[Relay][TOPI]Fix meaning of conv2d_transpose output_padding parameter (#4318)
* Add output_padding to generic
* Add output_padding to the reference impl
* Add output_padding to arm_cpu
* Add output_padding to the test
* Add output_padding for cuda
* Add output_padding for x86
* Make use of the new output_padding argument in Relay
* Adjust conv2d_transpose Relay test
* Fix lint errors
* Fix the VTA declaration of conv2d_transpose
* support for output padding in conv2d transpose
* some output padding will break IR pass
* Fix new conv2d_transpose test
* Update tophub
* Fix conv1d output_padding too.
* Fix the conv1d_transpose reference function.
* Fix the cuda impl
* fix the topi test for conv1d
* Update the versions in tophub.py
Co-authored-by: Thierry Moreau <tmoreau@octoml.ai>
diff --git a/python/vta/top/vta_conv2d_transpose.py b/python/vta/top/vta_conv2d_transpose.py
index a2750dc..ff10ff0 100644
--- a/python/vta/top/vta_conv2d_transpose.py
+++ b/python/vta/top/vta_conv2d_transpose.py
@@ -27,24 +27,28 @@
from ..environment import get_env
@autotvm.register_topi_compute(topi.nn.conv2d_transpose_nchw, 'vta', 'direct')
-def _declatation_conv2d_transpose(cfg,
+def _declaration_conv2d_transpose(cfg,
data,
kernel,
strides,
padding,
- out_dtype):
+ out_dtype,
+ output_padding=(0, 0)):
ishape = get_const_tuple(data.shape)
kshape = get_const_tuple(kernel.shape)
b, c_i, i_h, i_w, t_b, t_ci = ishape
c_o, _, k_h, k_w, t_co, t_ci = kshape
stride_h, stride_w = strides
+ opad_h, opad_w = output_padding
+ # FIXME(tmoreau89): currently IR pass breaks when output padding != (0,0)
+ assert opad_h == 0 and opad_w == 0, "VTA does not support output padding for now"
# derive padding parameters
fpad_top, fpad_left, fpad_bottom, fpad_right = get_pad_tuple(padding, (k_h, k_w))
bpad_top = k_h - 1 - fpad_top
- bpad_bottom = k_h - 1 - fpad_bottom
+ bpad_bottom = k_h - 1 - fpad_bottom + opad_h
bpad_left = k_w - 1 - fpad_left
- bpad_right = k_w - 1 - fpad_right
+ bpad_right = k_w - 1 - fpad_right + opad_w
# padding stage
dilated_input = topi.nn.dilate(data, [1, 1, stride_h, stride_w, 1, 1])
@@ -53,8 +57,8 @@
[0, 0, bpad_bottom, bpad_right, 0, 0])
# convolution transpose stage
- out_h = (i_h - 1) * stride_h - fpad_top - fpad_bottom + k_h
- out_w = (i_w - 1) * stride_w - fpad_left - fpad_right + k_w
+ out_h = (i_h - 1) * stride_h - fpad_top - fpad_bottom + k_h + opad_h
+ out_w = (i_w - 1) * stride_w - fpad_left - fpad_right + k_w + opad_w
oshape = (b, c_o, out_h, out_w, t_b, t_co)
d_c = tvm.reduce_axis((0, c_i), name='d_c')
d_h = tvm.reduce_axis((0, k_h), name='d_h')
diff --git a/scripts/tune_conv2d_transpose.py b/scripts/tune_conv2d_transpose.py
index 3e51d41..fa9900a 100644
--- a/scripts/tune_conv2d_transpose.py
+++ b/scripts/tune_conv2d_transpose.py
@@ -33,13 +33,15 @@
Workload = namedtuple("Conv2DTransposeWorkload",
['batch', 'height', 'width', 'in_filter', 'out_filter',
- 'hkernel', 'wkernel', 'hpad', 'wpad', 'hstride', 'wstride'])
+ 'hkernel', 'wkernel', 'hpad', 'wpad', 'hstride', 'wstride',
+ 'o_hpad', 'o_wpad'])
+# DCGAN workloads
dcgan_wkls = [
# dcgan
- ('DCGAN.CT1', Workload(env.BATCH, 4, 4, 1024, 512, 4, 4, 1, 1, 2, 2)),
- ('DCGAN.CT2', Workload(env.BATCH, 8, 8, 512, 256, 4, 4, 1, 1, 2, 2)),
- ('DCGAN.CT3', Workload(env.BATCH, 16, 16, 256, 128, 4, 4, 1, 1, 2, 2)),
+ ('DCGAN.CT1', Workload(env.BATCH, 4, 4, 1024, 512, 4, 4, 1, 1, 2, 2, 0, 0)),
+ ('DCGAN.CT2', Workload(env.BATCH, 8, 8, 512, 256, 4, 4, 1, 1, 2, 2, 0, 0)),
