You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@MaybeShewill-CV Hi.
I am planning to use MNN-LaneNet with VULKAN backend. Is it possible to make this repo with VULKAN backend ? I read cmakelists.txt but there is no any thing about vulkan backend.
I built MNN project and used libMNN.so in my MNN_LaneNet project but MNN_LaneNet is slower when i use VULKAN backend.
with mnn_config.type = MNN_FORWARD_CPU result :
detect, 211, cost time: 0.246000 ms
detect, 146, cost time: 0.838000 ms
detect, 153, cost time: 872.728027 ms
detect, 204, cost time: 29.430002 ms
with mnn_config.type = MNN_FORWARD_VULKAN result :
Vulkan don't support 2, ArgMax: lanenet_model/vgg_backend/binary_seg/ArgMax
Vulkan don't support 9, Cast: lanenet/Cast
detect, 146, cost time: 0.755000 ms
detect, 153, cost time: 2487.019043 ms
detect, 204, cost time: 31.821001 ms
The text was updated successfully, but these errors were encountered:
@naserpiltan You'd better raise an issue in MNN project where you may get a more satisfied answer. Actually I'm not a engineer major in Ai frameworks:)
@MaybeShewill-CV Hi.
I am planning to use MNN-LaneNet with VULKAN backend. Is it possible to make this repo with VULKAN backend ? I read cmakelists.txt but there is no any thing about vulkan backend.
I built MNN project and used libMNN.so in my MNN_LaneNet project but MNN_LaneNet is slower when i use VULKAN backend.
with mnn_config.type = MNN_FORWARD_CPU result :
detect, 211, cost time: 0.246000 ms
detect, 146, cost time: 0.838000 ms
detect, 153, cost time: 872.728027 ms
detect, 204, cost time: 29.430002 ms
with mnn_config.type = MNN_FORWARD_VULKAN result :
Vulkan don't support 2, ArgMax: lanenet_model/vgg_backend/binary_seg/ArgMax
Vulkan don't support 9, Cast: lanenet/Cast
detect, 146, cost time: 0.755000 ms
detect, 153, cost time: 2487.019043 ms
detect, 204, cost time: 31.821001 ms
The text was updated successfully, but these errors were encountered: