20 Commits (b56fc0c2af8a3dc8729237b5e1e6e4e4e5d45dfa)

Author SHA1 Message Date
  ling e938efc5dd matmul 5 years ago
  ling 336ba49ef2 unsqueeze slice space-to-depth matmul 5 years ago
  ling 8a78b1f362 [MSLITE] arithmetic optimize 5 years ago
  zhanyuan 54506678fe [MSLITE] Optimize fp32 matmul for Arm v7a 5 years ago
  lixian 0aa289b0fd add avx fp32 matmul kernel 5 years ago
  zhaozhenlong dd3c4f8a9a parameter memory save check and comments 5 years ago
  ling 6de20db54a [MSLITE] fullconnection int8 suppot perchannel 5 years ago
  lixian 7f9d65cce0 apply int8 4x16 kernel 5 years ago
  zhanyuan 371de47944 [MSLITE] Fix the bug of MatVecMul fp32 5 years ago
  yangruoqi713 665a7287ca [MSLITE][Develop] fix bug of arm cpu matmul: init const tensor in Init function bofore check infershape done 5 years ago
  ling a56ba41c94 [MSLITE][Develop] delete useless matmul func 5 years ago
  ling c6a8848700 [MS][LITE][Develop] conv1x1 per oc arm64 5 years ago
  ling a3cc26ffcc [MSLITE][Develop]int8 conv1x1 aupport arm32 5 years ago
  ling 25f29dcf50 [MS][LITE][Develop] conv1x1 int8 weight transpose bug 5 years ago
  fuzhiye d34c620dce optimize winograd 5 years ago
  ling 2cbac5b261 [MS][LITE][Develop] Deconv matmul 12x8 5 years ago
  mindspore-ci-bot 9dd4ab0e3e !4896 [MS][LITE][Develop]int8 conv op 5 years ago
  lixian 811940bc55 change fp32 kernel to 12x8 5 years ago
  ling 6cfcdaab3d [MS][LITE][Develop]conv1x1 int8 5 years ago
  chenjianping d88a98658c move nnacl to lite/ 5 years ago