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

[Relay] add ShapeFunc for tanh #6898

Merged
merged 3 commits into from
Mar 8, 2021
Merged

[Relay] add ShapeFunc for tanh #6898

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a2cbc8b
add ShapeFunc for tanh
monklof Nov 11, 2020
aadf4c9
_schedule_dense_small_batch turn autotvm off when dense's inner dim i…
monklof Jan 22, 2021
ee1a749
fix CI pylint
monklof Feb 23, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions python/tvm/relay/op/_tensor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -276,3 +276,4 @@ def elemwise_shape_func(attrs, inputs, _):
register_shape_func("clip", False, elemwise_shape_func)
register_shape_func("log2", False, elemwise_shape_func)
register_shape_func("sigmoid", False, elemwise_shape_func)
register_shape_func("tanh", False, elemwise_shape_func)
25 changes: 19 additions & 6 deletions python/tvm/topi/cuda/dense.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,13 +77,26 @@ def _callback(op):


def _schedule_dense_small_batch(cfg, s, C):
A, _ = C.op.input_tensors
_, in_dim = get_const_tuple(A.shape)
cfg.define_split("tile_k", in_dim, num_outputs=2)
if cfg.is_fallback:
cfg["tile_k"] = SplitEntity([-1, 64] if in_dim > 64 else [1, 64])
A, weights = C.op.input_tensors
icemelon marked this conversation as resolved.
Show resolved Hide resolved
_, in_dim_weights = get_const_tuple(weights.shape)
_, in_dim_A = get_const_tuple(A.shape)

if isinstance(in_dim_A, int):
in_dim = in_dim_A
elif isinstance(in_dim_weights, int):
in_dim = in_dim_weights
else:
in_dim = None

if in_dim is not None:
cfg.define_split("tile_k", in_dim, num_outputs=2)
if cfg.is_fallback:
cfg["tile_k"] = SplitEntity([-1, 64] if in_dim > 64 else [1, 64])
_, kf = cfg["tile_k"].apply(s, C, C.op.reduce_axis[0])
else:
tile_k = 64
_, kf = s[C].split(C.op.reduce_axis[0], tile_k)

_, kf = cfg["tile_k"].apply(s, C, C.op.reduce_axis[0])
CF = s.rfactor(C, kf)

if C.op in s.outputs:
Expand Down