Anaconda3、cuda10.1、python-3.8.5、 pytorch1.7.1、mmcv-full1.4.2, However, when running the command: python demo/quick_ Try.py demo/src.png -- style toy is always stuck in an error: Fail to importMultiScaleDeformableAttentionfrommmcv.ops.multi_ scale_ deform_ attn, You should installmmcv...
deform_attn performance optimization (open-mmlab#2616) * ms_opt_v2 * ms_opt_v2_1 * optimize MultiScaleDeformableAttention ops for MLU * ms_opt_v2_1 * [Feature] ms_deform_attn performance optimization V2 * [Feature] ms_deform_attn performance optimization V2 * [Feature] ms_deform_attn ...
wrong with the fp16, mmcv-full' version is 1.5.1. When I check the path of "mmcv/ops/csrc/pytorch/cuda/ms_deform_attn_cuda.cu", I found the "ms_deform_attn_cuda_forward" and function. Only by setting the fp16=None can I train the deformable-detr. So what can I do to use ...
7 changes: 0 additions & 7 deletions 7 mmcv/ops/csrc/common/cuda/ms_deform_attn_cuda_kernel.cuh Original file line numberDiff line numberDiff line change @@ -217,7 +217,6 @@ __global__ void ms_deformable_im2col_gpu_kernel( const int sampling_index = _temp; const int m_col = _...