DeepSpeed
GPT-J inference support
#1670
Merged

Commits
  • fix the top1gating logic when use_rts is set to false
    Reza Yazdani committed 4 years ago
  • Modify inference-api and add some kernels to support GPT-J inference
    Reza Yazdani committed 4 years ago
  • fix kernels header
    Reza Yazdani committed 4 years ago
  • fixing some issues in kernels and API
    Reza Yazdani committed 4 years ago
  • fix half version of rotary_pos_emb kernel
    Reza Yazdani committed 4 years ago
  • Merge branch 'master' into gptj-inference-support
    jeffra committed 4 years ago
Loading