Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Bugfix][TOPI] conv2d_transpose bugfix #3138

Merged
merged 2 commits into from
May 8, 2019
Merged

[Bugfix][TOPI] conv2d_transpose bugfix #3138

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3f35e10
deconv tests
Laurawly May 1, 2019
f8de3c2
deconv bug fixed for certain cases tests added
Laurawly May 6, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
9 changes: 4 additions & 5 deletions topi/python/topi/cuda/conv2d_transpose_nchw.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -174,7 +174,6 @@ def _callback(op):
by, vy, ty, yi = cfg["tile_y"].apply(s, output, y)
bx, vx, tx, xi = cfg["tile_x"].apply(s, output, x)

bf = s[output].fuse(n, bf)
s[output].bind(bf, tvm.thread_axis("blockIdx.z"))
s[output].bind(by, tvm.thread_axis("blockIdx.y"))
s[output].bind(bx, tvm.thread_axis("blockIdx.x"))
Expand All @@ -184,7 +183,7 @@ def _callback(op):
s[output].bind(tf, tvm.thread_axis("threadIdx.z"))
s[output].bind(ty, tvm.thread_axis("threadIdx.y"))
s[output].bind(tx, tvm.thread_axis("threadIdx.x"))
s[output].reorder(bf, by, bx, vf, vy, vx, tf, ty, tx, fi, yi, xi)
s[output].reorder(n, bf, by, bx, vf, vy, vx, tf, ty, tx, fi, yi, xi)
s[OL].compute_at(s[output], tx)

# tile reduction axes
Expand All @@ -193,13 +192,13 @@ def _callback(op):
rco, rcm, rci = cfg['tile_rc'].apply(s, OL, rc)
s[OL].reorder(rco, rcm, ry, rx, rci, n, f, y, x)

s[AA].compute_at(s[OL], rcm)
s[WW].compute_at(s[OL], rcm)
s[AA].compute_at(s[OL], rx)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@Laurawly Why cannot we compute at rcm？

Copy link
Contributor Author

@Laurawly Laurawly May 12, 2019 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We get blockidx.x not found error in certain cases such as the test I added in this pr

s[WW].compute_at(s[OL], rx)

# cooperative fetching
for load in [AA, WW]:
n, f, y, x = s[load].op.axis
fused = s[load].fuse(n, f, y, x)
fused = s[load].fuse(f, y, x)
tz, fused = s[load].split(fused, nparts=cfg["tile_f"].size[2])
ty, fused = s[load].split(fused, nparts=cfg["tile_y"].size[2])
tx, fused = s[load].split(fused, nparts=cfg["tile_x"].size[2])
Expand Down
1 change: 1 addition & 0 deletions topi/tests/python/test_topi_conv2d_transpose_nchw.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,6 +74,7 @@ def check_device(device):
def test_conv2d_transpose_nchw():
verify_conv2d_transpose_nchw(1, 3, 224, 32, 3, 1, 0)
verify_conv2d_transpose_nchw(1, 3, 224, 32, 3, 2, 1)
verify_conv2d_transpose_nchw(1, 3, 224, 32, 2, 2, 0)
verify_conv2d_transpose_nchw(1, 32, 32, 128, 5, 1, 0)
verify_conv2d_transpose_nchw(1, 32, 32, 128, 5, 2, 1)

Expand Down