Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on Dec 9, 2024

• Fix an integer overflow in permute_multi_embedding() (#3465)

  Show description for 0feee94
  

  shintaro-iwasaki

  authored and

  facebook-github-bot

  committed
  0feee94

  Copy full SHA for 0feee94

• Migrate `jagged_dense_elementwise_add` SLL ops to OSS (#3479)

  Show description for bcb1f8f
  

  q10

  authored and

  facebook-github-bot

  committed
  bcb1f8f

  Copy full SHA for bcb1f8f

Commits on Dec 7, 2024

• Fix for T209749755 ("Your diff, D66861418, broke one test") (#3480)

  Show description for 3fb2862
  

  generatedunixname537391475639613

  authored and

  facebook-github-bot

  committed
  3fb2862

  Copy full SHA for 3fb2862

• Update .SO loader in OSS (#3477)

  Show description for 887e5bf
  

  q10

  authored and

  facebook-github-bot

  committed
  887e5bf

  Copy full SHA for 887e5bf

• Implement Vec2 load/store for ROCm devices (#3475)

  Show description for 3bf676c
  

  avbokovoy

  authored and

  facebook-github-bot

  committed
  3bf676c

  Copy full SHA for 3bf676c

• FP8 Rowwise compile fix followup for AMD (#3478)

  Show description for 72da187
  

  jwfromm

  authored and

  facebook-github-bot

  committed
  72da187

  Copy full SHA for 72da187

• Properly define preallocated output as mutable in fp8 rowwise gemm (#3476)

  Show description for 60bbdd7
  

  jwfromm

  authored and

  facebook-github-bot

  committed
  60bbdd7

  Copy full SHA for 60bbdd7

Commits on Dec 6, 2024

• Release Triton jagged tensor ops to OSS (#3474)

  Show description for f1804aa
  

  q10

  authored and

  facebook-github-bot

  committed
  f1804aa

  Copy full SHA for f1804aa

• migrate jagged_flash_attention_basic (#3472)

  Show description for 29cd178
  

  brad-mengchi

  authored and

  facebook-github-bot

  committed
  29cd178

  Copy full SHA for 29cd178

• more CK FP8 rowwise GEMM instances and tuning (#3455)

  Show description for 82bc8f4
  

  mxz297

  authored and

  facebook-github-bot

  committed
  82bc8f4

  Copy full SHA for 82bc8f4

• Manual loop unroll for rocm inference (#3439)

  Show description for 0f8f5e6
  

  avbokovoy

  authored and

  facebook-github-bot

  committed
  0f8f5e6

  Copy full SHA for 0f8f5e6

• Extend TBE support for larger embedding dimensions (#3467)

  Show description for ff73361
  

  yyetim

  authored and

  facebook-github-bot

  committed
  ff73361

  Copy full SHA for ff73361

• Align sll function names (#3471)

  Show description for 0eb60e0
  

  brad-mengchi

  authored and

  facebook-github-bot

  committed
  0eb60e0

  Copy full SHA for 0eb60e0

• Swap output rows with input rows in flight for fp16 (#3462)

  Show description for 427befb
  

  yyetim

  authored and

  facebook-github-bot

  committed
  427befb

  Copy full SHA for 427befb

• Support config based bound check version via extended modes (#3418)

  Show description for 2e90424
  

  Fei Yu

  authored and

  facebook-github-bot

  committed
  2e90424

  Copy full SHA for 2e90424

• Add meta registrations for kv_cache operators (#3442)

  Show description for 32f154e
  

  jwfromm

  authored and

  facebook-github-bot

  committed
  32f154e

  Copy full SHA for 32f154e

• Add genai_requirement.txt (#3469)

  Show description for fa4bc6b
  

  spcyppt

  authored and

  facebook-github-bot

  committed
  fa4bc6b

  Copy full SHA for fa4bc6b

• Migrate `jagged_jagged_bmm_jagged_out` SLL ops to OSS (#3459)

