TEST: Local: unholy-v2-13b.Q5_K_M.gguf see description for details.

Описание: These video's are public, but more meant for my own feedback while i tweak settings. Feel free to comment.

This video is not cut or edited, RAW footage

Dev build: i won't announce which dev build i am using, for the reason several people are building their branch of it, and i will not name who for the reason if you perceive things as bad to not try their dev build. as "bad" is in the eye of the beholder. and i do not want these smart hard working individuals getting a bad reputation of their specific build. so this is why i do not disclose who's dev build i am currently using, hope you understand this reason. and again this test video's are for my own feedback mainly, but sharing them nonetheless.

system:

GPU:
RTX 3070 (8GB)
CPU:
11th Gen Intel(R) Core(TM) i5-11600K @ 3.90GHz 3.91 GHz
Ram :
Corsair CMH32GX4M2Z3600C18 2x16GB (32GB total)

model: unholy-v2-13b.Q5_K_M.gguf

Kobold:

Low Vram: Tagged
GPU layers: 8
Threads: 5
BLAS Batch Size: 2048
Context Size: 8192

Memory usage :
total VRAM used: 4507.41 MiB (model: 1737.77 MiB, context: 2769.64 MiB)

Results: Slow response time, but contextual talk was decent imho.