llama.cpp
69699be4 - CUDA: fix q_nope_absorbed prec for DS 2 Lite f16 (#13137)

Commit

270 days ago

CUDA: fix q_nope_absorbed prec for DS 2 Lite f16 (#13137)

References

#13137 - CUDA: fix q_nope_absorbed precision for Deepseek 2 Lite f16

Author

JohannesGaessler

JohannesGaessler

Parents

Loading