Commits

Branch selector

 f433d5f

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 9, 2021

• Fix uvm cache checkpoint load

  Show description for 00f4892
  

  nqjade

  authored and

  facebook-github-bot

  committed
  00f4892

Commits on Apr 7, 2021

• Add CacheState type to fbgemm_gpu (pytorch#591)

  Show description for 4b5e997
  

  r-barnes

  authored and

  facebook-github-bot

  committed
  4b5e997

Commits on Apr 6, 2021

• fix gpu device bug in split_table_batched_embeddings_benchmark (pytorch#590)

  Show description for b01c349
  

  gsethi523

  authored and

  facebook-github-bot

  committed
  b01c349

• optimize exact row-wise sparse adagrad on cpu (pytorch#589)

  Show description for a98ad84
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  a98ad84

• parallelize cpu emb fwd across B

  Show description for 1dfb0c3
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  1dfb0c3

• add input stride option to embedding spmdm

  Show description for 9b623bb
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  9b623bb

• More types for fbgemm_gpu (pytorch#585)

  Show description for c565348
  

  r-barnes

  authored and

  facebook-github-bot

  committed
  c565348

Commits on Apr 5, 2021

• correct use of contiguous in cpu embedding ops (pytorch#588)

  Show description for 4c43051
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  4c43051

• fix bug split_table_batched_embeddings_benchmark in reporting BW

  Show description for c109fc2
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  c109fc2

• make split_table_batched_embeddings_benchmark work on cpu

  Show description for f5dbb5a
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  f5dbb5a

• extract kernel for easier debugging (pytorch#587)

  Show description for 62c7209
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  62c7209

• fix cpu approx sparse adagrad (pytorch#586)

  Show description for 228cbac
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  228cbac

Commits on Apr 2, 2021

• int8 stochastic rounding

  Show description for 9c0408d
  

  amylittleyang

  authored and

  facebook-github-bot

  committed
  9c0408d

Commits on Apr 1, 2021

• Fix zero_point calculation (pytorch#584)

  Show description for f29eaad
  

  supriyar

  authored and

  facebook-github-bot

  committed
  f29eaad

Commits on Mar 31, 2021

• Re-apply "[fbgemm_gpu] have bf16 quantize kernels use 2d blocks with loop to cover large tensors" (pytorch#583)

  Show description for 5a8d11f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  5a8d11f

Commits on Mar 30, 2021

• Types for fbgemm (pytorch#580)

  Show description for e839120
  

  r-barnes

  authored and

  facebook-github-bot

  committed
  e839120

• table-wise sequence embeddings

  Show description for d8ac3eb
  

  xing-liu

  authored and

  facebook-github-bot

  committed
  d8ac3eb

• remove fp16 flag (pytorch#581)

  Show description for 2c2ae1f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  2c2ae1f

Commits on Mar 25, 2021

• Build fix with certain compilers versions (pytorch#575)

  Show description for 6c051e1
  

  dskhudia

  authored and

  facebook-github-bot

  committed
  6c051e1

Commits on Mar 24, 2021

• Pyre for fbgemm (pytorch#558)

  Show description for 203f7ff
  

  r-barnes

  authored and

  facebook-github-bot

  committed
  203f7ff

• add approx_sgd/rowwise_adagrad to setup.py (pytorch#579)

  Show description for a8097ee
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  a8097ee

• Remove Explicit prefetch argument from forward (pytorch#578)

  Show description for 7c0c486
  

  another-pjohnson

  authored and

  facebook-github-bot

  committed
  7c0c486

• Back out "have bf16 quantize kernels use 2d blocks with loop to cover large tensors" (pytorch#572)

  Show description for fc2c734
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  fc2c734

Commits on Mar 23, 2021

• Use more compact syntax for is_uvm_tensor (pytorch#577)

  Show description for 1d7eaf3
  

  ezyang

  authored and

  facebook-github-bot

  committed
  1d7eaf3

• use JIT'ed approx rowwise sparse adagrad on CPU (pytorch#574)

  Show description for a2b58df
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  a2b58df

• add grad stride option to fused rowwise sparse adagrad (pytorch#568)

  Show description for 8998e6f
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  8998e6f

• make ref code deterministic in split_table_batched_embeddings_test (pytorch#565)

  Show description for 7f78584
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  7f78584

• Fix OSS build failure due to approx sgd (pytorch#570)

  Show description for 28b130f
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  28b130f

• Move permute_pooled_embs_kernel into fbgemm_gpu (pytorch#569)

  Show description for 3657619
  

  jianyuh

  authored and

  facebook-github-bot

  committed
  3657619

Commits on Mar 22, 2021

• relax error bound of fp16 sparse adagrad (pytorch#567)

  Show description for f3abcb8
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  f3abcb8

• fix approx sgd on cpu (pytorch#566)

  Show description for ffff7a3
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  ffff7a3

• avoid unused variable warning for approx sgd (pytorch#564)

  Show description for 99f3034
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  99f3034

Commits on Mar 20, 2021

• avoid initializing embedding bag output to zeros (pytorch#560)

  Show description for 7060df8
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  7060df8

• array out of bound checking (pytorch#562)

  Show description for 9f0062e
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  9f0062e

• add one more entry to hash_size_cum to easier compute of hash size of each table (pytorch#561)

  Show description for 4ac3a49
  

  jspark1105

  authored and

  facebook-github-bot

  committed
  4ac3a49

Pagination

PreviousNext