30 Commits (296e0022df6021cd404e43f74b8f8744fcb3f35d)

Author SHA1 Message Date
  nihuini 296e0022df deconvolution output adj and output shape 6 years ago
  nihuini e4b44d293e more autopad SAME_LOWER 6 years ago
  nihuini 9a6ee37eef asymmetric padding parameter for convolution and deconvolution family 6 years ago
  nihui 29324771b1
Implemented hard swish layer (#1195) 6 years ago
  nihuini 3d5b7f20ff avgpool count_include_pad 6 years ago
  Hao Zeng 1f6919fd40 Implemented hard swish layer 6 years ago
  nihuini 1cce18bdde binaryop broadcast vulkan 6 years ago
  nihui 0b6d7b7096 use underscored offset 6 years ago
  nihuini 163fb92537 concat vulkan pack1to4 and pack4to1to4 6 years ago
  nihuini 1b910efea5 convert slice properly 6 years ago
  nihui cf42e7c254 deconvolutiondepthwise pack4 arm neon 6 years ago
  nihui 60c0890eaf crop region -234 is rarely used, fix out of channel range write, crop pack4 arm neon 6 years ago
  nihuini c4f23ae8ad rename Mat packing to elempack 6 years ago
  nihui c013bd9b7e vulkan convolution winograd f63 6 years ago
  nihuini c769437533 fix fp16p deconvolution and convolution-typed innerproduct 7 years ago
  PENGUINLIONG 084053fed8 Implemented hard sigmoid (#1046) 7 years ago
  nihui 21f79b8546 prefer cpu fp16 casting to reduce upload/download overhead on discrete gpu 7 years ago
  nihuini e09607bc22 add option to upload model function, pipeline creation honors option use flags, setting allocator per extractor do not make much sense 7 years ago
  nihui fe4b00f7a2 unroll outh 4 for winograd gemm 7 years ago
  nihuini 74276314bb unroll size 4 for conv1x1s1 pack4 7 years ago
  nihuini cd7559c639 more fix for fp16p, still disabled by default 7 years ago
  harhar539 5e317b98c5 fix illegal memory access at conv layer of vulkan (#1011) 7 years ago
  nihui 25b9736f82 shader fp16 packed 7 years ago
  nihuini 4b50a97e31 implement vulkan winograd23 7 years ago
  nihuini aa94e77e68 fix pipeline object leak 7 years ago
  nihui 3e003ffd98 fuse sigmoid 7 years ago
  nihui 5adfa290a5 1x1s1d1_lds_4_4_4 is non-optimal, delete it 7 years ago
  nihuini 8ac300c3a2 mat4 type in shared memory makes some driver unhappy .. 7 years ago
  nihuini f5ba97e7c6 lds optimize for conv3x3s1, conv1x1s1 and fc 7 years ago
  nihuini 7a8f68aca6 move vulkan code to subdir, new layer interface create_pipeline and destroy_pipeline for post-loading works 7 years ago