Crystalcareai commited on
Commit
29fe906
1 Parent(s): c306017

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -38,7 +38,7 @@ The Deepseek team's DeepseekMoE paper was a game-changer for me, providing criti
38
 
39
  ## The Future: Continuous Improvement and Community Collaboration
40
 
41
- GemMoE is currently in beta, and I am actively working on refining the model, architecture, and implementation for full integration into the transformers library. One of the remaining challenges is GemMoE's incompatibility with distributed training via axolotl (and possibly base accelerate). I am diligently investigating the root cause of this issue and welcome any assistance or insights from the community. Once this hurdle is overcome, I will release a comprehensive technical report detailing the development process and the innovative solutions employed in GemMoE.
42
 
43
  I am incredibly excited about GemMoE's future and potential. By collaborating with the community, we can continue to refine and improve GemMoE, pushing the boundaries of what is possible with MoE architectures and limited amounts of compute.
44
 
 
38
 
39
  ## The Future: Continuous Improvement and Community Collaboration
40
 
41
+ GemMoE is currently in beta, and I am actively working on refining the model, architecture, and implementation for full integration into the transformers library. I will release a comprehensive technical report detailing the development process and the solutions employed in GemMoE.
42
 
43
  I am incredibly excited about GemMoE's future and potential. By collaborating with the community, we can continue to refine and improve GemMoE, pushing the boundaries of what is possible with MoE architectures and limited amounts of compute.
44