models-and-training / index.html
SNAPKITTYWEST's picture
Add new October 2026 drop repositories
1012740 verified
Raw History Blame Contribute Delete
20.9 kB
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Models and Training · SnapKitty</title>
<meta name="description" content="Language models, fine-tune checkpoints, reward models and the training infrastructure behind them.">
<link rel="preconnect" href="https://fonts.googleapis.com"><link rel="preconnect" href="https://fonts.gstatic.com" crossorigin><link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&family=JetBrains+Mono:wght@500&display=swap" rel="stylesheet">
<link rel="stylesheet" href="style.css">
</head>
<body>
<header class="topbar"><a class="brand" href="https://huggingface.co/spaces/Snapkitty/README">🐈 SnapKitty</a><a class="back" href="https://huggingface.co/Snapkitty">All collections</a></header>
<main class="wrap">
<section class="hero">
<div class="eyebrow">🧠 Collection</div>
<h1>Models and Training</h1>
<p class="lede">Language models, fine-tune checkpoints, reward models and the training infrastructure behind them.</p>
</section>
<div class="toolbar">
<input id="q" type="search" placeholder="Search 21 repositories" aria-label="Search repositories">
<div class="chips"><button class="chip" data-mode="all" aria-pressed="true">All</button><button class="chip" data-mode="new" aria-pressed="false">New</button></div>
<span id="count" class="count"></span>
</div>
<section class="grid">
<article class="card" data-kind="model" data-new="true" data-search="astra-guardian-depth astra: looped-transformer reference with recurrent depth and latent reasoning, in c11, python and cuda. c python cuda model non-ai mpl-2.0">
<div class="card-top"><span class="badge new">New</span><span class="badge kind">model</span><span class="badge lic">NON-AI MPL-2.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/astra-guardian-depth">astra-guardian-depth</a></h3>
<p>Astra: looped-transformer reference with recurrent depth and latent reasoning, in C11, Python and CUDA.</p>
<div class="card-foot"><div class="tags"><span class="tag">c</span><span class="tag">python</span><span class="tag">cuda</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/astra-guardian-depth">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/astra-guardian-depth">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="true" data-search="sovereign-reduction-algebra array reduction algebra for an experimental subleq attention / bi-encoder, with exact j 9.7 and r gmp arithmetic. apl j r model agpl-3.0">
<div class="card-top"><span class="badge new">New</span><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-reduction-algebra">sovereign-reduction-algebra</a></h3>
<p>Array reduction algebra for an experimental SUBLEQ attention / bi-encoder, with exact J 9.7 and R gmp arithmetic.</p>
<div class="card-foot"><div class="tags"><span class="tag">apl</span><span class="tag">j</span><span class="tag">r</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-reduction-algebra">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovereign-reduction-algebra">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="true" data-search="tensor-roll recursive cuda-q model quantizer: compress a 30b-class teacher into an ~8b student under a hard parameter budget. python cuda-q model agpl-3.0">
<div class="card-top"><span class="badge new">New</span><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/tensor-roll">tensor-roll</a></h3>
<p>Recursive CUDA-Q model quantizer: compress a 30B-class teacher into an ~8B student under a hard parameter budget.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span><span class="tag">cuda-q</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/tensor-roll">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/tensor-roll">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="hilbert-4b-baseline-gguf hilbert 4b decision baseline as q8_0 and q4_k_m gguf, with row-level decision agreement against bf16 (q8_0: 99.3%). gguf llama.cpp model apache-2.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Apache-2.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/hilbert-4b-baseline-GGUF">hilbert-4b-baseline-GGUF</a></h3>
<p>Hilbert 4B decision baseline as Q8_0 and Q4_K_M GGUF, with row-level decision agreement against BF16 (Q8_0: 99.3%).</p>
<div class="card-foot"><div class="tags"><span class="tag">gguf</span><span class="tag">llama.cpp</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/hilbert-4b-baseline-GGUF">Hugging Face</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="lean-llm-starter starter kit connecting lean 4 formal verification with llm inference pipelines. lean4 python model mpl-2.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">mpl-2.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/lean-llm-starter">lean-llm-starter</a></h3>
<p>Starter kit connecting Lean 4 formal verification with LLM inference pipelines.</p>
<div class="card-foot"><div class="tags"><span class="tag">lean4</span><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/lean-llm-starter">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/lean-llm-starter">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="nano-gemini minimal gemini model implementation for local inference. python model tri-license">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Tri-License</span></div>
<h3><a href="https://huggingface.co/Snapkitty/nano-gemini">nano-gemini</a></h3>
