Welcome to mirror list, hosted at ThFree Co, Russian Federation.

github.com/marian-nmt/FBGEMM.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
AgeCommit message (Expand)Author
2020-05-21Remove an unnecessary memory allocationyouki/improve-mem-allocYoung Jin Kim
2020-05-21Add the missing header for std::runtime_error (#379)Jianyu Huang
2020-05-18Drop extraneous "src" from local includes (#377)Nikita Shulga
2020-05-17Float16 rowwise sparse adagrad with stochastic rounding (reattempt of D215191...Yong Wu
2020-05-16Back out "Float16 rowwise sparse adagrad with stochastic rounding"Natalia Gimelshein
2020-05-16Float16 rowwise sparse adagrad with stochastic rounding (#370)Yong Wu
2020-05-15Add conv_1d (#369)Hongzhang Shan
2020-05-14Minor improvements in GEMM Kernels (#368)Daya Khudia
2020-05-14dirsync fbgemm with xplatJongsoo Park
2020-05-13fix benchmark to consider offsets is now default instead of lengths (#373)Jongsoo Park
2020-05-13use lea to simplify code (#372)Jongsoo Park
2020-05-12minor fix (#371)Hongzhang Shan
2020-05-06unified conv to call dw conv with 2 oc per g (#360)Jongsoo Park
2020-05-06depthwise convolution with 2 output channels per group (#359)Jongsoo Park
2020-05-04Fix convert indexing bugs (#367)Pawel Garbacki
2020-05-01use vreg.ymm() instead of ymm(vreg.id()) (#361)Jongsoo Park
2020-05-01remove one-stage interface for sparse adagrad (#366)Jongsoo Park
2020-04-30implement L2 regularization for Adagrad in fbgemm (#365)Jongsoo Park
2020-04-30asmjit submodule update (#364)Daya Khudia
2020-04-23Zero point addition after rounding in quantization routines (#362)Daya Khudia
2020-04-14Move `transpose_simd` to TransposeUtils.cc (#352)Nikita Shulga
2020-04-14move multiplication with b_zp from jit'ed code to requantization (#358)Jongsoo Park
2020-04-13minor changes in UniConvTest and GConvTest (#357)Jongsoo Park
2020-04-13simplify transposeConvWeights (#356)Jongsoo Park
2020-04-13follow EXPECT_EQ(actual, expected) convention (#354)Jongsoo Park
2020-04-11use .half (#353)Jongsoo Park
2020-04-11Revert D20956364: Move `transpose_simd` to TransposeUtils.ccZafar Takhirov
2020-04-10Move `transpose_simd` to TransposeUtils.cc (#349)Nikita Shulga
2020-04-10add option to use int64_t offsets/lengths in embedding operators (#350)Jongsoo Park
2020-04-10Move FakeFP16 back to internal to remove dependency on MKL (#36297)Jianyu Huang
2020-04-10Bazel build definition to use filelists from defs.bzl (#348)Nikita Shulga
2020-04-10change default value of use_offsets to true (to match with PyTorch EmbeddingB...Jongsoo Park
2020-04-10Move filelists to defs.bzl file (#344)Nikita Shulga
2020-04-09Add bazel build files (#346)Nikita Shulga
2020-04-09Merge pull request #345 from jspark1105/fixup-T64383689-masterDaya Khudia
2020-04-09Re-sync with internal repositoryJongsoo Park
2020-04-09Open sourcing fbgemm_fp16 ops (#36212)Yinghai Lu
2020-04-07add license header (#341)Jongsoo Park
2020-04-07JIT depth-wise conv (#338)Jongsoo Park
2020-04-07use read-write lock in CodeCache (#339)Jongsoo Park
2020-04-07Match numerics of quantize with vector path and fake quantize (#340)Daya Khudia
2020-04-03test 1024^3 matrix mulmat performance (#332)Hongzhang Shan
2020-04-03add dummy compute to cache evict func (#337)Jongsoo Park
2020-03-31disable asan for QuantizeAvx2 (#336)Jongsoo Park
2020-03-31add offset-based interface for PyTorch (#334)Jongsoo Park
2020-03-27Rename `times` global to avoid conflictAndrew Gallagher
2020-03-24remove unnecessary bound check for prefetching; avoid prefetching for -1 inde...Jongsoo Park
2020-03-23clamping with 1 comparison trick by treating signed as if it's unsigned (#324)Jongsoo Park
2020-03-23use xor instead of mov 0 (#325)Jongsoo Park
2020-03-23minor fixes in embedding benchmarks (#326)Jongsoo Park
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-18 17:38:13 +0300