Adding IQ1_TN - 1.6875 bpw for TriLM ternary models#44

Merged

ikawrakow merged 9 commits intomainfrom

Sep 9, 2024

Commits on Sep 8, 2024

Adding iq1_tn - 1.6875 bpw for TriLM ternary models
Iwan Kawrakow
committed
iq1_tn: NEON
Iwan Kawrakow
committed
iq1_tn: faster NEON
Iwan Kawrakow
committed

Commits on Sep 9, 2024

iq2_bn: improve performance on NEON
Iwan Kawrakow
committed
iq1_tn: improve AVX2
Iwan Kawrakow
committed
iq1_tn: improve Zen4
Iwan Kawrakow
committed
iq2_bn: improve on Zen4
Iwan Kawrakow
committed
iq2_bn: improve AVX2 implementation
Iwan Kawrakow
committed
Remove unnecessary barrier in ggml_compute_forward_mul_mat
Iwan Kawrakow
committed