise addition on tuples.
def tuple_plus(t1,t2):
......@@ -177,31 +172,48 @@ def shape_str(shape):
result = result + (',%d' % d)
return result
def scalar_str(x):
result = ('%.1000g' % x)
# This next part is a bit stupid.
if "." not in result and "e" not in result:
result = result + ".0f"
else:
result = result + "f"
return result
def data_str(data):
result = ''
first = True
for x in np.nditer(data):
if first:
result = ('%.1000g' % x)
result = scalar_str(x)
first = False
else:
result = result + (',%.1000g' % x)
result = result + ',' + scalar_str(x)
return result
def emit_test(t,f):
test_name, input_batch_data, filter_data, move_strides, filter_dilation, below_pads, above_pads, image_dilation, bprop = t
test_name, input_batch_shape, filters_shape, move_strides, filter_dilation, below_pads, above_pads, image_dilation, bprop = t
input_batch_literals = random_array_float_literals(reduce(mul,input_batch_shape))
filters_literals = random_array_float_literals(reduce(mul,filters_shape))
input_batch_array = np.array(map(lambda s: np.float32(s),input_batch_literals))
input_batch_array.shape = input_batch_shape
filters_array = np.array(map(lambda s: np.float32(s),filters_literals))
filters_array.shape = filters_shape
print ("Generating convolution test '%s'..." % test_name)
output_batch_data = convolution_ref(input_batch_data,filter_data,move_strides,filter_dilation,below_pads,above_pads,image_dilation)
output_batch_data = convolution_ref(input_batch_array,filters_array,move_strides,filter_dilation,below_pads,above_pads,image_dilation)
template = '''
TEST (${BACKEND_NAME}, %s)
{
auto shape_a = Shape{%s};
auto A = make_shared<op::Parameter>(element::f64, shape_a);
auto A = make_shared<op::Parameter>(element::f32, shape_a);
auto shape_b = Shape{%s};
auto B = make_shared<op::Parameter>(element::f64, shape_b);
auto B = make_shared<op::Parameter>(element::f32, shape_b);
auto shape_r = Shape{%s};
auto make_graph = [A, B] {
return make_shared<Function>(make_shared<op::Convolution>(A, B,
......@@ -219,81 +231,81 @@ TEST (${BACKEND_NAME}, %s)
auto cf = backend->make_call_frame(external);
// Create some tensors for input/output
auto a = backend->make_primary_tensor_view(element::f64, shape_a);
copy_data(a, vector<double>{%s});
auto b = backend->make_primary_tensor_view(element::f64, shape_b);
copy_data(b, vector<double>{%s});
auto result = backend->make_primary_tensor_view(element::f64, shape_r);
auto a = backend->make_primary_tensor_view(element::f32, shape_a);
copy_data(a, vector<float>{%s});
auto b = backend->make_primary_tensor_view(element::f32, shape_b);
copy_data(b, vector<float>{%s});
auto result = backend->make_primary_tensor_view(element::f32, shape_r);
vector<double> expected_result{%s};
vector<float> expected_result{%s};
cf->call({a, b}, {result});
EXPECT_TRUE(all_close_d(vector<double>{expected_result}, read_vector<double>(result)));
EXPECT_TRUE(all_close<float>(vector<float>{expected_result}, read_vector<float>(result)));
// only test backprop for certain cases as it takes significant compute resources
if(%s) {
EXPECT_TRUE(autodiff_numeric_compare<double>(manager, backend, make_graph, {a, b}, .01, .01));
EXPECT_TRUE(autodiff_numeric_compare<float>(manager, backend, make_graph, {a, b}, .01f, .01f));
}
}
'''
f.write (template % (test_name,
shape_str(input_batch_data.shape),
shape_str(filter_data.shape),
shape_str(input_batch_shape),
shape_str(filters_shape),
shape_str(output_batch_data.shape),
shape_str(move_strides),
shape_str(filter_dilation),
shape_str(below_pads),
shape_str(above_pads),
shape_str(image_dilation),
data_str(input_batch_data),
data_str(filter_data),
",".join(map(lambda s: s + "f",input_batch_literals)),
",".join(map(lambda s: s + "f",filters_literals)),
data_str(output_batch_data),
bprop));
