Manual Resharding Handler #3398
Conversation
meta-cla
bot
added
the
CLA Signed
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
Summary: Pull Request resolved: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Differential Revision: D83392188
aporialiao
force-pushed
the
export-D83392188
branch
from
d1594d0 to
a1dd8d8
Compare
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
2 similar comments
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
Summary: Pull Request resolved: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Differential Revision: D83392188
aporialiao
force-pushed
the
export-D83392188
branch
from
a1dd8d8 to
5bf0303
Compare
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
Summary: Pull Request resolved: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Differential Revision: D83392188
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
aporialiao
force-pushed
the
export-D83392188
branch
from
5bf0303 to
c096241
Compare
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
Summary: X-link: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Differential Revision: D83392188
aporialiao
force-pushed
the
export-D83392188
branch
from
c096241 to
7f4cafb
Compare
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating diff in D83392188. |
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83392188. |
Summary: Pull Request resolved: meta-pytorch#3398 X-link: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Differential Revision: D83392188
aporialiao
force-pushed
the
export-D83392188
branch
from
7f4cafb to
5af1b1b
Compare
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83392188. |
1 similar comment
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83392188. |
aporialiao
force-pushed
the
export-D83392188
branch
from
5af1b1b to
f9ed566
Compare
Summary: Pull Request resolved: meta-pytorch#3398 X-link: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Reviewed By: isururanawaka Differential Revision: D83392188
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83392188. |
1 similar comment
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83392188. |
Summary: Pull Request resolved: meta-pytorch#3398 X-link: meta-pytorch#3398 * `DistributedModelParallel`: stores sharders passed into planner, this will be reused at resharding * Removed hardcoding for module FQNs and plan keys since not compatible with all models. Leveraging how planner/enumerator identifies sharders/EBCs to find the right plan Reviewed By: isururanawaka Differential Revision: D83392188
aporialiao
force-pushed
the
export-D83392188
branch
from
f9ed566 to
43315c0
Compare
|
@aporialiao has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83392188. |
Summary:
DistributedModelParallel: stores sharders passed into planner, this will be reused at reshardingDifferential Revision: D83392188