Skip to content

Releases: microsoft/DeepSpeed

v0.5.9: Patch release

04 Jan 23:00
d0ab722
Compare
Choose a tag to compare
Various small documentation text improvements (#1665)

Co-authored-by: Jeff Rasley <[email protected]>

v0.5.8: Patch release

01 Dec 01:16
9aa288d
Compare
Choose a tag to compare
Remove unused import of ssl.OP_ENABLE_MIDDLEBOX_COMPAT (#1601)

v0.5.7: Patch release

01 Dec 01:16
a8a17f2
Compare
Choose a tag to compare
Several fixes for our read-the-docs build (#1579)

v0.5.6: Patch release

11 Nov 17:14
2665c8b
Compare
Choose a tag to compare
Fix 1bit extra issue (#1542)

v0.5.5: Patch release

05 Nov 16:11
91defd7
Compare
Choose a tag to compare
Prevent creation of local temp directory (#1494)

Co-authored-by: Olatunji Ruwase <[email protected]>
Co-authored-by: Jeff Rasley <[email protected]>

v0.5.4: Patch release

06 Oct 16:23
bc7778e
Compare
Choose a tag to compare
Fix the workspace allocation for the transformer kernel (#1397)

* fix the workspace allocation for the transformer kernel

* change layer-id type & rm one unit test due to OOM

v0.5.3: Patch release

18 Sep 05:07
30537e7
Compare
Choose a tag to compare
[zero_to_fp32] adapt to 4-bytes alignment in z2 (#1372)

Co-authored-by: Olatunji Ruwase <[email protected]>

v0.5.2: Patch release

14 Sep 22:50
8e577c9
Compare
Choose a tag to compare
Update setup.py (#1361)

updated classifiers

v0.5.1: Patch release

26 Aug 22:24
49b6a63
Compare
Choose a tag to compare
Reducing the memory-overhead of creating model for multi-GPU run (#1244)

Co-authored-by: Jeff Rasley <[email protected]>

DeepSpeed v0.5.0

17 Aug 05:29
f284324
Compare
Choose a tag to compare
  • Mixture of Experts (MoE) support
  • Curriculum learning