# filter image
# test name input image batch filters stride dilation below-pads above-pads dilation
# filter image
# test name batch shape filts shape stride dilation below-pads above-pads dilation bprop?
tests = [
("convolution_2d_1image", shaped_linspace((1,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (1,1), (0,0), (0,0), (1,1), "true"),
("convolution_2d_1image_padded_1_1x1_1", shaped_linspace((1,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (1,1), (1,1), (1,1), (1,1), "true"),
("convolution_2d_1image_padded_2_3x4_5", shaped_linspace((1,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (1,1), (2,3), (4,5), (1,1), "true"),
("convolution_2d_2images", shaped_linspace((2,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (1,1), (0,0), (0,0), (1,1), "true"),
("convolution_2d_2images_strided", shaped_linspace((2,1,3,5)), shaped_linspace((2,1,2,2)), (2,2), (1,1), (0,0), (0,0), (1,1), "true"),
("convolution_2d_2images_strided_padded", shaped_linspace((2,1,3,5)), shaped_linspace((2,1,2,2)), (2,2), (1,1), (4,2), (5,7), (1,1), "true"),
("convolution_2d_1image", (1,1,3,5), (2,1,2,2), (1,1), (1,1), (0,0), (0,0), (1,1), "true"),
("convolution_2d_1image_padded_1_1x1_1", (1,1,3,5), (2,1,2,2), (1,1), (1,1), (1,1), (1,1), (1,1), "true"),
("convolution_2d_1image_padded_2_3x4_5", (1,1,3,5), (2,1,2,2), (1,1), (1,1), (2,3), (4,5), (1,1), "true"),
("convolution_2d_2images", (2,1,3,5), (2,1,2,2), (1,1), (1,1), (0,0), (0,0), (1,1), "true"),
("convolution_2d_2images_strided", (2,1,3,5), (2,1,2,2), (2,2), (1,1), (0,0), (0,0), (1,1), "true"),
("convolution_2d_2images_strided_padded", (2,1,3,5), (2,1,2,2), (2,2), (1,1), (4,2), (5,7), (1,1), "true"),
("convolution_2d_2images_strided_padded_same",
shaped_linspace((2,1,3,5)), shaped_linspace((2,1,2,2)), (2,2), (1,1), (2,2), (2,2), (1,1), "true"),
("convolution_2d_2images_dilated", shaped_linspace((2,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (2,2), (0,0), (0,0), (1,1), "true"),
("convolution_2d_2images_dilated_padded", shaped_linspace((2,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (2,2), (4,2), (5,7), (1,1), "true"),
("convolution_3d_2images", shaped_linspace((2,1,3,5,8)), shaped_linspace((2,1,2,2,3)), (1,1,1), (1,1,1), (0,0,0), (0,0,0), (1,1,1), "true"),
("convolution_4d_2images", shaped_linspace((2,1,3,5,8,7)),shaped_linspace((2,1,2,2,3,1)),(1,1,1,1),(1,1,1,1),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images", shaped_linspace((4,3,3,5,8,7)),shaped_linspace((4,3,2,2,3,1)),(1,1,1,1),(1,1,1,1),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_padded_neg", shaped_linspace((4,3,3,5,8,7)),shaped_linspace((4,3,2,2,3,1)),(1,1,1,1),(1,1,1,1),(-1,2,-3,2),(1,0,0,-3), (1,1,1,1), "false"),
("convolution_4d_4images_strided", shaped_linspace((4,3,3,5,8,7)),shaped_linspace((4,3,2,2,3,1)),(2,1,3,2),(1,1,1,1),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_dilated", shaped_linspace((4,3,3,5,8,7)),shaped_linspace((4,3,2,2,3,1)),(1,1,1,1),(2,1,3,2),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_strided_dilated",shaped_linspace((4,3,8,8,8,8)),shaped_linspace((4,3,2,2,3,1)),(3,2,2,3),(2,1,3,2),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
(2,1,3,5), (2,1,2,2), (2,2), (1,1), (2,2), (2,2), (1,1), "true"),
