Heykuki News
Top
New
Best
Ask
Show
Jobs
Toggle theme
Login
Top
New
Best
Ask
Show
Jobs
YaFSDP: a sharded data parallelism framework, faster for pre-training LLMs
github.com/yandex
135 points
wiradikusuma
2 years ago
16 comments
Loading...
YaFSDP: a sharded data parallelism framework, faster for pre-training LLMs | Heykuki News