Pull Requests microsoft/DeepSpeed

Refactor/Pydantify comms config

#2641 opened 2022-12-21 22:55 by mrwyattii

Refactor/Pydantify nebula checkpoint engine config

#2635 opened 2022-12-21 00:31 by mrwyattii

Refactor/Pydantify activation checkpointing config

#2633 opened 2022-12-20 23:21 by mrwyattii

Add validation for injection policy in inference config

#2630 opened 2022-12-20 18:37 by mrwyattii

Migrate W8A16 Inference to Dequantization Utility

#2580 opened 2022-12-07 02:40 by cmikeh2

Remove all unused quantize settings and flags.

#2555 opened 2022-11-28 22:38 by awan-10

Inference support for encoder-decoder architecture

#2451 opened 2022-10-28 20:45 by RezaYazdaniAminabadi

Create GPU library abstraction layer

#2437 opened 2022-10-20 22:15 by mrwyattii

Unify benchmark scripts and knowledge

#2374 opened 2022-09-28 18:10 by mrwyattii

Pydantify sub-configs

#2210 opened 2022-08-11 21:26 by mrwyattii

Configure fused fp16 mode

#1882 opened 2022-04-06 00:53 by tjruwase

Modify unit test to cover more cases

#1775 opened 2022-02-15 19:33 by RezaYazdaniAminabadi

[zero-3] add support for new params added during fwd pass

#1606 opened 2021-12-01 23:11 by jeffra

Optimizer state loading fix for bitsandbytes 8-bit optimizers.

#1582 opened 2021-11-22 06:10 by TimDettmers

Transformer/fix layer norm

#1404 opened 2021-09-27 20:49 by RezaYazdaniAminabadi

Add function `to_sequential` to PipelineModule

#1014 opened 2021-04-28 21:26 by sdtblck