<p>Minimal Gemini model implementation for local inference.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/nano-gemini">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/nano-gemini">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="pocketlearn compact learning framework for on-device model training and inference. python model tri-license">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Tri-License</span></div>
<h3><a href="https://huggingface.co/Snapkitty/pocketlearn">pocketlearn</a></h3>
<p>Compact learning framework for on-device model training and inference.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/pocketlearn">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/pocketlearn">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="snapkitty-merged nemotron mini 4.2b gguf (q4_k_m). qlora fine-tune experiment on sovereign corpus. 2.6 gb. gguf python model nvidia community">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">NVIDIA Community</span></div>
<h3><a href="https://huggingface.co/Snapkitty/snapkitty-merged">snapkitty-merged</a></h3>
<p>Nemotron Mini 4.2B GGUF (Q4_K_M). QLoRA fine-tune experiment on sovereign corpus. 2.6 GB.</p>
<div class="card-foot"><div class="tags"><span class="tag">gguf</span><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/snapkitty-merged">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/snapkitty-merged">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="snapkitty-mlc snapkitty machine learning in c: a compact c autograd and ml core library. c model agpl-3.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/snapkitty-mlc">snapkitty-mlc</a></h3>
<p>SnapKitty Machine Learning in C: a compact C autograd and ML core library.</p>
<div class="card-foot"><div class="tags"><span class="tag">c</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/snapkitty-mlc">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/snapkitty-mlc">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="snapkitty-nemotron nemotron fine-tune checkpoint on sovereign domain data. python model nvidia community">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">NVIDIA Community</span></div>
<h3><a href="https://huggingface.co/Snapkitty/snapkitty-nemotron">snapkitty-nemotron</a></h3>
<p>Nemotron fine-tune checkpoint on sovereign domain data.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/snapkitty-nemotron">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/snapkitty-nemotron">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="snapkitty-nemotron-harness deterministic local harness for nemotron. emojicode persona control, prolog and lean gates, sealed receipts. python prolog lean4 model agpl-3.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/snapkitty-nemotron-harness">snapkitty-nemotron-harness</a></h3>
<p>Deterministic local harness for Nemotron. EmojiCode persona control, Prolog and Lean gates, sealed receipts.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span><span class="tag">prolog</span><span class="tag">lean4</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/snapkitty-nemotron-harness">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/snapkitty-nemotron-harness">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="snapkitty-transformer corpus classification and training gate with datalog rules and x86-64 assembly. worm-sealed receipts. python asm datalog model bsl-1.1">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">BSL-1.1</span></div>
<h3><a href="https://huggingface.co/Snapkitty/snapkitty-transformer">snapkitty-transformer</a></h3>
<p>Corpus classification and training gate with Datalog rules and x86-64 assembly. WORM-sealed receipts.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span><span class="tag">asm</span><span class="tag">datalog</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/snapkitty-transformer">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/snapkitty-transformer">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovereign-gemini-gguf zero-dependency gguf binary parser and neural graph ir for gemini-class models without loading weights. python model tri-license">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Tri-License</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-gemini-gguf">sovereign-gemini-gguf</a></h3>
<p>Zero-dependency GGUF binary parser and neural Graph IR for Gemini-class models without loading weights.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-gemini-gguf">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovereign-gemini-gguf">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovereign-mamba mamba ssm architecture integration for sovereign inference with state-space model routing. python cuda model agpl-3.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-mamba">sovereign-mamba</a></h3>
<p>Mamba SSM architecture integration for sovereign inference with state-space model routing.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span><span class="tag">cuda</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-mamba">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovereign-mamba">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovereign-memory-twin recurrent memory twin network: evidence/silence verdicts, worm-sealed, with an ibm i ledger bridge. python model agpl-3.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-memory-twin">sovereign-memory-twin</a></h3>
<p>Recurrent Memory Twin Network: Evidence/Silence verdicts, WORM-sealed, with an IBM i ledger bridge.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-memory-twin">Hugging Face</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovereign-mimo-4b 4b parameter code reward model pruned from mimo-7b. fsm pipeline, ere gates, worm sealing. runs local on rtx 3080. python cuda gguf model tri-license">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Tri-License</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-mimo-4b">sovereign-mimo-4b</a></h3>
