comaniac commented on a change in pull request #10036:
URL: https://github.com/apache/tvm/pull/10036#discussion_r791069518
##########
File path: python/tvm/contrib/cutlass/build.py
##########
@@ -272,7 +274,13 @@ def handle_conv2d(
def tune_cutlass_kernels(
- mod, sm, use_3xtf32=True, profile_all=True, use_multiprocessing=False,
tmp_dir="./tmp"
+ mod,
+ sm,
+ use_3xtf32=True,
+ profile_all_alignments=False,
+ profile_all=True,
Review comment:
It seems to me that `profile_all_alignment` and `profile_all` may
confuse people. Maybe changing to the following?
- `profile_all_alignment=False` -> `include_all_alighments=False`: Later if
we have more options, we may consider having a unified `candidate_config` to
include all rules.
- `profile_all=True` -> `find_first_valid=False`: This name might be clearer.
##########
File path: python/tvm/contrib/cutlass/build.py
##########
@@ -272,7 +274,13 @@ def handle_conv2d(
def tune_cutlass_kernels(
- mod, sm, use_3xtf32=True, profile_all=True, use_multiprocessing=False,
tmp_dir="./tmp"
+ mod,
+ sm,
+ use_3xtf32=True,
+ profile_all_alignments=False,
+ profile_all=True,
Review comment:
Sure I'm fine with that. It's also fine to call it `profile_configs` in
the future once we want to expose more configs to users.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
