Skip to content

/ DeepSpeed

New pull request New
37 Open 664 Closed
37 Open 664 Closed
Author
Filter by author
Label
Filter by label
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Milestones
Filter by milestone
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀
Added both CPU and GPU profiling Support!!
#1311 opened Aug 17, 2021 by harishsg99 • Review required
4
Add more synchronizations and barriers for the multi-gpu inference case
#1309 opened Aug 16, 2021 by RezaYazdaniAminabadi • Review required
Elasticity updates (squash) (#997)
#1290 opened Aug 9, 2021 by Kaiming-Qu • Draft
add pod level retry relaunch with worker number unchanged
#1287 opened Aug 7, 2021 by shadowtudark
FP16 fused and unfused grad norm query.
#1286 opened Aug 7, 2021 by ShadenSmith • Draft
2
CPU-Adam: add compile-flag to enable param-copy from CPU to GPU
#1280 opened Aug 5, 2021 by RezaYazdaniAminabadi • Review required
2
Correctness fix PP+ZeRO for gradient accumulation
#1264 opened Jul 29, 2021 by jeffra • Approved
1
Use scalar cpu-adam in case of exception in op-builder
#1259 opened Jul 28, 2021 by RezaYazdaniAminabadi • Review required
1
Activation checkpointing improvements
#1254 opened Jul 28, 2021 by tjruwase • Approved
1
3
Reducing the memory-overhead of creating large-models for multi-GPU run
#1244 opened Jul 20, 2021 by RezaYazdaniAminabadi • Approved
1
4
ZeRO3, improved parameter all-gather operation
#1188 opened Jun 25, 2021 by zarzen • Review required
7
switch bin files to use python3 instead of python
#1185 opened Jun 24, 2021 by jeffra • Approved
1
ZeRO-3 Slowdown
#1170 opened Jun 17, 2021 by tjruwase • Review required
18
fix checkpoint loading
#1167 opened Jun 17, 2021 by conglongli • Review required
2
Allow global steps to be reset when fine tuning
#1160 opened Jun 14, 2021 by gcooper-isi • Review required
Added retain_graph as a kwarg to the main engine backward function
#1149 opened Jun 8, 2021 by ncilfone • Review required
1
Refactor to use local vars for rank, world_size, group.
#1102 opened May 25, 2021 by awan-10 • Approved
@awan-10
2
Allow FixedSparseAttention with num_global_blocks = 0
#1027 opened May 1, 2021 by sdtblck • Review required
Add function to_sequential to PipelineModule
#1014 opened Apr 28, 2021 by sdtblck • Review required
7
[launcher/runner] respect CUDA_VISIBLE_DEVICES for a single node
#960 opened Apr 14, 2021 by stas00 • Approved
4
Add HF transformers tests
#958 opened Apr 14, 2021 by awan-10 • Review required
2
Triangular mask support for Transformer kernel
#933 opened Apr 7, 2021 by RezaYazdaniAminabadi • Review required
zero-3 dynamic tracing support
#873 opened Mar 18, 2021 by jeffra • Draft
[ignore] Check if fp16_partitions called
#871 opened Mar 18, 2021 by sshleifer • Draft
1
[Fix] Ensure we remove existing hooks before assigning new hooks using ZeRO 3
#870 opened Mar 18, 2021 by SeanNaren • Review required
4
ProTip! Type g i on any issue or pull request to go back to the issue listing page.