<p>4B parameter code reward model pruned from MiMo-7B. FSM pipeline, ERE gates, WORM sealing. Runs local on RTX 3080.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span><span class="tag">cuda</span><span class="tag">gguf</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-mimo-4b">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovereign-mimo-4b">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovereign-qra deterministic routing tensor replacing softmax in moe dispatch. zero entropy, proved bijective. lean 4 zero-sorry. lean4 verilog haskell cuda model tri-license">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Tri-License</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-qra">sovereign-qra</a></h3>
<p>Deterministic routing tensor replacing softmax in MoE dispatch. Zero entropy, proved bijective. Lean 4 zero-sorry.</p>
<div class="card-foot"><div class="tags"><span class="tag">lean4</span><span class="tag">verilog</span><span class="tag">haskell</span><span class="tag">cuda</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-qra">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovereign-qra">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovereign-transformer transformer architecture for sovereign inference pipeline with deterministic execution. python model sovereign source">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Sovereign Source</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovereign-transformer">sovereign-transformer</a></h3>
<p>Transformer architecture for sovereign inference pipeline with deterministic execution.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovereign-transformer">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovereign-transformer">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="sovlm sovereign statistical language model. n-grams, kneser-ney smoothing, bm25 retrieval, qrng sampling. python model tri-license">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">Tri-License</span></div>
<h3><a href="https://huggingface.co/Snapkitty/sovlm">sovlm</a></h3>
<p>Sovereign statistical language model. N-grams, Kneser-Ney smoothing, BM25 retrieval, QRNG sampling.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/sovlm">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/sovlm">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="summon pip-installable framework for sovereign fine-tuning. qlora, functional api, layered corpus builder, worm chain sealing. python model agpl-3.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/summon">summon</a></h3>
<p>Pip-installable framework for sovereign fine-tuning. QLoRA, functional API, layered corpus builder, WORM chain sealing.</p>
<div class="card-foot"><div class="tags"><span class="tag">python</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/summon">Hugging Face</a><a class="link" href="https://github.com/SNAPKITTYWEST/summon">GitHub</a></div></div>
</article>
<article class="card" data-kind="model" data-new="false" data-search="symbolic-morphology from-scratch symbolic learner for latin verb morphology: hand-derived gradients, no ml framework, implemented and benchmarked across 9 runtimes down to nand gates. rust f# python apl model agpl-3.0">
<div class="card-top"><span class="badge kind">model</span><span class="badge lic">AGPL-3.0</span></div>
<h3><a href="https://huggingface.co/Snapkitty/symbolic-morphology">symbolic-morphology</a></h3>
<p>From-scratch symbolic learner for Latin verb morphology: hand-derived gradients, no ML framework, implemented and benchmarked across 9 runtimes down to NAND gates.</p>
<div class="card-foot"><div class="tags"><span class="tag">rust</span><span class="tag">f#</span><span class="tag">python</span><span class="tag">apl</span></div><div class="links"><a class="link" href="https://huggingface.co/Snapkitty/symbolic-morphology">Hugging Face</a></div></div>
</article>
</section>
<section class="commercial">
<div>
<h2>Commercial license</h2>
<p>SnapKitty code is free and open under AGPL-3.0 for open-source use. Building a commercial product or service? A proprietary commercial license from Snapkitty Collective LLC lets you ship SnapKitty code without the AGPL's source-sharing and network-use obligations.</p>
</div>
<a class="btn" href="mailto:A.parr@belespritdaccord.uk?subject=SnapKitty%20commercial%20license">Get a commercial license</a>
</section>
</main>
<footer class="footer">
<span>© 2026 Snapkitty Collective LLC</span>
<span><a href="https://huggingface.co/Snapkitty">huggingface.co/Snapkitty</a> · <a href="mailto:A.parr@belespritdaccord.uk">A.parr@belespritdaccord.uk</a></span>
</footer>
<script>
(() => {
const q = document.getElementById('q'), chips = [...document.querySelectorAll('.chip')], cards = [...document.querySelectorAll('.card')];
const count = document.getElementById('count'); let mode = 'all';
function apply() {
const term = (q.value || '').trim().toLowerCase(); let n = 0;
for (const c of cards) {
const okMode = mode === 'all' || (mode === 'new' ? c.dataset.new === 'true' : c.dataset.kind === mode);
const ok = okMode && (!term || c.dataset.search.includes(term));
c.hidden = !ok; if (ok) n++;
}
count.textContent = n + (n === 1 ? ' repository' : ' repositories');
}
q.addEventListener('input', apply);
chips.forEach(ch => ch.addEventListener('click', () => { chips.forEach(x => x.setAttribute('aria-pressed', 'false')); ch.setAttribute('aria-pressed', 'true'); mode = ch.dataset.mode; apply(); }));
apply();
})();
</script>
</body>
</html>