alkinun

AtAndDev

·

alkinun

AI & ML interests

decentralize

Recent Activity

liked a model about 1 hour ago

s3nh/Tess-4-27B-abliterated

reacted to s3nh's post with 👀 about 1 hour ago

Existing methods — GPTQ, AWQ, llama.cpp's k-quants — minimize empirical loss heuristically. None of them prove they are optimal in any information-theoretic sense. ICRB-Q builds a quantization scheme that is provably optimal via the Cramér-Rao lower bound (CRB): no unbiased estimator of a weight can have lower variance than [F(θ)]⁻¹, where F is the Fisher information matrix.

reacted to s3nh's post with ❤️ about 1 hour ago

Existing methods — GPTQ, AWQ, llama.cpp's k-quants — minimize empirical loss heuristically. None of them prove they are optimal in any information-theoretic sense. ICRB-Q builds a quantization scheme that is provably optimal via the Cramér-Rao lower bound (CRB): no unbiased estimator of a weight can have lower variance than [F(θ)]⁻¹, where F is the Fisher information matrix.

View all activity

Organizations

liked a model about 1 hour ago

s3nh/Tess-4-27B-abliterated

Image-Text-to-Text • 27B • Updated 7 days ago • 107 • 2

reacted to s3nh's post with 👀❤️ about 1 hour ago

Post

310

Existing methods — GPTQ, AWQ, llama.cpp's k-quants — minimize empirical loss heuristically. None of them prove they are optimal in any information-theoretic sense. ICRB-Q builds a quantization scheme that is provably optimal via the Cramér-Rao lower bound (CRB): no unbiased estimator of a weight can have lower variance than [F(θ)]⁻¹, where F is the Fisher information matrix.

1 reply

·

replied to ProCreations's post about 1 hour ago

man.. someone get this man a sponsor

reacted to ProCreations's post with 😔🔥🤝 about 1 hour ago

Post

2631

sad grug news:
I have burnt a lot of GPU credits already just making grug and it’s variants (I do not own a workstation yet) and grug 35b has a pretty bad issue but if I continue burning GPU credits to fix it then it will take even longer to get the workstation I am working for… so the fix may take a few days or never happen.

The error:
I tested it in opencode after benchmarks were good and after release (my mistake) and it had pretty bad repetition failure and simply didn’t work. so.. yea.

4 replies

·

liked 2 models about 2 hours ago

alpindale/miquella-120b

Text Generation • 118B • Updated Feb 1, 2024 • 6 • 22

miqudev/miqu-1-70b

69B • Updated Feb 4, 2024 • 28.1k • 986

upvoted a paper about 9 hours ago

GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models

Paper • 2508.06471 • Published Aug 8, 2025 • 213

liked a model about 9 hours ago

zai-org/GLM-4.7

Text Generation • 358B • Updated Jan 29 • 62k • • 2.05k

liked a model 2 days ago

thinkingmachines/Inkling

Image-Text-to-Text • 952B • Updated 1 day ago • 7.87k • • 948

updated a model 4 days ago

aethercompute/LFM2.5-1.2B-Thinking

1B • Updated 4 days ago • 14

published a model 4 days ago

aethercompute/LFM2.5-1.2B-Thinking

1B • Updated 4 days ago • 14

liked 3 models 4 days ago

Qwen/Qwen3.5-9B

Image-Text-to-Text • 10B • Updated Mar 2 • 9.22M • • 1.72k

Qwen/Qwen3.5-9B-Base

Image-Text-to-Text • 10B • Updated Apr 23 • 381k • 94

Qwen/Qwen3.5-4B-Base

Image-Text-to-Text • 5B • Updated Apr 23 • 207k • 75

liked a dataset 4 days ago

nvidia/Open-SWE-Traces

Viewer • Updated about 23 hours ago • 207k • 8.01k • 55

liked a model 5 days ago

unsloth/Qwen3.6-27B-NVFP4

Image-Text-to-Text • 21B • Updated 5 days ago • 1.92M • 224

updated a model 5 days ago

aethercompute/lfm25-1.2b-reasoning-lora-adapter

Updated 4 days ago • 148