mirrors
/
mmcv
mirror of https://github.com/open-mmlab/mmcv.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

[Enhancement] support dipu_mock_cuda=False in dipu for mmcv ext ops with cpu fallback (#2839) 

* Support cuda_mock

* Fix

* Support roi_align

* Fix lint

* Support bbox
pull/2851/head
CokeDong 2023-06-20 17:22:11 +08:00 committed by GitHub
parent fc038a386a
commit 99a8d05766
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 11 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
mmcv/ops/csrc/pytorch/bbox_overlaps.cpp
View File

@ -33,7 +33,9 @@ void bbox_overlaps_diopi(const Tensor bboxes1, const Tensor bboxes2,
diopiContextHandle_t ch = &ctx; diopiContextHandle_t ch = &ctx;
auto bboxes2_p = toDiopiTensorHandle(bboxes2); auto bboxes2_p = toDiopiTensorHandle(bboxes2);
auto ious_p = toDiopiTensorHandle(ious); auto ious_p = toDiopiTensorHandle(ious);
if (reinterpret_cast<void *>(diopiBboxOverlapsMmcv) != nullptr) { bool is_mock_cuda = bboxes1.device().type() == c10::DeviceType::PrivateUse1;
if (is_mock_cuda &&
reinterpret_cast<void *>(diopiBboxOverlapsMmcv) != nullptr) {
auto ret = diopiBboxOverlapsMmcv(ch, ious_p, bboxes1_p, bboxes2_p, mode, auto ret = diopiBboxOverlapsMmcv(ch, ious_p, bboxes1_p, bboxes2_p, mode,
offset, aligned); offset, aligned);
if (ret == diopiSuccess) return; if (ret == diopiSuccess) return;

3
mmcv/ops/csrc/pytorch/nms.cpp
View File

@ -42,7 +42,8 @@ Tensor nms_diopi(Tensor boxes, Tensor scores, float iou_threshold, int offset) {
auto outp = toDiopiTensorHandle(out); auto outp = toDiopiTensorHandle(out);
diopiTensorHandle_t* outhandle = &outp; diopiTensorHandle_t* outhandle = &outp;
auto scores_p = toDiopiTensorHandle(scores); auto scores_p = toDiopiTensorHandle(scores);
if (reinterpret_cast<void*>(diopiNmsMmcv) != nullptr) { bool is_mock_cuda = boxes.device().type() == c10::DeviceType::PrivateUse1;
if (is_mock_cuda && reinterpret_cast<void*>(diopiNmsMmcv) != nullptr) {
auto ret = auto ret =
diopiNmsMmcv(ch, outhandle, boxes_p, scores_p, iou_threshold, offset); diopiNmsMmcv(ch, outhandle, boxes_p, scores_p, iou_threshold, offset);
if (ret == diopiSuccess) { if (ret == diopiSuccess) {

8
mmcv/ops/csrc/pytorch/roi_align.cpp
View File

@ -53,7 +53,8 @@ void roi_align_forward_diopi(Tensor input, Tensor rois, Tensor output,
auto out_p = toDiopiTensorHandle(output); auto out_p = toDiopiTensorHandle(output);
auto argmax_y_p = toDiopiTensorHandle(argmax_y); auto argmax_y_p = toDiopiTensorHandle(argmax_y);
auto argmax_x_p = toDiopiTensorHandle(argmax_x); auto argmax_x_p = toDiopiTensorHandle(argmax_x);
if (reinterpret_cast<void*>(diopiRoiAlignMmcv) != nullptr) { bool is_mock_cuda = input.device().type() == c10::DeviceType::PrivateUse1;
if (is_mock_cuda && reinterpret_cast<void*>(diopiRoiAlignMmcv) != nullptr) {
auto ret = diopiRoiAlignMmcv( auto ret = diopiRoiAlignMmcv(
ch, out_p, argmax_y_p, argmax_x_p, input_p, rois_p, aligned_height, ch, out_p, argmax_y_p, argmax_x_p, input_p, rois_p, aligned_height,
aligned_width, sampling_ratio, pool_mode, spatial_scale, aligned); aligned_width, sampling_ratio, pool_mode, spatial_scale, aligned);
@ -91,7 +92,10 @@ void roi_align_backward_diopi(Tensor grad_output, Tensor rois, Tensor argmax_y,
auto grad_input_ = toDiopiTensorHandle(grad_input); auto grad_input_ = toDiopiTensorHandle(grad_input);
diopiContext ctx(dipu::getCurrentDIPUStream().rawstream()); diopiContext ctx(dipu::getCurrentDIPUStream().rawstream());
diopiContextHandle_t ch = &ctx; diopiContextHandle_t ch = &ctx;
if (reinterpret_cast<void*>(diopiRoiAlignBackwardMmcv) != nullptr) { bool is_mock_cuda =
grad_output.device().type() == c10::DeviceType::PrivateUse1;
if (is_mock_cuda &&
reinterpret_cast<void*>(diopiRoiAlignBackwardMmcv) != nullptr) {
auto ret = diopiRoiAlignBackwardMmcv(ch, grad_input_, grad_output_, rois_, auto ret = diopiRoiAlignBackwardMmcv(ch, grad_input_, grad_output_, rois_,
argmax_y_, argmax_x_, aligned_height, argmax_y_, argmax_x_, aligned_height,
aligned_width, sampling_ratio, aligned_width, sampling_ratio,