llama.cpp
e1e8e099 - CUDA: batched+noncont MMQ, refactor bs>1 MoE code (#13199)

Commit

261 days ago

CUDA: batched+noncont MMQ, refactor bs>1 MoE code (#13199)

References

#13199 - CUDA: batched+noncont MMQ, refactor bs>1 MoE code

Author

JohannesGaessler

JohannesGaessler

Parents

Loading