Commits

Branch selector

 master

User selector

jspark1105

Datepicker

All time

Commit History

Commits on Apr 9, 2020

• Re-sync with internal repository

  jspark1105

  committed
  45ad43d

  Copy full SHA for 45ad43d

Commits on Apr 7, 2020

• add license header (pytorch#341)

  Show description for c97495c
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  c97495c

  Copy full SHA for c97495c

• JIT depth-wise conv (pytorch#338)

  Show description for 1f42be5
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  1f42be5

  Copy full SHA for 1f42be5

• use read-write lock in CodeCache (pytorch#339)

  Show description for 35e486b
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  35e486b

  Copy full SHA for 35e486b

Commits on Apr 3, 2020

• add dummy compute to cache evict func (pytorch#337)

  Show description for 1bda6ae
  jspark1105

  committed
  1bda6ae

  Copy full SHA for 1bda6ae

Commits on Mar 31, 2020

• disable asan for QuantizeAvx2 (pytorch#336)

  Show description for 1798e56
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  1798e56

  Copy full SHA for 1798e56

• add offset-based interface for PyTorch (pytorch#334)

  Show description for df34258
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  df34258

  Copy full SHA for df34258

Commits on Mar 24, 2020

• remove unnecessary bound check for prefetching; avoid prefetching for -1 index (pytorch#329)

  Show description for 60b734a
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  60b734a

  Copy full SHA for 60b734a

Commits on Mar 23, 2020

• clamping with 1 comparison trick by treating signed as if it's unsigned (pytorch#324)

  Show description for 58c002d
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  58c002d

  Copy full SHA for 58c002d

• use xor instead of mov 0 (pytorch#325)

  Show description for 38ee061
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  38ee061

  Copy full SHA for 38ee061

• minor fixes in embedding benchmarks (pytorch#326)

  Show description for 49a56f7
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  49a56f7

  Copy full SHA for 49a56f7

Commits on Mar 22, 2020

• remove unnecessary prefetch for scale/bias in corner cases (pytorch#323)

  Show description for 4b36034
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  4b36034

  Copy full SHA for 4b36034

• prefetch scale/bias (pytorch#322)

  Show description for d7c4a34
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  d7c4a34

  Copy full SHA for d7c4a34

Commits on Mar 21, 2020

• use simd_info in GenerateKernel.h (pytorch#320)

  Show description for fff5d32
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  fff5d32

  Copy full SHA for fff5d32

• Update RefImplementations.cc

  jspark1105

  authored
  db31580

  Copy full SHA for db31580

• Update CMakeLists.txt

  jspark1105

  authored
  6e21cf4

  Copy full SHA for 6e21cf4

• reduce Windows CI warnings (pytorch#309)

  Show description for 99df751
  jspark1105

  committed
  99df751

  Copy full SHA for 99df751

• make cmake less verbose (pytorch#308)

  Show description for 4942ea9
  jspark1105

  committed
  4942ea9

  Copy full SHA for 4942ea9

• fbgemm handles block size 1 sparse adagrad (pytorch#306)

  Show description for e2950ea
  jspark1105

  committed
  e2950ea

  Copy full SHA for e2950ea

• add copyright to EmbeddingSpMDMTestUtils.{cc,h} (pytorch#303)

  Show description for ca1e41c
  jspark1105

  committed
  ca1e41c

  Copy full SHA for ca1e41c

• fix index mapping table type to int32_t (pytorch#302)

  Show description for 1a7be35
  jspark1105

  committed
  1a7be35

  Copy full SHA for 1a7be35

• clean up embedding spmdm tests (pytorch#300)

  Show description for 6a4ed1c
  jspark1105

  committed
  6a4ed1c

  Copy full SHA for 6a4ed1c

• make order btw div and mul in adgrad consistent (#32974)

  Show description for 86dd3bf
  jspark1105

  committed
  86dd3bf

  Copy full SHA for 86dd3bf

• rowwise sparse adagrad fused (pytorch#298)

  Show description for cfda552
  jspark1105

  committed
  cfda552

  Copy full SHA for cfda552

• simplify avx2 mask usage in embedding kernels (pytorch#287)

  Show description for ccd91d6
  jspark1105

  committed
  ccd91d6

  Copy full SHA for ccd91d6

• clean up temporary code for isa-dependent serialized packed weight (pytorch#290)

  Show description for 365633b
  jspark1105

  committed
  365633b

  Copy full SHA for 365633b

• row-wise sparse fp32/fp16/int8 EmbeddingSpMDM (pytorch#288)

  Show description for b8297c2
  jspark1105

  committed
  b8297c2

  Copy full SHA for b8297c2

• enable positional weight for EmbeddingSpMDMNBitRowWiseSparse (pytorch#281)

  Show description for fb7a373
  jspark1105

  committed
  fb7a373

  Copy full SHA for fb7a373

• enable EmbeddingSpMDMNBitRowWiseSparse avx2 (pytorch#280)

  Show description for 0b0e2d9
  jspark1105

  committed
  0b0e2d9

  Copy full SHA for 0b0e2d9

• fix overflow in cache flush (pytorch#260)

  Show description for 2a881a1
  jspark1105

  committed
  2a881a1

  Copy full SHA for 2a881a1

• 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

• make gconv work for any W and H (pytorch#269)

  Show description for 176a2bb
  jspark1105

  committed
  176a2bb

  Copy full SHA for 176a2bb

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

  Show description for 174b354
  jspark1105

  committed
  174b354

  Copy full SHA for 174b354

• add missing copyright header (pytorch#267)

  Show description for 56fe514
  jspark1105

  committed
  56fe514

  Copy full SHA for 56fe514

Pagination

Previous

Next