("convolution_2d_2images_dilated", (2,1,3,5), (2,1,2,2), (1,1), (2,2), (0,0), (0,0), (1,1), "true"),
("convolution_2d_2images_dilated_padded", (2,1,3,5), (2,1,2,2), (1,1), (2,2), (4,2), (5,7), (1,1), "true"),
("convolution_3d_2images", (2,1,3,5,8), (2,1,2,2,3), (1,1,1), (1,1,1), (0,0,0), (0,0,0), (1,1,1), "true"),
("convolution_4d_2images", (2,1,3,5,8,7),(2,1,2,2,3,1),(1,1,1,1),(1,1,1,1),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images", (4,3,3,5,8,7),(4,3,2,2,3,1),(1,1,1,1),(1,1,1,1),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_padded_neg", (4,3,3,5,8,7),(4,3,2,2,3,1),(1,1,1,1),(1,1,1,1),(-1,2,-3,2),(1,0,0,-3), (1,1,1,1), "false"),
("convolution_4d_4images_strided", (4,3,3,5,8,7),(4,3,2,2,3,1),(2,1,3,2),(1,1,1,1),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_dilated", (4,3,3,5,8,7),(4,3,2,2,3,1),(1,1,1,1),(2,1,3,2),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_strided_dilated",(4,3,8,8,8,8),(4,3,2,2,3,1),(3,2,2,3),(2,1,3,2),(0,0,0,0), (0,0,0,0), (1,1,1,1), "false"),
("convolution_4d_4images_strided_dilated_padded",
shaped_linspace((4,3,8,8,8,8)),shaped_linspace((4,3,2,2,3,1)),(3,2,2,3),(2,1,3,2),(2,4,6,8), (1,3,5,7), (1,1,1,1), "false"),
(4,3,8,8,8,8),(4,3,2,2,3,1),(3,2,2,3),(2,1,3,2),(2,4,6,8), (1,3,5,7), (1,1,1,1), "false"),
("convolution_4d_4images_strided_dilated_padded_neg",
shaped_linspace((4,3,8,8,8,8)),shaped_linspace((4,3,2,2,3,1)),(3,2,2,3),(2,1,3,2),(-2,4,0,5), (1,3,-1,-4),(1,1,1,1), "false"),
(4,3,8,8,8,8),(4,3,2,2,3,1),(3,2,2,3),(2,1,3,2),(-2,4,0,5), (1,3,-1,-4),(1,1,1,1), "false"),
("convolution_4d_4images_strided_dilated_padded_same",
shaped_linspace((4,3,8,8,8,8)),shaped_linspace((4,3,2,2,3,1)),(3,2,2,3),(2,1,3,2),(3,3,3,3), (3,3,3,3), (1,1,1,1), "false"),
("convolution_2d_1image_1o1i_img_dilated",shaped_linspace((1,1,3,5)), shaped_linspace((1,1,2,2)), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_1image_2o1i_img_dilated",shaped_linspace((1,1,3,5)), shaped_linspace((2,1,2,2)), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_1image_2o2i_img_dilated",shaped_linspace((1,2,3,5)), shaped_linspace((2,2,2,2)), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_1image_5o3i_img_dilated",shaped_linspace((1,3,3,5)), shaped_linspace((5,3,2,2)), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_8image_5o3i_img_dilated",shaped_linspace((8,3,3,5)), shaped_linspace((5,3,2,2)), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
(4,3,8,8,8,8),(4,3,2,2,3,1),(3,2,2,3),(2,1,3,2),(3,3,3,3), (3,3,3,3), (1,1,1,1), "false"),
("convolution_2d_1image_1o1i_img_dilated",(1,1,3,5), (1,1,2,2), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_1image_2o1i_img_dilated",(1,1,3,5), (2,1,2,2), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_1image_2o2i_img_dilated",(1,2,3,5), (2,2,2,2), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_1image_5o3i_img_dilated",(1,3,3,5), (5,3,2,2), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_8image_5o3i_img_dilated",(8,3,3,5), (5,3,2,2), (1,1), (1,1), (0,0), (0,0), (2,2), "true"),
("convolution_2d_8image_large_5o3i_img_dilated",
shaped_linspace((8,3,16,16)), shaped_linspace((5,3,2,2)), (1,1), (1,1), (0,0), (0,0), (2,2), "false"),
(8,3,16,16), (5,3,2,2), (1,1), (1,1), (0,0), (0,0), (2,2), "false"),
