s1arsky
s1arsky
AI & ML interests
I create pretraining datasets for fine tuning base->adventure/RP open models. Also, on occasion I fill gaps with quants for choosen ERP/RP model. Dataset related questions welcome on discord on Silly Tavern's server, under 's1arsky'.
Recent Activity
new activity about 21 hours ago
alesha-pro/Qwen3.8-Flash-Next-abliterated-GSQ-RCO-Strata-GGUF:Refusal direction for Orcarouter ? new activity 1 day ago
turboderp/Qwen3.8-Flash-Next-exl3:Uncensored EXL3Organizations
Refusal direction for Orcarouter ?
#3 opened 1 day ago
by s1arsky
Uncensored EXL3
1
#15 opened 3 days ago
by 
rvelloso
Orcarouter uncensored for Strata
#1 opened 1 day ago
by s1arsky
Ablation strength ~1.955 and coding loops: results with correction to 1.0
👀👍 4
2
#1 opened 5 days ago
by 
alesha-pro
Self calibrated quants
👍 3
1
#14 opened 10 days ago
by 
Pedromontesano
model_of_choice
👀 1
7
#8 opened about 1 month ago
by s1arsky
thumbs up
➕ 1
#1 opened 23 days ago
by s1arsky
Field manual to setting up local inference for exl3 quant model.
#11 opened 26 days ago
by s1arsky
3.05 HumanEval
👍🔥 3
9
#6 opened about 1 month ago
by 
johnlakness
4bpw work on my 3090 + 64 RAM - excellence.
❤️👍 11
22
#2 opened about 1 month ago
by s1arsky
UD_Q4_XL benchmarks ?
1
#1 opened about 1 month ago
by s1arsky
context type MTP requested but model doesn't contain MTP layers
👀 4
6
#47 opened about 1 month ago
by 
Manhandle6292
My findings for this setup.
🚀 2
1
#1 opened about 2 months ago
by s1arsky
Keeps getting stuck in thinking loops
👍 7
3
#12 opened 3 months ago
by 
jpschroeder
QAT versions please 🙏
👍 13
3
#2 opened 4 months ago
by 
Iwaku-Real
Accessing the weights
#2 opened 5 months ago
by s1arsky
New activity in groxaxo/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-heretic-EXL3-6bpw 6 months ago
Interest in 4bpw
❤️ 2
1
#1 opened 7 months ago
by 
19440harry
Request for access / IQ4_XS imatrix quantization
2
#3 opened 6 months ago
by s1arsky
Hallucinations resistance ?
🔥 1
3
#3 opened 6 months ago
by s1arsky