ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-27 01:29:51 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
94aa54df763aecde4cc5f528a26f72e9c5f2d4ff
ik_llama.cpp/github-data/pull_requests/106-Bitnet changes.md
Thomas 94aa54df76 Add GitHub data (#637)
2025-07-22 18:18:40 +02:00

1.4 KiB
Raw Blame History

🔀 #106 - Bitnet changes

Author ikawrakow
State ❌ Closed
Created 2024-10-24
Updated 2024-10-25

Description

  • Change IQ1_BN and IQ2_BN to have per row scales. In that way we can handle Bitnet models with and without separate tensor scales
  • Remove IQ1_TN and IQ2_TN. With the above change these are now redundant. IQ1_BN and IQ2_BN are also faster, so no reason to keep these around
  • Change build_bitnet() to use the standard llm_build_kv() function for the self attention portion. I was hoping this would also allow to use FA, but nope, the Bitnet models have a strange head size of 100 that is not supported by the FA implementations.

Everything works except - can you guess? - Metal. There is something wrong with the dot product kernels and I simply don't see what. I have to fix Metal before merging.

On CUDA (RTX-4080) we now get 368 t/s for TG-128 with the 3.3B Bitnet model (IQ2_BN). When I first added Bitnet support we were at ~320 t/s, so quite an improvement since then.

Update

I wasted quite some time trying to figure out why the Bitnet changes don't work on Metal. At the end it turned out that it is PR #98 that breaks the Metal back-end. So, this PR reverts #98.

@agray3 Do you have the ability to investigate why #98 breaks the Metal back-end?