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

Comm broadcast #6213

Merged
merged 18 commits into from
Sep 9, 2021
Merged

Comm broadcast #6213

merged 18 commits into from
Sep 9, 2021

Conversation

liufengwei0103
Copy link
Contributor

No description provided.

@liufengwei0103 liufengwei0103 changed the base branch from master to add_allgather September 9, 2021 06:43
clackhan
clackhan reviewed Sep 9, 2021
View reviewed changes
oneflow/core/framework/op_interpreter/eager_mirrored_op_interpreter.cpp Outdated
@@ -280,7 +272,7 @@ Maybe<Tensor> GetSyncedTensorIfBroadcast(const std::shared_ptr<Tensor>& tensor,
JUST(GetTensorDevice4CurrentProcessCtx(parallel_desc, &parallel_id));
if (!parallel_id.has_value()) { return tensor; }
const auto& broadcast_parallel_desc = JUST(GetBroadcastSubParallelDesc(parallel_desc, nd_sbp));
return Broadcast(tensor, broadcast_parallel_desc, false);
return Broadcast(tensor, /* rank id */ 0, broadcast_parallel_desc, false);
Copy link
Contributor

Choose a reason for hiding this comment

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

/* root */ 0

Copy link
Contributor Author

Choose a reason for hiding this comment

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

/* root */ 0

已改

Base automatically changed from add_allgather to master September 9, 2021 09:36
@oneflow-ci-bot oneflow-ci-bot removed their request for review September 9, 2021 11:04
@oneflow-ci-bot oneflow-ci-bot requested review from oneflow-ci-bot and removed request for oneflow-ci-bot September 9, 2021 13:10
@oneflow-ci-bot oneflow-ci-bot requested review from oneflow-ci-bot and removed request for oneflow-ci-bot September 9, 2021 14:04
@github-actions
Copy link
Contributor

github-actions bot commented Sep 9, 2021

Speed stats: 
GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 128.4ms (= 6419.3ms / 50, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.2ms (= 7107.7ms / 50, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.11 (= 142.2ms / 128.4ms)

OneFlow resnet50 time: 74.8ms (= 3737.7ms / 50, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 86.0ms (= 4301.8ms / 50, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.15 (= 86.0ms / 74.8ms)

OneFlow resnet50 time: 48.4ms (= 2420.0ms / 50, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.1ms (= 2906.4ms / 50, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.20 (= 58.1ms / 48.4ms)

OneFlow resnet50 time: 43.4ms (= 2172.1ms / 50, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 50.4ms (= 2522.1ms / 50, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.16 (= 50.4ms / 43.4ms)

OneFlow resnet50 time: 39.3ms (= 1966.5ms / 50, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.2ms (= 1908.0ms / 50, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.97 (= 38.2ms / 39.3ms)

OneFlow resnet50 time: 155.2ms (= 7757.7ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 157.4ms (= 7870.8ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.01 (= 157.4ms / 155.2ms)

OneFlow resnet50 time: 99.6ms (= 4980.8ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.8ms (= 5239.4ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.05 (= 104.8ms / 99.6ms)

OneFlow resnet50 time: 77.9ms (= 3895.3ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.1ms (= 3904.8ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.00 (= 78.1ms / 77.9ms)

OneFlow resnet50 time: 76.6ms (= 3828.9ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 64.2ms (= 3211.6ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.84 (= 64.2ms / 76.6ms)

OneFlow resnet50 time: 70.5ms (= 3523.2ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 62.8ms (= 3139.3ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.89 (= 62.8ms / 70.5ms)

@oneflow-ci-bot oneflow-ci-bot merged commit cf515ce into master Sep 9, 2021
@oneflow-ci-bot oneflow-ci-bot deleted the comm_broadcast branch September 9, 2021 16:59
@liufengwei0103 liufengwei0103 mentioned this pull request Sep 12, 2021
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@clackhan clackhan clackhan approved these changes

@oneflow-ci-bot oneflow-ci-bot Awaiting requested review from oneflow-ci-bot

Assignees
No one assigned
Labels
automerge feature system
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@liufengwei0103 @clackhan @oneflow-ci-bot