@@ -1007,12 +1007,12 @@ void ggml_cuda_op_conv2d_implicit(ggml_backend_cuda_context & ctx, ggml_tensor *
10071007 const int cc = ggml_cuda_info ().devices [ctx.device ].cc ;
10081008
10091009 const int32_t * p = (const int32_t *) dst->op_params ;
1010- const int ST_X = p[0 ]; // stride_x
1011- const int ST_Y = p[1 ]; // stride_y
1012- const int PD_X = p[2 ]; // padding_x
1013- const int PD_Y = p[3 ]; // padding_y
1014- const int DL_X = p[4 ]; // dilation_x
1015- const int DL_Y = p[5 ]; // dilation_y
1010+ const uint ST_X = p[0 ]; // stride_x
1011+ const uint ST_Y = p[1 ]; // stride_y
1012+ const uint PD_X = p[2 ]; // padding_x
1013+ const uint PD_Y = p[3 ]; // padding_y
1014+ const uint DL_X = p[4 ]; // dilation_x
1015+ const uint DL_Y = p[5 ]; // dilation_y
10161016 // const int LT = p[6]; // layout
10171017
10181018 // GGML_ASSERT(LT == 0 || LT == 1);
@@ -1022,16 +1022,16 @@ void ggml_cuda_op_conv2d_implicit(ggml_backend_cuda_context & ctx, ggml_tensor *
10221022 // No cwhn
10231023 GGML_ASSERT (p[6 ] == false );
10241024
1025- const int IW = input->ne [0 ]; // input_w
1026- const int IH = input->ne [1 ]; // input_h
1027- const int OW = dst->ne [0 ]; // output_w
1028- const int OH = dst->ne [1 ]; // output_h
1029- const int KW = kernel->ne [0 ]; // kernel_w
1030- const int KH = kernel->ne [1 ]; // kernel_h
1031- const int IC = input->ne [2 ]; // input_channels
1025+ const uint IW = input->ne [0 ]; // input_w
1026+ const uint IH = input->ne [1 ]; // input_h
1027+ const uint OW = dst->ne [0 ]; // output_w
1028+ const uint OH = dst->ne [1 ]; // output_h
1029+ const uint KW = kernel->ne [0 ]; // kernel_w
1030+ const uint KH = kernel->ne [1 ]; // kernel_h
1031+ const uint IC = input->ne [2 ]; // input_channels
10321032
1033- const int OC = kernel->ne [3 ]; // ouptut_chanles
1034- const int B = input->ne [3 ]; // n_batches
1033+ const uint OC = kernel->ne [3 ]; // ouptut_chanles
1034+ const uint B = input->ne [3 ]; // n_batches
10351035
10361036 const int64_t total = B * OC * OH * OW;
10371037
0 commit comments