Commits

Breadcrumbs

History for

one1abaec

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 21, 2020

• fix "invalid min and max arguments for uniform_int" error (pytorch#282)

  Show description for defaa88
  

  Hongzhang Shan

  authored and

  jspark1105

  committed
  defaa88

  Copy full SHA for defaa88

• fix overflow in cache flush (pytorch#260)

  Show description for 2a881a1
  jspark1105

  committed
  2a881a1

  Copy full SHA for 2a881a1

• add 2-stage SparseAdagrad interface (pytorch#276)

  Show description for 4684696
  

  jiecaoyu

  authored and

  jspark1105

  committed
  4684696

  Copy full SHA for 4684696

• clang-format the code (pytorch#278)

  Show description for ff316a6
  

  jianyuh

  authored and

  jspark1105

  committed
  ff316a6

  Copy full SHA for ff316a6

• 2/4-bit EmbeddingSPMDM with rowwise sparsity (pytorch#263)

  Show description for 492a0b8
  jspark1105

  committed
  492a0b8

  Copy full SHA for 492a0b8

• 2bit JIT'ed SLS kernel (pytorch#234)

  Show description for ef59325
  jspark1105

  committed
  ef59325

  Copy full SHA for ef59325

• Remove extra paranthesis (pytorch#270)

  Show description for 1de1d5b
  

  jianyuh

  authored and

  jspark1105

  committed
  1de1d5b

  Copy full SHA for 1de1d5b

• use 2-stage EmbeddingSpMDM interface from benchmark and test (pytorch#256)

  Show description for 174b354
  jspark1105

  committed
  174b354

  Copy full SHA for 174b354

• EmbeddingSpMDM with float16 input (pytorch#265)

  Show description for 99ce6bf
  jspark1105

  committed
  99ce6bf

  Copy full SHA for 99ce6bf

• modified instrinsic fp16 kernel for windows build (pytorch#259)

  Show description for c373524
  jspark1105

  committed
  c373524

  Copy full SHA for c373524

• Intrinsic implementation of FP16 kernels for windows build (pytorch#254)

  Show description for 04e4804
  

  ykim362

  authored and

  jspark1105

  committed
  04e4804

  Copy full SHA for 04e4804

• add 2-stage EmbeddingSpMDM interface and change CodeCache to use read-write lock (pytorch#250)

  Show description for fdcdfeb
  jspark1105

  committed
  fdcdfeb

  Copy full SHA for fdcdfeb

• Fix multi-instance benchmarking for fp16 (pytorch#246)

  Show description for e3b453a
  

  dskhudia

  authored and

  jspark1105

  committed
  e3b453a

  Copy full SHA for e3b453a

• Githubaction (pytorch#237)

  Show description for ec75a9a
  

  Hongzhang Shan

  authored and

  jspark1105

  committed
  ec75a9a

  Copy full SHA for ec75a9a

• 4bit JIT'ed SLS kernel (pytorch#233)

  Show description for 621c739
  jspark1105

  committed
  621c739

  Copy full SHA for 621c739

• Adding missing copyright message

  Show description for cc2bc13
  

  dskhudia

  authored and

  jspark1105

  committed
  cc2bc13

  Copy full SHA for cc2bc13

• simplify SparseAdagrad (pytorch#227)

  Show description for b4ed0ac
  jspark1105

  committed
  b4ed0ac

  Copy full SHA for b4ed0ac

• simplify EmbeddingSpMDM (pytorch#228)

  Show description for 70aefac
  jspark1105

  committed
  70aefac

  Copy full SHA for 70aefac

• Fixing cblas.h for mac build in fp16 benchmark

  Show description for ececc2f
  

  Hongzhang Shan

  authored and

  jspark1105

  committed
  ececc2f

  Copy full SHA for ececc2f

• Adding FP32 JIT SparseAdagrad and Rowwise Adagrad (pytorch#224)

  Show description for 98b6534
  

  Protonu Basu

  authored and

  jspark1105

  committed
  98b6534

  Copy full SHA for 98b6534

• Explicit template parameters

  Show description for 535d400
  

  dskhudia

  authored and

  jspark1105

  committed
  535d400

  Copy full SHA for 535d400

• Things like iota/accumulate are defined in numeric header

  Show description for 4029550
  

  dskhudia

  authored and

  jspark1105

  committed
  4029550

  Copy full SHA for 4029550

• Aligned memory allocation that works with MSVC as well

  Show description for 7ecd3fa
  

  dskhudia

  authored and

  jspark1105

  committed
  7ecd3fa

  Copy full SHA for 7ecd3fa

• Back out "Revert D18138146: [caffe2] Add support for AVX512-256(YMM) in FBGEMM16" (pytorch#220)

  Show description for 996dcfc
  

  efiks

  authored and

  jspark1105

  committed
  996dcfc

  Copy full SHA for 996dcfc

• Revert D18138146: Add support for AVX512-256(YMM) in FBGEMM16

  Show description for 2bb8c8f
  

  jianyuh

  authored and

  jspark1105

  committed
  2bb8c8f

  Copy full SHA for 2bb8c8f

• Add support for AVX512-256(YMM) in FBGEMM16 (pytorch#209)

  Show description for 2c87645
  

  efiks

  authored and

  jspark1105

  committed
  2c87645

  Copy full SHA for 2c87645

• Reapply fix for make_unique not supported by C++11 (pytorch#217)

  Show description for 13e3366
  

  efiks

  authored and

  jspark1105

  committed
  13e3366

  Copy full SHA for 13e3366

• Fix non-MKL build in CI (pytorch#216)

  Show description for 5954322
  

  efiks

  authored and

  jspark1105

  committed
  5954322

  Copy full SHA for 5954322

• FBGEMM CI avoid call to C++14 (pytorch#215)

  Show description for 3a3339f
  

  efiks

  authored and

  jspark1105

  committed
  3a3339f

  Copy full SHA for 3a3339f

• Fix FBGEMM OSS CI (pytorch#212)

  Show description for c31324f
  

  efiks

  authored and

  jspark1105

  committed
  c31324f

  Copy full SHA for c31324f

• Fix FBGEMM OSS CI (pytorch#211)

  Show description for c8ef7b3
  

  jianyuh

  authored and

  jspark1105

  committed
  c8ef7b3

  Copy full SHA for c8ef7b3

• Add additional execution arguments to the benchmark (pytorch#207)

  Show description for a52df75
  

  efiks

  authored and

  jspark1105

  committed
  a52df75

  Copy full SHA for a52df75

• Pre-transpose Weight matrix (B) before sending to MKL routine (pytorch#205)

  Show description for ee1a771
  

  efiks

  authored and

  jspark1105

  committed
  ee1a771

  Copy full SHA for ee1a771

• Enhance benchmark to include cache eviction and multi instance with OMP (pytorch#204)

  Show description for 4131b25
  

  efiks

  authored and

  jspark1105

  committed
  4131b25

  Copy full SHA for 4131b25

• Adding FP 32 SLS, and unifying it with 8 Bit SLS (pytorch#206)

  Show description for 81cc0a1
  

  Protonu Basu

  authored and

  jspark1105

  committed
  81cc0a1

  Copy full SHA for 81cc0a1

Pagination

Previous

Next