-
-
Notifications
You must be signed in to change notification settings - Fork 719
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.
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
Fix import error when distributed.rmm.pool-size is setted #6482
Fix import error when distributed.rmm.pool-size is setted #6482
Conversation
Can one of the admins verify this patch? |
The linting is failed, so I try to fix. |
61e0539
to
21e2033
Compare
Unit Test Results 15 files ±0 15 suites ±0 6h 27m 37s ⏱️ + 18m 15s For more details on these failures, see this check. Results for commit 6575251. ± Comparison against base commit c2b28cf. ♻️ This comment has been updated with latest results. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @KoyamaSohei! @pentschev @charlesbluca -- do either of you have a moment to look at this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I would prefer to handle this differently, as per my suggestions. When the user is requesting an initial pool size, one must assume that is intended, otherwise we just silently do nothing which can be difficult to debug later.
distributed/comm/ucx.py
Outdated
|
||
pool_size_str = dask.config.get("distributed.rmm.pool-size") | ||
if pool_size_str is not None: | ||
pool_size = parse_bytes(pool_size_str) | ||
rmm.reinitialize( | ||
pool_allocator=True, managed_memory=False, initial_pool_size=pool_size | ||
) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
pool_size_str = dask.config.get("distributed.rmm.pool-size") | |
if pool_size_str is not None: | |
pool_size = parse_bytes(pool_size_str) | |
rmm.reinitialize( | |
pool_allocator=True, managed_memory=False, initial_pool_size=pool_size | |
) |
Thank you, @pentschev !!! But calling |
I added that warning message on ignoring |
Fair enough. GitHub won't allow me to suggest changes beyond the scope of the PR, but I would suggest a couple more things:
|
c4dd066
to
6575251
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM now, thanks @KoyamaSohei !
Thanks @pentschev for the review :) |
Thanks all merging in. |
Hi,
I try to use dask-cuda with UCX integration for our research.
On executing the following commands,
An error has occurred on
rmm.reinitialize()
.This happened when
numba
is installed andrmm
is not installed.Because the previous process (set
device_array
) just checked whetherrmm
ornumba
is installed.This seems to be solved just by
import rmm
.stack trace
Regards,