v0.15.0
版本发布时间: 2024-08-23 06:46:51
microsoft/DeepSpeed最新发布版本:v0.15.1(2024-09-05 09:30:51)
What's Changed
- Update version.txt after 0.14.5 release by @loadams in https://github.com/microsoft/DeepSpeed/pull/5982
- move pynvml install to setup.py by @Rohan138 in https://github.com/microsoft/DeepSpeed/pull/5840
- add moe topk(k>2) gate support by @inkcherry in https://github.com/microsoft/DeepSpeed/pull/5881
- Move inf_or_nan_tracker to cpu for cpu offload by @BacharL in https://github.com/microsoft/DeepSpeed/pull/5826
- Enable dynamic shapes for pipeline parallel engine inputs by @tohtana in https://github.com/microsoft/DeepSpeed/pull/5481
- Add and Remove ZeRO 3 Hooks by @jomayeri in https://github.com/microsoft/DeepSpeed/pull/5658
- DeepNVMe GDS by @jomayeri in https://github.com/microsoft/DeepSpeed/pull/5852
- Pin transformers version on nv-nightly by @loadams in https://github.com/microsoft/DeepSpeed/pull/6002
- DeepSpeed on Window blog by @tjruwase in https://github.com/microsoft/DeepSpeed/pull/6364
- Bug Fix 5880 by @jomayeri in https://github.com/microsoft/DeepSpeed/pull/6378
- Update linear.py compatible with torch 2.4.0 by @terry-for-github in https://github.com/microsoft/DeepSpeed/pull/5811
- GDS Swapping Fix by @jomayeri in https://github.com/microsoft/DeepSpeed/pull/6386
- Long sequence parallelism (Ulysses) integration with HuggingFace by @samadejacobs in https://github.com/microsoft/DeepSpeed/pull/5774
- reduce cpu host overhead when using moe by @ranzhejiang in https://github.com/microsoft/DeepSpeed/pull/5578
- fix fp16 Qwen2 series model to DeepSpeed-FastGen by @ZonePG in https://github.com/microsoft/DeepSpeed/pull/6028
- Add Japanese translation of Windows support blog by @tohtana in https://github.com/microsoft/DeepSpeed/pull/6394
- Correct op_builder path to xpu files for trigger XPU tests by @loadams in https://github.com/microsoft/DeepSpeed/pull/6398
- add pip install cutlass version check by @GuanhuaWang in https://github.com/microsoft/DeepSpeed/pull/6393
- [XPU] API align with new intel pytorch extension release by @YizhouZ in https://github.com/microsoft/DeepSpeed/pull/6395
- Pydantic v2 migration by @mrwyattii in https://github.com/microsoft/DeepSpeed/pull/5167
- Fix torch check by @loadams in https://github.com/microsoft/DeepSpeed/pull/6402
New Contributors
- @Rohan138 made their first contribution in https://github.com/microsoft/DeepSpeed/pull/5840
- @terry-for-github made their first contribution in https://github.com/microsoft/DeepSpeed/pull/5811
- @ranzhejiang made their first contribution in https://github.com/microsoft/DeepSpeed/pull/5578
Full Changelog: https://github.com/microsoft/DeepSpeed/compare/v0.14.5...v0.15.0