("convolution_2d_8image_large_5o3i_uneven_filter_img_dilated",
shaped_linspace((8,3,16,16)), shaped_linspace((5,3,2,3)), (1,1), (1,1), (0,0), (0,0), (2,2), "false"),
(8,3,16,16), (5,3,2,3), (1,1), (1,1), (0,0), (0,0), (2,2), "false"),
("convolution_2d_8image_large_5o3i_uneven_filter_uneven_img_dilation_img_dilated",
shaped_linspace((8,3,16,16)), shaped_linspace((5,3,2,3)), (1,1), (1,1), (0,0), (0,0), (2,3), "false"),
(8,3,16,16), (5,3,2,3), (1,1), (1,1), (0,0), (0,0), (2,3), "false"),
("convolution_3d_2image_large_5o3i_uneven_filter_uneven_img_dilation_img_dilated",
shaped_linspace((2,3,8,8,8)), shaped_linspace((5,3,2,3,4)), (1,1,1), (1,1,1), (0,0,0), (0,0,0), (2,3,2), "false"),
(2,3,8,8,8), (5,3,2,3,4), (1,1,1), (1,1,1), (0,0,0), (0,0,0), (2,3,2), "false"),
("convolution_3d_1image_large_5o3i_padded_uneven_filter_uneven_img_dilation_img_dilated",
shaped_linspace((1,3,8,8,8)), shaped_linspace((5,3,2,3,4)), (1,1,1), (1,1,1), (2,1,2), (1,2,3), (2,3,2), "false"),
(1,3,8,8,8), (5,3,2,3,4), (1,1,1), (1,1,1), (2,1,2), (1,2,3), (2,3,2), "false"),
("convolution_3d_2image_large_5o3i_padded_strided_uneven_filter_uneven_img_dilation_img_dilated",
shaped_linspace((2,3,8,8,8)), shaped_linspace((5,3,2,3,4)), (2,3,2), (1,1,1), (2,1,2), (1,2,3), (2,3,2), "false"),
(2,3,8,8,8), (5,3,2,3,4), (2,3,2), (1,1,1), (2,1,2), (1,2,3), (2,3,2), "false"),
("convolution_3d_2image_large_5o3i_padded_strided_uneven_filter_uneven_img_dilation_filter_dilated_img_dilated",
shaped_linspace((2,3,8,8,8)), shaped_linspace((5,3,2,3,4)), (2,3,2), (3,2,2), (2,1,2), (1,2,3), (2,3,2), "false"),
(2,3,8,8,8), (5,3,2,3,4), (2,3,2), (3,2,2), (2,1,2), (1,2,3), (2,3,2), "false"),
]
def main():
......@@ -301,6 +313,8 @@ def main():
f = open(sys.argv[1],'w')
f.write('''
// clang-format off
// ----------------------------------------------------------------------------
// Copyright 2017 Nervana Systems Inc.
// Licensed under the Apache License, Version 2.0 (the "License");
......@@ -322,8 +336,7 @@ def main():
// If you want to add new tests, you should edit test/ref_generators/generate_convolution_ref.py
// and regenerate this file.
//
// To regenerate (NOTE: this script will run apply-code-format.sh and reformat all source files
// in your tree):
// To regenerate:
//
// $ cd <ngraph source dir>/test
// $ ./update_reference.sh
......@@ -342,10 +355,11 @@ def main():
using namespace std;
using namespace ngraph;
static bool all_close_d(const std::vector<double>& a,
const std::vector<double>& b,
double rtol = 1e-5,
double atol = 1e-8)
template<typename T>
static bool all_close(const std::vector<T>& a,
const std::vector<T>& b,
T rtol = T(1e-4),
T atol = T(1e-7))
{
assert(a.size() == b.size());
......@@ -362,8 +376,14 @@ static bool all_close_d(const std::vector<double>& a,
}
''')
for t in tests:
emit_test(t,f)
f.write('''
// clang-format on
''')
f.close()
if __name__ == "__main__":
......
test/update_reference.sh
View file @ 2b0a5489
#!/bin/bash
declare THIS_SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
python ${THIS_SCRIPT_DIR}/ref_generators/generate_convolution_ref.py ${THIS_SCRIPT_DIR}/convolution_test.in.cpp
${THIS_SCRIPT_DIR}/../maint/apply-code-format.sh
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment