Files
ik_llama.cpp/github-data/pull_requests/174 - On Zen4 repack fp16 models to bf16_r16.md
2025-07-23 13:31:53 +02:00

456 B

🔀 #174 - On Zen4 repack fp16 models to bf16_r16

Author ikawrakow
State Closed
Created 2025-01-21
Updated 2025-01-21

Description

...when run-time-repacking is requested via -rtr

This massively improves performance. As this is opt-in, we do not worry about possible precision loss in the f16 -> bf16 conversion.