Oleg Afonin
aoleg
AI & ML interests
None yet
Recent Activity
new activity 13 days ago
mradermacher/model_requests:trohrbaugh/Qwen3.8-27B-heretic-ara new activity 19 days ago
napalm/krea2_loras:When extracting loras from models, use bf16 checkpointsOrganizations
None yet
trohrbaugh/Qwen3.8-27B-heretic-ara
1
#2798 opened 13 days ago
by
aoleg
When extracting loras from models, use bf16 checkpoints
4
#3 opened 20 days ago
by
aoleg
A few more tics to fight off (not Scotoma-2 specific; pure Gemma 4 stuff).
4
#1 opened 20 days ago
by
WetRat
Sabomako/gemma-4-31B-it-heretic - new abliteration method (ORBA)
👀 1
6
#2599 opened 2 months ago
by
aoleg
Any plans to re-quantize with your improved scheme?
2
#2 opened 2 months ago
by
aoleg
Good!
🤗🚀 1
13
#1 opened 2 months ago
by
usermma
NVFP4 calibration dataset?
5
#1 opened 3 months ago
by
nohurry
Surprisingly coherent
1
#1 opened 3 months ago
by
aoleg
Bravo
❤️ 3
2
#1 opened 4 months ago
by
tcclaviger
Bro cooked
🔥 5
3
#1 opened 4 months ago
by
checki333
Version for gemma-4-26B-A4B?
1
#3 opened 4 months ago
by
Decivox
Q6 seems a perfect middle ground
❤️ 2
3
#1 opened 4 months ago
by
aoleg
Won't load in SwarmUI
2
#1 opened 4 months ago
by
aoleg
AuriAetherwiing/G4-31B-Musica-v1
1
#2267 opened 4 months ago
by
aoleg
What's the Difference Between the v1 and the v2?
👀 1
4
#1 opened 4 months ago
by
llmfan46
Size discrepancies
👍 3
4
#1 opened 4 months ago
by
aoleg
How does this method compare to other recent abliteration methods such as MPOA and ARA?
8
#1 opened 5 months ago
by
aoleg
coder3101/gemma-4-31B-it-heretic
3
#2117 opened 5 months ago
by
aoleg
MS3.2-PaintedFantasy-v4.1 is out
👀 1
3
#2 opened 6 months ago
by
aoleg