Search results

w/o model-parallel usability numbers reproduce

Status: Open.

#284 In microsoft/DeepSpeed;

· g-karthik opened on Jul 6, 2020

hidden_dim constraint in transformer cuda kernel

Status: Open.

#491 In microsoft/DeepSpeed;

· szhengac opened on Oct 28, 2020

NaN loss due to mixed precision?

Status: Open.

#504 In microsoft/DeepSpeed;

· sabetAI opened on Nov 5, 2020

Question about the definitions in block sparse attention

Status: Open.

#519 In microsoft/DeepSpeed;

· szhengac opened on Nov 10, 2020

AssertionError: Default process group is not initialized

Status: Open.

#510 In microsoft/DeepSpeed;

· drfinkus opened on Nov 8, 2020

using deepspeed initialize huggingface GPT2LMHeadModel, it occured RuntimeError: "Function 'LogSoftmaxBackward' returned nan values in its 0th output."

Status: Open.

#554 In microsoft/DeepSpeed;

· LIZHICHAOUNICORN opened on Nov 24, 2020

'CUDA error: an illegal memory access was encountered' in forward

Status: Open.

#308 In microsoft/DeepSpeed;

· gongwei-130 opened on Aug 7, 2020

unnecessary gradients in case of multiple optimizers / engines

Status: Open.

#399 In microsoft/DeepSpeed;

· kracwarlock opened on Sep 10, 2020

Method for Converting a Model from Other Trainers

Status: Open.

#386 In microsoft/DeepSpeed;

· trisongz opened on Sep 9, 2020

DeepspeeD AI was started in 2019 by me, why are you all using my moniker.

Status: Open.

#406 In microsoft/DeepSpeed;

· mcwebdev opened on Sep 12, 2020

Fully half precision optimizer

Status: Open.

#447 In microsoft/DeepSpeed;

· eric-haibin-lin opened on Sep 26, 2020

For model-parallel&multi-gpu training and inference

Status: Open.

#448 In microsoft/DeepSpeed;

· hiun opened on Sep 27, 2020