You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
Megvii Engine Team 2aba0378b9 refactor(mgb/dnn): fix group conv is_available 3 years ago
..
adaptive_pooling chore(mge): update copyright years 4 years ago
add_update feat(log): opt log 4 years ago
argmxx chore(mge): update copyright years 4 years ago
argsort feat(mge/random): add some random op and remove random/distrbution.py 3 years ago
batch_conv_bias feat(mgb/dnn): add accuracy shake checker 4 years ago
batch_normalization feat(log): opt log 4 years ago
batched_matrix_mul feat(mgb/dnn): add accuracy shake checker 4 years ago
check_has_inf feat(mgb/opr): add check_has_inf kernel and opr 3 years ago
checksum chore(mge): update copyright years 4 years ago
concat chore(mge): update copyright years 4 years ago
cond_take chore(mge): update copyright years 4 years ago
conv_bias refactor(mgb/dnn): fix group conv is_available 3 years ago
convolution refactor(mgb/dnn): fix group conv is_available 3 years ago
convolution3d refactor(mgb/dnn): fix group conv is_available 3 years ago
convolution_helper perf(cuda/conv): cache serval cudnn api 4 years ago
convpooling chore(mge): update copyright years 4 years ago
correlation feat(megbrain): add correlation opr 4 years ago
cub feat(dnn/cuda): topk support fp16 3 years ago
cumsum feat(log): opt log 4 years ago
cv chore(mge): update copyright years 4 years ago
cvt_color chore(mge): update copyright years 4 years ago
dct chore(mge): update copyright years 4 years ago
deformable_conv refactor(mgb/dnn): add negative attribute for algo 4 years ago
deformable_ps_roi_pooling chore(mge): update copyright years 4 years ago
dot fix(mgb/dnn): fix cub potential issues 4 years ago
elemwise feat(mge/opr): add silu and gelu 3 years ago
elemwise_multi_type feat(mge/opr): add silu and gelu 3 years ago
eye chore(mge): update copyright years 4 years ago
fake_quant chore(mge): update copyright years 4 years ago
flip chore(mge): update copyright years 4 years ago
gaussian_blur chore(mge): update copyright years 4 years ago
group_local chore(mge): update copyright years 4 years ago
images2neibs feat(mge/module): add python wrapper for unfold 4 years ago
indexing_multi_axis_vec feat(dnn/cuda): support float16 for index_incr_multi_axis_vec 4 years ago
indexing_one_hot feat(log): opt log 4 years ago
linspace chore(mge): update copyright years 4 years ago
local chore(mge): update copyright years 4 years ago
local_share fix(mgb): add usable-depend-on-shape attr 4 years ago
lrn chore(mge): update copyright years 4 years ago
lsq feat(mgb/opr): add lsq opr 3 years ago
mask_conv chore(mge): update copyright years 4 years ago
matrix_inverse chore(mge): update copyright years 4 years ago
matrix_mul feat(dnn/cuda): generate cutlass kimpls using cmake and bazel 3 years ago
max_tensor_diff chore(mge): update copyright years 4 years ago
megcore feat(log): opt log 4 years ago
mesh_indexing chore(mge): update copyright years 4 years ago
param_pack chore(mge): update copyright years 4 years ago
pooling feat(dnn/cuda): add relayout format kernel for nchw <-> nhwc 3 years ago
powc chore(mge): update copyright years 4 years ago
reduce feat(mgb/opr): add check_has_inf kernel and opr 3 years ago
reduce_helper chore(mge): update copyright years 4 years ago
relayout feat(dnn/cuda): support warp perspective/pooling op when channel not aligned to 64 4 years ago
relayout_format refactor(mgb/dnn): fix group conv is_available 3 years ago
remap feat(log): opt log 4 years ago
repeat chore(mge): update copyright years 4 years ago
resize feat(log): opt log 4 years ago
rng feat(mge/random): add some random op and remove random/distrbution.py 3 years ago
roi_align chore(mge): update copyright years 4 years ago
roi_copy chore(mge): update copyright years 4 years ago
roi_pooling chore(mge): update copyright years 4 years ago
rotate chore(mge): update copyright years 4 years ago
separable_conv chore(mge): update copyright years 4 years ago
separable_filter chore(mge): update copyright years 4 years ago
sleep chore(mge): update copyright years 4 years ago
sliding_window_transpose feat(mge): add SlidingWindowTranspose opr 3 years ago
split chore(mge): update copyright years 4 years ago
svd chore(mge): update copyright years 4 years ago
tensor_remap chore(mge): update copyright years 4 years ago
tile chore(mge): update copyright years 4 years ago
topk feat(dnn/cuda): topk support fp16 3 years ago
tqt fix(mgb/dnn): fix backward computation of tqt 4 years ago
transpose chore(mge): update copyright years 4 years ago
type_cvt feat(dnn/cuda): support warp perspective/pooling op when channel not aligned to 64 4 years ago
warp_affine feat(log): opt log 4 years ago
warp_perspective feat(dnn/cuda): add relayout format kernel for nchw <-> nhwc 3 years ago
api_cache.h fix(api_cache): fix serialization for conv_desc 4 years ago
cuda_shfl_compat.cuh chore(mge): update copyright years 4 years ago
cudnn_with_check.h chore(mge): update copyright years 4 years ago
cudnn_wrapper.cpp refactor(mgb/dnn): fix group conv is_available 3 years ago
cudnn_wrapper.h feat(dnn): add deterministic max pooling 3 years ago
elemwise_helper.cpp feat(dnn/cuda): add cuda elemwise int4 4 years ago
elemwise_helper.cuh feat(mgb/opr): add lsq opr 3 years ago
elemwise_helper_q4.cuh feat(dnn/cuda): add cuda elemwise int4 4 years ago
error_info.cuh chore(mge): update copyright years 4 years ago
fp16_help.cuh chore(mge): update copyright years 4 years ago
handle.cpp fix(mgb): remove unnecessary cudnn8 warning 3 years ago
handle.h perf(cuda/conv): cache serval cudnn api 4 years ago
handle_create.cpp feat(mge): add SlidingWindowTranspose opr 3 years ago
int_fastdiv.cpp chore(mge): update copyright years 4 years ago
int_fastdiv.cuh chore(mge): update copyright years 4 years ago
integer_subbyte_utils.cuh refactor(dnn/cuda): refactor relayout format kernels 3 years ago
memory_utils.cuh feat(dnn/cuda): support warp perspective/pooling op when channel not aligned to 64 4 years ago
query_blocksize.cpp chore(mge): update copyright years 4 years ago
query_blocksize.cuh chore(mge): update copyright years 4 years ago
query_blocksize_impl.cu chore(mge): update copyright years 4 years ago
reduce_helper.cuh chore(mge): update copyright years 4 years ago
reduce_helper.cuinl chore(mge): update copyright years 4 years ago
utils.cpp chore(mge): update copyright years 4 years ago
utils.cuh feat(dnn/cuda): support warp perspective/pooling op when channel not aligned to 64 4 years ago
utils.h chore(mge): update copyright years 4 years ago

MegEngine 安装包中集成了使用 GPU 运行代码所需的 CUDA 环境,不用区分 CPU 和 GPU 版。 如果想要运行 GPU 程序,请确保机器本身配有 GPU 硬件设备并安装好驱动。 如果你想体验在云端 GPU 算力平台进行深度学习开发的感觉,欢迎访问 MegStudio 平台

C++ Cuda Python C SVG other