Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 19, 2024

• DeepSpeed content for 1.19.0

  Show description for 9c2d043
  

  SW publisher

  authored and

  Jenkins

  committed
  9c2d043

Commits on Jun 20, 2024

• [XPU] support op builder from intel_extension_for_pytorch kernel path (microsoft#5425)

  Show description for d254d75
  

  3 peopleauthored
  d254d75

Commits on Jun 19, 2024

• Update BUFSIZE to come from autotuner's constants.py, not numpy (microsoft#5686)

  loadams

  authored
  0c979d6

Commits on Jun 18, 2024

• Fix numpy upgrade to 2.0.0 BUFSIZE import error (microsoft#5680)

  Show description for 4000cee
  

  Yejing-Lai

  and

  loadams

  authored
  4000cee

• Fix latest pytorch '_get_socket_with_port' import error (microsoft#5654)

  Show description for a4cd550
  Yejing-Lai

  authored
  a4cd550

• enable yuan autotp & add conv tp (microsoft#5428)

  Show description for 8ea995e
  

  4 peopleauthored
  8ea995e

• Fixing the reshape bug in sequence parallel alltoall, which corrupted all QKV data (microsoft#5664)

  Show description for 3bdd187
  

  3 peopleauthored
  3bdd187

• Fix hpZ with zero element (microsoft#5652)

  Show description for b33873d
  

  samadejacobs

  and

  loadams

  authored
  b33873d

Commits on Jun 17, 2024

• Remove compile wrapper to simplify access to model attributes (microsoft#5581)

  Show description for 2a0c0e3
  

  tohtana

  and

  tjruwase

  authored
  2a0c0e3

• fix IDEX dependence in xpu accelerator (microsoft#5666)

  Show description for 8831b57
  Liangliang-Ma

  authored
  8831b57

• [XPU] adapt lazy_call func to different versions (microsoft#5670)

  Show description for 7331630
  YizhouZ

  authored
  7331630

• _exec_forward_pass: place zeros(1) on the same device as the param (microsoft#5576)

  Show description for 05cb79d
  

  nelyahu

  and

  loadams

  authored
  05cb79d

Commits on Jun 12, 2024

• [CPU] SHM based allreduce improvement for small message size (microsoft#5571)

  Show description for eda5075
  

  3 peopleauthored
  eda5075

• Update version.txt after 0.14.3 release (microsoft#5651)

  Show description for dfcade2
  mrwyattii

  authored
  dfcade2

• Monitor was always enabled causing performance degradation (microsoft#5633)

  Show description for 54f98fd
  

  3 peopleauthored
  54f98fd

• pipe/_exec_backward_pass: fix immediate grad update (microsoft#5605)

  Show description for 91a9986
  nelyahu

  authored
  91a9986

• assumption of torch.initial_seed function accepting seed arg in DeepSpeedAccelerator abstract class is incorrect (microsoft#5569)

  Show description for ac935c7
  

  polisettyvarma

  and

  tjruwase

  authored
  ac935c7

Commits on Jun 11, 2024

• fixes in _partition_param_sec function (microsoft#5613)

  Show description for b6e24ad
  mmhab

  authored
  b6e24ad

Commits on Jun 10, 2024

• Fix overlap communication of ZeRO stage 1 and 2 (microsoft#5606)

  Show description for a41729f
  

  5 peopleauthored
  a41729f

• stage_1_and_2: optimize clip calculation to use clamp (microsoft#5632)

  Show description for 1ef9b02
  nelyahu

  authored
  1ef9b02

• WA for Torch-compile-Z3-act-apt accuracy issue from the Pytorch repo (microsoft#5590)

  Show description for 6e2899f
  

  NirSonnenschein

  and

  tjruwase

  authored
  6e2899f

Commits on Jun 7, 2024

• Pin transformers version for MII tests (microsoft#5629)

  Show description for 31a57fa
  loadams

  authored
  31a57fa

Commits on Jun 6, 2024

• Updated hpu-gaudi2 tests content. (microsoft#5622)

  Show description for 69adeab
  vshekhawat-hlab

  authored
  69adeab

Commits on Jun 5, 2024

• Add Compressedbackend for Onebit optimizers (microsoft#5473)

  Show description for 11a62a0
  

  Liangliang-Ma

  and

  tjruwase

  authored
  11a62a0

• fix sequence parallel(Ulysses) grad scale for zero0 (microsoft#5555)

  Show description for 6b6d641
  inkcherry

  authored
  6b6d641

• Fix cuda hardcode for inference woq (microsoft#5565)

  Show description for af4356b
  

  Liangliang-Ma

  and

  loadams

  authored
  af4356b

Commits on Jun 4, 2024

• estimate_zero2_model_states_mem_needs: fixing memory estiamtion (microsoft#5099)

  Show description for f4cb866
  

  nelyahu

  and

  tjruwase

  authored
  f4cb866

Commits on May 31, 2024

• Fixed the Windows build. (microsoft#5596)

  Show description for e7dd28a
  

  3 peopleauthored
  e7dd28a

• Add slide deck for meetup in Japan (microsoft#5598)

  tohtana

  authored
  77c9494

• Update minor CUDA version compatibility (microsoft#5591)

  Show description for 1baf688
  adk9

  authored
  1baf688

Commits on May 28, 2024

• DeepSpeedCheckpoint: support custom final ln idx (microsoft#5506)

  Show description for 2fc702e
  

  3 peopleauthored
  2fc702e

• Update to fix sidebar over text (microsoft#5567)

  Show description for 4deb40d
  loadams

  authored
  4deb40d

• Rename files in fp_quantize op from quantize.* to fp_quantize.* (microsoft#5577)

  Show description for bf66acd
  loadams

  authored
  bf66acd

• [MiCS] Remove the handle print on DeepSpeed side (microsoft#5574)

  Show description for fd8051a
  ys950902

  authored
  fd8051a

• Update HPU docker version (microsoft#5566)

  loadams

  authored
  988372b

Pagination

PreviousNext