Oleg Kibirev
catplusplus
AI & ML interests
None yet
Recent Activity
updated a model 1 day ago
catplusplus/gemma-4-12B-it-uncensored-Grand-Hybrid-NVFP4-MXFP8 published a model 1 day ago
catplusplus/gemma-4-12B-it-uncensored-Grand-Hybrid-NVFP4-MXFP8 updated a model 1 day ago
catplusplus/dspark-gemma-4-12b-it-Grand-Hybrid-Unsloth-Dynamic-NVFP4-MXFP8Organizations
None yet
Optimized version and usage tips
🚀🔥 2
#12 opened 17 days ago
by
catplusplus
Where can I get your patched Nunchaku version for Qwen-Image 2.1?
1
#1 opened 27 days ago
by
mjbot
Leave it running in opencode for the whole day and come back to something useful!
#67 opened about 2 months ago
by
catplusplus
Finetuned MTP head to work better with this checkpoint
#1 opened 2 months ago
by
catplusplus
sorry, how to get this working on comfy UI?
2
#1 opened 3 months ago
by
Genaiuser9990202020awrfs
Update README.md
#1 opened 7 months ago
by
catplusplus
Pathname confusion
#1 opened 7 months ago
by
catplusplus
Made a version with full precision attention and kv cache scales
3
#1 opened 7 months ago
by
catplusplus
Would you consider model adoption to solve your storage issues?
1
#2 opened 7 months ago
by
catplusplus
Only generates nulls for me
2
#2 opened 8 months ago
by
catplusplus
Generates incoherent outputs for me with VLLM 0.18
1
#5 opened 8 months ago
by
catplusplus
VLLM + MTP + NVFP4 doesn't work
👀 1
2
#16 opened 8 months ago
by
catplusplus
Run on DGX Spark
19
#14 opened 8 months ago
by
LimeemiL
Doesn't work with latest vllm, even tried to recompile vLLM and transformers from git
➕ 1
4
#8 opened 8 months ago
by
catplusplus
All this talk about NVFP4 - why is it dog slow?
15
#13 opened 8 months ago
by
josephbreda
NVFP4 cannot be loaded in SGLang
4
#12 opened 8 months ago
by
mratsim
vLLM MTP unusable on RTX 6000 Pro, as spec decoding consumes 20GB+ VRAM at start-up, causing OOM
5
#9 opened 8 months ago
by
lsmc
Made NVFP4 version
#3 opened 9 months ago
by
catplusplus
Got better even though the higher KLD!
👍 4
3
#2 opened 9 months ago
by
FORNAX20