Skip to content

rel_2022.4 release

Pre-release
Pre-release
Compare
Choose a tag to compare
@rogerxfeng8 rogerxfeng8 released this 16 Dec 09:34
· 88 commits to main since this release
rel_2022.4
1d4d5dc

Initial preproduction release of Intel Extension to DeepSpeed that supports GPT-2 3.6 billion parameters Large Language Model training, with below features verified on Intel GPUs:
• DeepSpeed ZeRO stage 2
• Model parallel size is 1

Assets 2
Loading