Discussions: microsoft/DeepSpeed

Discussions

• 1 You must be logged in to vote

  🙏

  LMOD: Error while loading shared libraries: libpython3.10.so.1.0: cannot open shared object file: No such file or directory

  BramVanroy asked Jan 20, 2023 in Q&A · Answered

  

  

  4
• 1 You must be logged in to vote

  🙏

  params_in_partition in DS stage 3

  jrplatin asked Jan 20, 2023 in Q&A · Unanswered

  

  

  0
• 1 You must be logged in to vote

  🙏

  ZeRO-Offload CPU, memory utilization

  taehyunzzz asked Jan 18, 2023 in Q&A · Answered

  

  

  1
• 1 You must be logged in to vote

  💬

  I've created a working whl that installs 0.8.0 version but I need help...

  PleezDeez started Jan 16, 2023 in General

  

  

  2
• 2 You must be logged in to vote

  💬

  Some random russian guy got Triton 2.0.0 to build on windows

  PleezDeez started Jan 12, 2023 in General

  

  

  0
• 3 You must be logged in to vote

  🙏

  About kernel fusion and customized GeMM

  ariesjzj asked Nov 24, 2022 in Q&A · Unanswered

  

  

  1
• 1 You must be logged in to vote

  🙏

  CPU utilization, CPU memory for ZeRO-Offload

  taehyunzzz asked Dec 27, 2022 in Q&A · Answered

  

  

  2
• 1 You must be logged in to vote

  🙏

  parallel transformer

  raghukiran1224 asked Jan 6, 2023 in Q&A · Unanswered

  

  

  0
• 3 You must be logged in to vote

  🙏

  Triton 1.0.0 required for sparse attention but does not exist

  maxmaier59 asked Oct 8, 2022 in Q&A · Unanswered

  

  

  1
• 1 You must be logged in to vote

  🙏

  Potential configuration problem. Gradient offload not working.

  tok99bc asked Dec 7, 2022 in Q&A · Unanswered

  

  

  2
• 1 You must be logged in to vote

  💬

  Encoder-decoder inference with fast kernels

  saikrishb started Nov 2, 2022 in General

  

  

  

  

  2
• 0 You must be logged in to vote

  🙏

  About ZeRO-Offload method

  simey1128 asked Nov 21, 2022 in Q&A · Unanswered

  

  

  3
• 1 You must be logged in to vote

  💬

  how to change batch size during training?

  tnq177 started Nov 12, 2022 in General

  

  

  1
• 1 You must be logged in to vote

  💬

  how to implement custom learning rate scheduler?

  tnq177 started Oct 6, 2022 in General

  

  

  4
• 1 You must be logged in to vote

  🙏

  Adafactor optimizer work?

  toriving asked Nov 7, 2022 in Q&A · Unanswered

  

  

  1
• 1 You must be logged in to vote

  🙏

  about delayed parameter update in ZeRO-Offload

  simey1128 asked Nov 1, 2022 in Q&A · Answered

  

  

  2
• 1 You must be logged in to vote

  🙏

  Is there a way of inferring on models with int8 MoQ?

  EdouardVilain-Git asked Oct 11, 2022 in Q&A · Unanswered

  

  

  0
• 5 You must be logged in to vote

  🙏

  Overriding DeepSpeed autocasting

  gahdritz asked Jan 10, 2022 in Q&A · Unanswered

  

  

  1
• 1 You must be logged in to vote

  🙏

  How to gather loss across GPUs?

  tnq177 asked Sep 29, 2022 in Q&A · Answered

  

  

  1
• 1 You must be logged in to vote

  🙏

  is engine(input) the same as running each part of the model separately?

  tnq177 asked Sep 29, 2022 in Q&A · Unanswered

  

  

  0
• 1 You must be logged in to vote

  💬

  How to keep some layers in fp32?

  Guitaricet started Sep 27, 2022 in General

  

  

  0
• 1 You must be logged in to vote

  💡

  Understanding the need for train_batch_size in the config / variable batch size

  vince62s started Sep 27, 2022 in Ideas

  

  

  1
• 1 You must be logged in to vote

  🙏

  offload parameters during forward propagation but the CUDA memory actually be not freed

  superhao1995 asked Sep 26, 2022 in Q&A · Unanswered

  

  

  0
• 1 You must be logged in to vote

  🙏

  ZeRo-Inference across multiple GPU's

  joaopcm1996 asked Sep 20, 2022 in Q&A · Answered

  

  

  7
• 3 You must be logged in to vote

  💬

  DeepSpeed and TF XLA GPU comparison

  bharatv007 started Jul 29, 2022 in General

  

  

  0