llama.cpp
8185710a - CUDA: use only 1 thread if fully offloaded (#2915)

Commit

1 year ago

CUDA: use only 1 thread if fully offloaded (#2915)

References

#2915 - CUDA: use 1 thread if model is fully offloaded

Author

JohannesGaessler

JohannesGaessler

Parents

Loading