+ ('DCGAN.CT3', Workload(env.BATCH, 16, 16, 256, 128, 4, 4, 1, 1, 2, 2, 0, 0)),
]
@tvm.tag_scope(tag=topi.tag.ELEMWISE)
@@ -51,7 +53,7 @@
x = tvm.compute(x.shape, lambda *i: tvm.max(x(*i), const_min), name="clipB")
return x
-def conv2d_transpose(N, CI, H, W, CO, KH, KW, strides, padding):
+def conv2d_transpose(N, CI, H, W, CO, KH, KW, strides, padding, opadding):
data_shape = (N//env.BATCH, CI//env.BLOCK_IN, H, W, env.BATCH, env.BLOCK_IN)
kernel_shape = (CO//env.BLOCK_OUT, CI//env.BLOCK_IN, KH, KW, env.BLOCK_OUT, env.BLOCK_IN)
@@ -64,7 +66,9 @@
Filter=kernel,
strides=strides,
padding=padding,
- out_dtype=env.acc_dtype)
+ out_dtype=env.acc_dtype,
+ output_padding=opadding
+ )
res = topi.right_shift(res, env.WGT_WIDTH)
res = my_clip(res, 0, (1 << env.OUT_WIDTH - 1) - 1)
res = topi.cast(res, env.out_dtype)
@@ -109,11 +113,12 @@
KW = wl.wkernel
strides = (wl.hstride, wl.wstride)
padding = (wl.hpad, wl.wpad)
+ opadding = (wl.o_hpad, wl.o_wpad)
# Create task
task = autotvm.task.create(
conv2d_transpose,
- args=(N, CI, H, W, CO, KH, KW, strides, padding),
+ args=(N, CI, H, W, CO, KH, KW, strides, padding, opadding),
target=tvm.target.vta(),
target_host=env.target_host,
template_key='direct')
diff --git a/tests/python/integration/test_benchmark_topi_conv2d_transpose.py b/tests/python/integration/test_benchmark_topi_conv2d_transpose.py
index e2601d1..235076c 100644
--- a/tests/python/integration/test_benchmark_topi_conv2d_transpose.py
+++ b/tests/python/integration/test_benchmark_topi_conv2d_transpose.py
@@ -37,7 +37,8 @@
Workload = namedtuple("Conv2DTransposeWorkload",
['batch', 'height', 'width', 'in_filter', 'out_filter',
- 'hkernel', 'wkernel', 'hpad', 'wpad', 'hstride', 'wstride'])
+ 'hkernel', 'wkernel', 'hpad', 'wpad', 'hstride', 'wstride',
+ 'o_hpad', 'o_wpad'])
# Get batch info from env
env = vta.get_env()
@@ -45,9 +46,9 @@
# DCGAN workloads
dcgan_wklds = [
# dcgan
- ('DCGAN.CT1', Workload(env.BATCH, 4, 4, 1024, 512, 4, 4, 1, 1, 2, 2)),
- ('DCGAN.CT2', Workload(env.BATCH, 8, 8, 512, 256, 4, 4, 1, 1, 2, 2)),
- ('DCGAN.CT3', Workload(env.BATCH, 16, 16, 256, 128, 4, 4, 1, 1, 2, 2)),
+ ('DCGAN.CT1', Workload(env.BATCH, 4, 4, 1024, 512, 4, 4, 1, 1, 2, 2, 0, 0)),
+ ('DCGAN.CT2', Workload(env.BATCH, 8, 8, 512, 256, 4, 4, 1, 1, 2, 2, 0, 0)),
+ ('DCGAN.CT3', Workload(env.BATCH, 16, 16, 256, 128, 4, 4, 1, 1, 2, 2, 0, 0)),
]
# FIXME: we need a custom clip operator to circumvent a pattern detection limitation
@@ -102,7 +103,8 @@
# Define base computation schedule
with target:
res = topi.nn.conv2d_transpose_nchw(
- data, kernel, (wl.hstride, wl.wstride), (wl.hpad, wl.wpad), env.acc_dtype)
+ data, kernel, (wl.hstride, wl.wstride),
+ (wl.hpad, wl.wpad), env.acc_dtype, (wl.o_hpad, wl.o_wpad))
res = topi.right_shift(res, env.WGT_WIDTH)
res = my_clip(res, 0, (1 << env.OUT_WIDTH - 1) - 1)
res = topi.cast(res, env.out_dtype)
@@ -112,8 +114,8 @@
print(vta.lower(s, [data, kernel, res], simple_mode=True))
# Derive number of ops
- fout_height = (wl.height - 1) * wl.hstride - 2 * wl.hpad + wl.hkernel
- fout_width = (wl.width - 1) * wl.wstride - 2 * wl.wpad + wl.wkernel
+ fout_height = (wl.height - 1) * wl.hstride - 2 * wl.hpad + wl.hkernel + wl.o_hpad
+ fout_width = (wl.width - 1) * wl.wstride - 2 * wl.wpad + wl.wkernel + wl.o_wpad
num_ops = 2 * wl.batch * fout_height * fout_width * wl.hkernel * wl.wkernel * wl.out_filter * wl.in_filter
# @memoize("vta.tests.test_benchmark_topi.conv2d.verify_nhwc")
