Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Nov 28, 2023

• use all_gather_into_tensor instead of all_gather (deepspeedai#4705)

  Show description for fd0a52c
  

  taozhiwei

  and

  --local

  authored
  fd0a52c

  Copy full SHA for fd0a52c

• [docs] Intel inference blog (deepspeedai#4734)

  jeffra

  authored
  6b8103b

  Copy full SHA for 6b8103b

Commits on Nov 21, 2023

• Communication Optimization for Large-Scale Training (deepspeedai#4695)

  Show description for 2afa1c7
  

  RezaYazdaniAminabadi

  and

  tjruwase

  authored
  2afa1c7

  Copy full SHA for 2afa1c7

• Update flops profiler to recurse (deepspeedai#4374)

  Show description for 6139122
  loadams

  authored
  6139122

  Copy full SHA for 6139122

Commits on Nov 17, 2023

• Remove unneeded dict reinit (fix for deepspeedai#4565) (deepspeedai#4702)

  Show description for a603a21
  eisene

  authored
  a603a21

  Copy full SHA for a603a21

• Add get and set APIs for the ZeRO-3 partitioned parameters (deepspeedai#4681)

  Show description for 0ec2d3e
  

  4 peopleauthored
  0ec2d3e

  Copy full SHA for 0ec2d3e

• infV2 fix for OPT size variants (deepspeedai#4694)

  Show description for a3926bb
  

  mrwyattii

  and

  jeffra

  authored
  a3926bb

  Copy full SHA for a3926bb

Commits on Nov 16, 2023

• [Bug fix] WarmupCosineLR issues (deepspeedai#4688)

  Show description for ce0ebda
  sbwww

  authored
  ce0ebda

  Copy full SHA for ce0ebda

• Add new MII unit tests (deepspeedai#4693)

  mrwyattii

  authored
  bcdabf4

  Copy full SHA for bcdabf4

• fixes issues introduced in deepspeedai#4668

  jeffra

  committed
  5acef9e

  Copy full SHA for 5acef9e

Commits on Nov 15, 2023

• universal-ckp: support megatron-deepspeed llama model (deepspeedai#4666)

  Show description for ce5e56a
  

  3 peopleauthored
  ce5e56a

  Copy full SHA for ce5e56a

• Fix for when prompt contains an odd num of apostrophes (deepspeedai#4660)

  Show description for 00e7dc5
  

  oelayan7

  and

  loadams

  authored
  00e7dc5

  Copy full SHA for 00e7dc5

• Ensure all parameters are aligned

  cmikeh2

  committed
  25e9cd5

  Copy full SHA for 25e9cd5

Commits on Nov 14, 2023

• KV Cache Improved Flexibility (deepspeedai#4668)

  Show description for 901d807
  

  cmikeh2

  and

  tjruwase

  authored
  901d807

  Copy full SHA for 901d807

• Inference Checkpoints in V2 (deepspeedai#4664)

  Show description for 5411030
  

  8 peopleauthored
  5411030

  Copy full SHA for 5411030

• [CANN] Support cpu offload optimizer for Ascend NPU (deepspeedai#4568)

  Show description for c1ba6a1
  

  hipudding

  and

  tjruwase

  authored
  c1ba6a1

  Copy full SHA for c1ba6a1

Commits on Nov 13, 2023

• Add stable diffusion unit test (deepspeedai#2496)

  Show description for efd4556
  

  4 peopleauthored
  efd4556

  Copy full SHA for efd4556

• [op-builder] use unique exceptions for cuda issues (deepspeedai#4653)

  Show description for 953e3e3
  

  jeffra

  and

  loadams

  authored
  953e3e3

  Copy full SHA for 953e3e3

• [NPU] Add NPU support for unit test (deepspeedai#4569)

  Show description for 4b7cae7
  

  3 peopleauthored
  4b7cae7

  Copy full SHA for 4b7cae7

• [MII] catch error wrt HF version and Mistral (deepspeedai#4634)

  jeffra

  authored
  0a6095f

  Copy full SHA for 0a6095f

• Update version.txt after 0.12.3 release (deepspeedai#4673)

  Show description for 0abf4df
  

  mrwyattii

  and

  lekurile

  authored
  0abf4df

  Copy full SHA for 0abf4df

Commits on Nov 10, 2023

• fix num_kv_heads sharding in autoTP for the new in-repo Falcon-40B (deepspeedai#4654)

  Show description for 6ea44d0
  

  3 peopleauthored
  6ea44d0

  Copy full SHA for 6ea44d0

• Fix UNET and VAE implementations for new diffusers version (deepspeedai#4663)

  Show description for a361bac
  lekurile

  authored
  a361bac

  Copy full SHA for a361bac

• Update lr_schedules.py (deepspeedai#4563)

  Show description for 4388a60
  

  3 peopleauthored
  4388a60

  Copy full SHA for 4388a60

Commits on Nov 9, 2023

• Update .gitignore [Adding comments , Improved documentation] (deepspeedai#4631)

  Show description for da652d0
  

  Nadav23AnT

  and

  loadams

  authored
  da652d0

  Copy full SHA for da652d0

• Universal ckp fixes (deepspeedai#4588)

  Show description for 8ad187d
  

  mosheisland

  and

  tjruwase

  authored
  8ad187d

  Copy full SHA for 8ad187d

• Fix the openfold training. (deepspeedai#4657)

  Show description for 1d1a20c
  

  cctry

  and

  conglongli

  authored
  1d1a20c

  Copy full SHA for 1d1a20c

• Diffusers attention script update triton2.1 (deepspeedai#4573)

  Show description for 3b1cf1f
  

  3 peopleauthored
  3b1cf1f

  Copy full SHA for 3b1cf1f

Commits on Nov 8, 2023

• Update ds-chat workflow to work w/ deepspeed-chat install (deepspeedai#4598)

  Show description for 8ad50cf
  lekurile

  authored
  8ad50cf

  Copy full SHA for 8ad50cf

• tl.dot(a,b, trans_b=True) is not supported by triton2.0+ , updating this api (deepspeedai#4541)

  Show description for 217b15d
  

  bmedishe

  and

  loadams

  authored
  217b15d

  Copy full SHA for 217b15d

• Fix rope_theta arg for diffusers_attention (deepspeedai#4656)

  Show description for 00757a1
  

  lekurile

  and

  mrwyattii

  authored
  00757a1

  Copy full SHA for 00757a1

• Added __HIP_PLATFORM_AMD__=1 for non JIT build (deepspeedai#4585)

  Show description for 0daf7f2
  

  rraminen

  and

  loadams

  authored
  0daf7f2

  Copy full SHA for 0daf7f2

• add option to disable pipeline partitioning (deepspeedai#4322)

  Show description for c2074b3
  

  3 peopleauthored
  c2074b3

  Copy full SHA for c2074b3

• Enable ZeRO3 allgather for multiple dtypes (deepspeedai#4647)

  Show description for b8e1664
  

  3 peopleauthored
  b8e1664

  Copy full SHA for b8e1664

• Enhance the robustness of module_state_dict (deepspeedai#4587)

  Show description for 01af3e1
  LZHgrla

  authored
  01af3e1

  Copy full SHA for 01af3e1

Pagination

Previous

Next