I made a PWM fan hub/controller for my passive cooling gpu. Support DS18B20, NTC, manual as temperature source, Curve-based controller, scheduling,... Temporarily closed-source for development, open-source soon.
For PWM, I'm using an NPN open-collector driver, combined with a tachometer level-shifter/divider which I found from here
Looking for V100 users to share your config and it's performance.
GPU: Tesla V100 PCIE 32Gb
Qwen3.6 27B Q4_K_M + Q8_0 MTP
128K context length
Pi coding agent
llama.cpp model preset:
[*]
spec-default = 1
ctx-size = 131072
mmap = 1
kv-unified = 1
n-gpu-layers = 999
threads = 18
prio = 3
seed = 3407
image-min-tokens = 1024
batch-size = 4096
ubatch-size = 2048
parallel = 1
flash-attn = true
[Qwen3.6-27B]
model = /models/Qwen3.6/Qwen3.6-27B-Q4_K_M.gguf
mmproj = /models/mmproj/mmproj-Qwen3.6-27B-Q8_0.gguf
spec-draft-model = /models/mtp/mtp-Qwen3.6-27B-Q8_0.gguf
chat-template-file = /templates/froggeric_chat_template_v21-3.jinja
spec-type = draft-mtp
spec-draft-n-max = 1
temperature = 0.6
top-p = 0.95
top-k = 20
min-p = 0.05
presence-penalty = 0.0
repeat-penalty = 1.0
chat-template-kwargs = {"preserve_thinking": true}
And the performance:
Credits page showing ~$20 but my harness(configured and using until the problem happened) show Insufficient balance??? The API key is not credit limited, I also tries create new API key and that doesn't fix. Is there any idea what happening? Thanks.
Just appeared on my HF newsfeed. Did anyone tried yet ?
https://huggingface.co/empero-ai/Qwythos-27B-v1
https://huggingface.co/empero-ai/Qwythos-27B-v1-GGUF