  Show description for e2a1f8c
  

  q10

  authored and

  facebook-github-bot

  committed
  e2a1f8c

  Copy full SHA for e2a1f8c

• Temporary disable nbit_forward_test on OSS rocm clang (#3445)

  Show description for 5a389d0
  

  spcyppt

  authored and

  facebook-github-bot

  committed
  5a389d0

  Copy full SHA for 5a389d0

• Migrate `array_jagged_bmm_jagged_out` SLL op to OSS (#3456)

  Show description for 837b14f
  

  q10

  authored and

  facebook-github-bot

  committed
  837b14f

  Copy full SHA for 837b14f

• Fix FP8 Rowwise Gemm Compilation with Auto-functionalize V2 (#3457)

  Show description for 03129ae
  

  jwfromm

  authored and

  facebook-github-bot

  committed
  03129ae

  Copy full SHA for 03129ae

• address mem over used during flushing (#3460)

  Show description for e560190
  

  duduyi2013

  authored and

  facebook-github-bot

  committed
  e560190

  Copy full SHA for e560190

• [FBGEMM open-source SLL jagged2_softmax (#3428)

  Show description for 1d3da55
  

  brad-mengchi

  authored and

  facebook-github-bot

  committed
  1d3da55

  Copy full SHA for 1d3da55

• Modularize CMake Build [5B/N] (#3451)

  Show description for 9d78337
  

  q10

  authored and

  facebook-github-bot

  committed
  9d78337

  Copy full SHA for 9d78337

• Clean up sparse bucketize (#3302)

  Show description for e58c49f
  

  q10

  authored and

  facebook-github-bot

  committed
  e58c49f

  Copy full SHA for e58c49f

• Hot fix to skip VBE CPU reshaping for MTIA (#3466)

  Show description for 4e18e7b
  

  spcyppt

  authored and

  facebook-github-bot

  committed
  4e18e7b

  Copy full SHA for 4e18e7b

Commits on Dec 5, 2024

• add enable_async_update into tbe signature (#3461)

  Show description for 38518bf
  

  duduyi2013

  authored and

  facebook-github-bot

  committed
  38518bf

  Copy full SHA for 38518bf

• Modularize CMake Build [4/N] (#3417)

  Show description for 09b7b96
  

  q10

  authored and

  facebook-github-bot

  committed
  09b7b96

  Copy full SHA for 09b7b96

• Add options for verbose builds in CI (#3446)

  Show description for 8f214ec
  

  q10

  authored and

  facebook-github-bot

  committed
  8f214ec

  Copy full SHA for 8f214ec

• Reorganize sparse block bucketize macros (#3296)

  Show description for df95b17
  

  q10

  authored and

  facebook-github-bot

  committed
  df95b17

  Copy full SHA for df95b17

• Refactor repeat code in sparse_block_bucketize (#3295)

  Show description for eeecd78
  

  q10

  authored and

  facebook-github-bot

  committed
  eeecd78

  Copy full SHA for eeecd78

• open-source SLL jagged_softmax (#3458)

  Show description for 8e23e33
  

  brad-mengchi

  authored and

  facebook-github-bot

  committed
  8e23e33

  Copy full SHA for 8e23e33

• create pack_segments_v2 with additional pad_minf and presence_mask functionality (#3427)

  Show description for be6c6c3
  

  brad-mengchi

  authored and

  facebook-github-bot

  committed
  be6c6c3

  Copy full SHA for be6c6c3

• Support zero-size inputs in FP8 cuda quantize kernel (#3448)

  Show description for 1a0d837
  

  jiawenliu64

  authored and

  facebook-github-bot

  committed
  1a0d837

  Copy full SHA for 1a0d837

Commits on Dec 4, 2024

• Use int64_t index for SplitOptimizer grad (#3447)

  Show description for 264f946
  

  sryap

  authored and

  facebook-github-bot

  committed
  264f946

  Copy full SHA for 264f946