def test_conv2d():
# im2col and np_conv2d are helper functions
def im2col(X, filter_H, filter_W, padding, stride):
N, C, H, W = X.shape
assert (H + 2 * padding - filter_H) % stride == 0
assert (W + 2 * padding - filter_W) % stride == 0
out_H = (H + 2 * padding - filter_H) / stride + 1
out_W = (W + 2 * padding - filter_W) / stride + 1
y_row_size = C * filter_H * filter_W
y_col_size = out_H * out_W
y_shape = (N, y_row_size, y_col_size)
Y = np.empty(y_shape, dtype = X.dtype)
for batch_index in range(N):
for col_index in range(y_col_size):
out_y = col_index / out_W
out_x = col_index % out_W
in_y = out_y * stride - padding
in_x = out_x * stride - padding
row_idx = 0
for c in range(0, C):
for y in range(in_y, in_y + filter_H):
for x in range(in_x, in_x + filter_W):
if (x < 0 or x >= W or y < 0 or y >= H):
Y[batch_index, row_idx, col_index] = 0
else:
Y[batch_index, row_idx, col_index] = X[batch_index, c, y, x]
row_idx += 1
return Y
def np_conv2d(X, Filter, padding=0, stride=1):
"""Implement a conv2d as a matrix multiply after im2col."""
filter_outChannel, filter_inChannel, filter_H, filter_W = Filter.shape
N, C, H, W = X.shape
assert (H + 2 * padding - filter_H) % stride == 0
assert (W + 2 * padding - filter_W) % stride == 0
out_H = (H + 2 * padding - filter_H) / stride + 1
out_W = (W + 2 * padding - filter_W) / stride + 1
im2col_matrix = im2col(X, filter_H, filter_W, padding, stride)
filter_matrix = Filter.reshape(filter_outChannel, -1)
return np.matmul(filter_matrix, im2col_matrix).reshape(N, filter_outChannel, out_H, out_W)
shapeX = (100, 3, 28, 28)
shapeF = (10, 3, 5, 5)
shapeY = (100, 10, 24, 24)
x = np.random.uniform(0, 10, size=shapeX).astype(dtype)
f = np.random.uniform(0, 10, size=shapeF).astype(dtype)
y = np.zeros(shapeY).astype(dtype)
arr_x = tvm.nd.array(x, ctx=ctx)
arr_f = tvm.nd.array(f, ctx=ctx)
arr_y = tvm.nd.array(y, ctx=ctx)
conv2d = tvm_op.make_conv2d(shapeX, shapeF, tgt, tgt_host, "conv2d")
conv2d(arr_x, arr_f, arr_y)
y = arr_y.asnumpy()
np.testing.assert_allclose(np_conv2d(x, f), y, rtol=1e-5)
def test_conv2d():
# im2col and np_conv2d are helper functions
def im2col(X, filter_H, filter_W, padding, stride):
N, C, H, W = X.shape
assert (H + 2 * padding - filter_H) % stride == 0
assert (W + 2 * padding - filter_W) % stride == 0
out_H = (H + 2 * padding - filter_H) / stride + 1
out_W = (W + 2 * padding - filter_W) / stride + 1
y_row_size = int(C * filter_H * filter_W)
y_col_size = int(out_H * out_W)
y_shape = (N, y_row_size, y_col_size)
Y = np.empty(y_shape, dtype = X.dtype)
for batch_index in range(N):
for col_index in range(y_col_size):
out_y = col_index / out_W
out_x = col_index % out_W
in_y = int(out_y * stride - padding)
in_x = int(out_x * stride - padding)
row_idx = 0
for c in range(0, C):
for y in range(in_y, in_y + filter_H):
for x in range(in_x, in_x + filter_W):
if (x < 0 or x >= W or y < 0 or y >= H):
Y[batch_index, row_idx, col_index] = 0
else:
Y[batch_index, row_idx, col_index] = X[batch_index, c, y, x]
row_idx += 1
return Y
def np_conv2d(X, Filter, padding=0, stride=1):
"""Implement a conv2d as a matrix multiply after im2col."""
filter_outChannel, filter_inChannel, filter_H, filter_W = Filter.shape
N, C, H, W = X.shape
assert (H + 2 * padding - filter_H) % stride == 0
assert (W + 2 * padding - filter_W) % stride == 0
out_H = int((H + 2 * padding - filter_H) / stride + 1)
out_W = int((W + 2 * padding - filter_W) / stride + 1)
im2col_matrix = im2col(X, filter_H, filter_W, padding, stride)
filter_matrix = Filter.reshape(filter_outChannel, -1)
return np.matmul(filter_matrix, im2col_matrix).reshape(N, filter_outChannel, out_H, out_W)
shapeX = (100, 3, 28, 28)
shapeF = (10, 3, 5, 5)
shapeY = (100, 10, 24, 24)
x = np.random.uniform(0, 10, size=shapeX).astype(dtype)
f = np.random.uniform(0, 10, size=shapeF).astype(dtype)
y = np.zeros(shapeY).astype(dtype)
arr_x = tvm.nd.array(x, ctx=ctx)
arr_f = tvm.nd.array(f, ctx=ctx)
arr_y = tvm.nd.array(y, ctx=ctx)
conv2d = tvm_op.make_conv2d(shapeX, shapeF, tgt, tgt_host, "conv2d")
conv2d(arr_x, arr_f, arr_y)
y = arr_y.asnumpy()
np.testing.assert_allclose(np_conv2d(x, f), y, rtol=1e-5)
def test_conv2d_small():
shapeX = (2, 3, 4, 4)
shapeF = (4, 3, 3, 3)
shapeY = (2, 4, 2, 2)
x = np.random.uniform(0, 10, size=shapeX).astype(dtype)
f = np.random.uniform(0, 10, size=shapeF).astype(dtype)
y = np.zeros(shapeY).astype(dtype)
arr_x = tvm.nd.array(x, ctx=ctx)
arr_f = tvm.nd.array(f, ctx=ctx)
arr_y = tvm.nd.array(y, ctx=ctx)
conv2d = tvm_op.make_conv2d(shapeX, shapeF, tgt, tgt_host, "conv2d")
conv2d(arr_x, arr_f, arr_y)
y = arr_y.asnumpy()
np.testing.assert_allclose(y, np_conv2d(x, f), rtol=1e-5)
def test_conv2d():
shapeX = (100, 3, 28, 28)
shapeF = (10, 3, 5, 5)
shapeY = (100, 10, 24, 24)
x = np.random.uniform(0, 10, size=shapeX).astype(dtype)
f = np.random.uniform(0, 10, size=shapeF).astype(dtype)
y = np.zeros(shapeY).astype(dtype)
arr_x = tvm.nd.array(x, ctx=ctx)
arr_f = tvm.nd.array(f, ctx=ctx)
arr_y = tvm.nd.array(y, ctx=ctx)
conv2d = tvm_op.make_conv2d(shapeX, shapeF, tgt, tgt_host, "conv2d")
conv2d(arr_x, arr_f, arr_y)
y = arr_y.asnumpy()
np.testing.assert_allclose(np_conv2d(x, f), y, rtol=1e-5)