Skip to content

[Feature] Support engine with NPU backend. #2262

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

Merged
zhouzaida merged 23 commits into from
Sep 30, 2022
Merged

[Feature] Support engine with NPU backend. #2262

Changes from 1 commit
Commits
Show all changes
23 commits
Select commit Hold shift + click to select a range
093f462
init npu
wangjiangben-hw Sep 14, 2022
ba9c36b
Merge pull request #2261 from wangjiangben-hw/npu-dev
ckirchhoff2021 Sep 15, 2022
716b3b3
add npu extension and focal loss adapter
ckirchhoff2021 Sep 19, 2022
6e53b3f
clean code
ckirchhoff2021 Sep 20, 2022
08f0a16
clean code
ckirchhoff2021 Sep 20, 2022
da659cb
clean code
ckirchhoff2021 Sep 20, 2022
448476e
clean code
ckirchhoff2021 Sep 20, 2022
afbd351
fix autocast bugs on npu (#2273)
wangjiangben-hw Sep 20, 2022
26f35e0
code format
ckirchhoff2021 Sep 20, 2022
58618ac
code format
ckirchhoff2021 Sep 20, 2022
d75dcb1
code format
ckirchhoff2021 Sep 21, 2022
7af2a6f
bug fix
ckirchhoff2021 Sep 21, 2022
268ff0e
pytorch_npu_helper.hpp clean code
ckirchhoff2021 Sep 21, 2022
9b0133c
Merge pull request #2269 from ckirchhoff2021/npu-dev
ckirchhoff2021 Sep 28, 2022
a043541
Npu dev (#2306)
wangjiangben-hw Sep 29, 2022
90fc3dc
raise ImportError when compile with npu
wangjiangben-hw Sep 29, 2022
6ffdb57
add npu test case (#2307)
wangjiangben-hw Sep 29, 2022
3f11861
Update focal_loss.py
wangjiangben-hw Sep 29, 2022
c81daa3
add comment
wangjiangben-hw Sep 30, 2022
1cee865
clean lint
wangjiangben-hw Sep 30, 2022
5841f92
update dtype assert
wangjiangben-hw Sep 30, 2022
4ce938c
update DDP forward and comment
wangjiangben-hw Sep 30, 2022
ea1d8f8
fix bug
wangjiangben-hw Sep 30, 2022
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
Prev Previous commit
Next Next commit
clean code
  • Loading branch information
@ckirchhoff2021
ckirchhoff2021 committed Sep 20, 2022
commit 08f0a16a8ae227bd23866d7b6f87bc415d090758
16 changes: 8 additions & 8 deletions mmcv/ops/csrc/pytorch/npu/focal_loss_npu.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@ using namespace NPU_NAME_SPACE;
using namespace std;


void sigmoid_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
void sigmoid_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
Tensor output, float gamma, float alpha) {

at::Tensor target_y = at::reshape(target, input.sizes());
Expand All @@ -30,10 +30,10 @@ void sigmoid_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
.Run();
}

void sigmoid_focal_loss_forward_impl(Tensor input, Tensor target, Tensor weight,
void sigmoid_focal_loss_forward_impl(Tensor input, Tensor target, Tensor weight,
Tensor output, float gamma, float alpha);

void sigmoid_focal_loss_backward_npu(Tensor input, Tensor target, Tensor weight,
void sigmoid_focal_loss_backward_npu(Tensor input, Tensor target, Tensor weight,
Tensor grad_input, float gamma, float alpha) {

at::Tensor target_y = at::reshape(target, input.sizes());
Expand All @@ -58,10 +58,10 @@ void sigmoid_focal_loss_backward_npu(Tensor input, Tensor target, Tensor weight,
.Run();
}

void sigmoid_focal_loss_backward_impl(Tensor input, Tensor target, Tensor weight,
void sigmoid_focal_loss_backward_impl(Tensor input, Tensor target, Tensor weight,
Tensor grad_input, float gamma, float alpha);

void softmax_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
void softmax_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
Tensor output, float gamma, float alpha) {

int64_t n_class = input.size(1);
Expand All @@ -73,7 +73,7 @@ void softmax_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
if(weight_size > 0) {
weight_y = at_npu::native::NPUNativeFunctions::npu_broadcast(weight, input.sizes());
}

OpCommand cmd;
cmd.Name("SoftmaxFocalLoss")
.Input(input)
Expand All @@ -87,7 +87,7 @@ void softmax_focal_loss_forward_npu(Tensor input, Tensor target, Tensor weight,
.Run();
}

void softmax_focal_loss_forward_impl(Tensor input, Tensor target, Tensor weight,
void softmax_focal_loss_forward_impl(Tensor input, Tensor target, Tensor weight,
Tensor grad_input, float gamma, float alpha);

void softmax_focal_loss_backward_npu(Tensor input, Tensor target, Tensor weight, Tensor buff,
Expand All @@ -102,7 +102,7 @@ void softmax_focal_loss_backward_npu(Tensor input, Tensor target, Tensor weight,
if(weight_size > 0) {
weight_y = at_npu::native::NPUNativeFunctions::npu_broadcast(weight, input.sizes());
}

OpCommand cmd;
cmd.Name("SoftmaxFocalLossGrad")
.Input(input)
Expand Down