About 4% slower than Q6_K for PP-512, but 10% faster for TG-128. Someone has screwed up Q6_K TG performance on Metal? With the cobntinuous "improvements" in ggml I wouldn't be surprised. Need to look into it later. |
||
|---|---|---|
| .. | ||
| cmake | ||
| include | ||
| src | ||
| .gitignore | ||
| CMakeLists.txt | ||