Oleg Kibirev

catplusplus's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in nunchaku-ai/nunchaku-z-image-turbo 4 days ago

Ported nuchaku to triton kernels for NVIDIA Thor (sm_110) compatibility, would need community help to polish

#13 opened 4 days ago by

catplusplus

New activity in catplusplus/Qwen3.6-27B-uncensored-heretic-v2-NVFP4-MTP 29 days ago

Update README.md

#1 opened 29 days ago by

catplusplus

New activity in saricles/MiniMax-M2.7-REAP-172B-A10B-BF16 about 1 month ago

Pathname confusion

#1 opened about 1 month ago by

catplusplus

New activity in saricles/MiniMax-M2.7-REAP-172B-A10B-NVFP4-GB10 about 1 month ago

Made a version with full precision attention and kv cache scales

3

#1 opened about 1 month ago by

catplusplus

New activity in llmfan46/gemma-4-31B-it-uncensored-heretic about 1 month ago

Would you consider model adoption to solve your storage issues?

1

#2 opened about 1 month ago by

catplusplus

New activity in coder3101/Qwen3.5-122B-A10B-heretic-v2 about 1 month ago

quantizing to MLX 4bit removed abliteration

9

#2 opened 3 months ago by

giediprime

do the same thing you did with 27b because it does a lot better.

1

#3 opened 2 months ago by

drmcbride

New activity in coder3101/Qwen3.5-122B-A10B-heretic-v2 about 2 months ago

Uploaded an NVFP4 quant for Thor/Spark

#5 opened about 2 months ago by

catplusplus

New activity in saricles/MiniMax-M2.5-REAP-172B-A10B-NVFP4-GB10 about 2 months ago

Only generates nulls for me

2

#2 opened about 2 months ago by

catplusplus

New activity in tacos4me/Step-3.5-Flash-NVFP4 about 2 months ago

Generates incoherent outputs for me with VLLM 0.18

1

#5 opened about 2 months ago by

catplusplus

New activity in nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 2 months ago

VLLM + MTP + NVFP4 doesn't work

👀 1

2

#16 opened 2 months ago by

catplusplus

Run on DGX Spark

18

#14 opened 2 months ago by

LimeemiL

Doesn't work with latest vllm, even tried to recompile vLLM and transformers from git

➕ 1

4

#8 opened 2 months ago by

catplusplus

All this talk about NVFP4 - why is it dog slow?

15

#13 opened 2 months ago by

josephbreda

NVFP4 cannot be loaded in SGLang

4

#12 opened 2 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 2 months ago by

lsmc

New activity in llmfan46/Qwen3.5-35B-A3B-ultra-uncensored-heretic 3 months ago

Made NVFP4 version

#3 opened 3 months ago by

catplusplus

Got better even though the higher KLD!

👍 4

3

#2 opened 3 months ago by

FORNAX20

New activity in llmfan46/Qwen3.5-35B-A3B-uncensored-heretic 3 months ago

Any revision plan?

6

#2 opened 3 months ago by

Lck0427

Made catplusplus/Qwen3.5-35B-A3B-heretic-NVFP4 for Blackwell users

1

#3 opened 3 months ago by

catplusplus