remove trailing whitespaces

Signed-off-by: Junhee Yoo <junhee.yoo@navercorp.com>
This commit is contained in:
Junhee Yoo 2024-10-21 09:28:13 +09:00
parent 1467a7a064
commit e81462dda1
2 changed files with 4 additions and 4 deletions

View file

@ -1983,7 +1983,7 @@ kernel void kernel_im2col_ext(
const int32_t chw = tgpig[0] % CHW; const int32_t chw = tgpig[0] % CHW;
const int32_t tgpig_0 = chw / KHW; // 0 ~ (IC - 1) const int32_t tgpig_0 = chw / KHW; // 0 ~ (IC - 1)
const int32_t HW = tgpig[0] % KHW; const int32_t HW = tgpig[0] % KHW;
const int32_t tpitg_0 = (d * ntg[0]) + tpitg[0]; const int32_t tpitg_0 = (d * ntg[0]) + tpitg[0];
if (tpitg_0 >= N) { if (tpitg_0 >= N) {
return; return;

View file

@ -3325,13 +3325,13 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {12, 12, 2, 2048}, {3, 3, 2, 2048}, 1, 1, 1, 1, 1, 1, true)); test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {12, 12, 2, 2048}, {3, 3, 2, 2048}, 1, 1, 1, 1, 1, 1, true));
test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {12, 12, 1, 2560}, {3, 3, 1, 2560}, 1, 1, 1, 1, 1, 1, true)); test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {12, 12, 1, 2560}, {3, 3, 1, 2560}, 1, 1, 1, 1, 1, 1, true));
test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {12, 12, 2, 2560}, {3, 3, 2, 2560}, 1, 1, 1, 1, 1, 1, true)); test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {12, 12, 2, 2560}, {3, 3, 2, 2560}, 1, 1, 1, 1, 1, 1, true));
// sycl backend will limit task global_range < MAX_INT // sycl backend will limit task global_range < MAX_INT
// test cases for 2D im2col with large input W and H (occurs in stable-diffusion) // test cases for 2D im2col with large input W and H (occurs in stable-diffusion)
// however these cases need to alloc more memory which may fail in some devices (Intel Arc770, etc.) // however these cases need to alloc more memory which may fail in some devices (Intel Arc770, etc.)
// these cases are verified (pass) in Intel(R) Data Center GPU Max 1100 (sycl backend) and NV A30 (cuda backend) // these cases are verified (pass) in Intel(R) Data Center GPU Max 1100 (sycl backend) and NV A30 (cuda backend)
// test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {1024, 1024, 256, 1}, {3, 3, 256, 1}, 1, 1, 1, 1, 1, 1, true)); // test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F16, {1024, 1024, 256, 1}, {3, 3, 256, 1}, 1, 1, 1, 1, 1, 1, true));
// test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F32, {1024, 1024, 256, 1}, {3, 3, 256, 1}, 1, 1, 1, 1, 1, 1, true)); // test_cases.emplace_back(new test_im2col(GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_F32, {1024, 1024, 256, 1}, {3, 3, 256, 1}, 1, 1, 1, 1, 1, 1, true));
test_cases.emplace_back(new test_conv_transpose_1d()); test_cases.emplace_back(new test_conv_transpose_1d());
test_cases.emplace_back(new test_conv_transpose_1d({3,2,1,1}, {2,3,2,1}, 3, 0, 1)); test_cases.emplace_back(new test_conv_transpose_1d({3,2,1,1}, {2,3,2,1}, 3, 0, 1));