turboderp
turboderp
AI & ML interests
All the things
Recent Activity
updated a collection 1 day ago
EXL3 models updated a model 1 day ago
turboderp/DeepSeek-V4-Flash-Vision-Exp-exl3 published a model 1 day ago
turboderp/DeepSeek-V4-Flash-Vision-Exp-exl3Organizations
"All GPUs must be ampere (30 series) or newer."
1
#8 opened 4 days ago
by
laserbeans
Missing vision weights from 2/3/5/6-bit quants?
4
#9 opened 3 days ago
by
HermiHg
The quantized vision tower is excellent
1
#10 opened 3 days ago
by
HermiHg
4.05bpw thinking in other languages and/or gibberish
👀 1
5
#5 opened 7 days ago
by
khronnuz
reasoning content not split on Qwen3.8-Flash-Next EXL3 — thinking stays inline in content
3
#4 opened 8 days ago
by
khronnuz
Does this support ngram nvme offloading for people with 32GB system RAM?
4
#3 opened 10 days ago
by
LukeC110
UD_Q4_XL benchmarks ?
1
#1 opened 11 days ago
by
s1arsky
Question about Calibration Traces
4
#3 opened 21 days ago
by
Interpause
Add recipe.yaml artifacts.
1
#7 opened 15 days ago
by
cloover
Is there a docker way?
2
#6 opened 16 days ago
by
JohnJ990
I love you + self calibrated 6 bit?
➕ 3
1
#4 opened 20 days ago
by
Gaboo
Thank you for your work
👍 8
6
#1 opened 27 days ago
by
vonamakitsune
Making sense of the graph
👍❤️ 2
1
#2 opened 25 days ago
by
spanspek
general
12
#1 opened about 1 month ago
by
anoane
5bpw and 6bpw are in 4.00bpw tree
1
#1 opened about 2 months ago
by
eepos
Misleading model size chart
3
#1 opened 2 months ago
by
llama-anone
Answer loop
2
#1 opened 3 months ago
by
DoktorMincs
CAN'T load in exllamv3
1
#1 opened 5 months ago
by
Jakry
About the EXL3 mention
👍 2
#1 opened 5 months ago
by
turboderp