LocalLLaMA

LocalLLaMA's activity

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

MaziyarPanahi 

posted an update about 8 hours ago

view post

Post

84

Used nvidia/NV-Reason-CT by Nvidia and Hugging Face Jobs to make this cool demo! Let me know if you think we should add more features to it: 🔥

https://www.linkedin.com/feed/update/urn:li:activity:7515789448077398016

See translation

👀

1

1

+

Reply

codelion 

in LocalLLaMA/typed-decisions about 14 hours ago

Submission: FINAL-Bench/Darwin-27B-ZTC (general, zero-shot), accuracy 0.743

🔥 1

3

#17 opened 7 days ago by

SeaWolf-AI

Submission: FINAL-Bench/Darwin-27B-ZTC-v5 (general, then fine-tuned on train): accuracy 0.812

1

#33 opened about 18 hours ago by

SeaWolf-AI

codelion 

updated a dataset about 14 hours ago

LocalLLaMA/typed-decisions

Benchmark • Updated about 14 hours ago • 3.2k • 45.3k • 188

OzTianlu 

posted an update about 16 hours ago

view post

Post

67

zixi-51M: Not a stock voice. A voice. 🐾

NoesisLab/zixi-51m
Everyone is shipping 500B-parameter models. I shipped 51 million. With one voice. On purpose. 🎤

What: open-weights Chinese TTS. VITS2 lineage. 51.43M params. Runs on your laptop CPU — no cloud, no API key, no GPU bill, no meter running.💸

The voice: not a stock voice, not a 5,000-speaker average. One real human's voice — mine. Timbre, rhythm, breath included. The sample on themodel card is one take, default parameters, recorded by absolutely no one. 🎧

What's NOT inside:

• ❌ BERT (zero. the text encoder vibes on phones + tones alone)
• ❌ language routing (Chinese-exclusive: 77 symbols, 6 tones, one job)
• ❌ my training data (the corpus is my own writing and stays mine — open weights, closed corpus, that's the deal 🤝)

The trick: embedding transplant. Phone, tone and speaker rows inherited from the MeloTTS-Chinese base by symbol name — no random init. Themodel woke up already knowing Chinese. 🧠

Built solo, corpus to packaging, by a writer who is also a violinist. Prosody is my day job. 🎻

Use it, fork it, fine-tune your own voice. Just don't put words in a real person's mouth. 🙏

🔗 Weights: NoesisLab/zixi-51m
💻 Code: https://github.com/lizixi-0x2F/zixi-51M

See translation

• 

👀

1

1

+

Reply

codelion 

updated a dataset about 20 hours ago

LocalLLaMA/local-model-explorer-data

Viewer • Updated 33 minutes ago • 4.1k • 6.01k • 3

codelion 

in LocalLLaMA/typed-decisions about 21 hours ago

Amos 1 medium and large: general zero-shot scores, plus the large fitted on train

🔥 1

1

#31 opened 1 day ago by

mukaddamzaid

Submission: rony/ministral3-3b-reasoning-typed-decisions (Specialist) (Ministral 3 3B Reasoning, fine-tuned on train)

1

#30 opened 1 day ago by

rony

Add Amos 1 medium and large (general, zero-shot) and the large's fitted row

1

#32 opened about 22 hours ago by

mukaddamzaid

codelion 

in LocalLLaMA/typed-decisions 1 day ago

Add jev-vision-27b (general, zero-shot): acc 0.714, KL 0.318

🔥🚀 1

2

#28 opened 1 day ago by

e12ex2

Add ProKope-421M-8K (Alan Canto) to Table 2 (71.85% acc, 8K context)

1

#24 opened 3 days ago by

AlanCantoFTW

Submission: rony/gemma3-270m-typed-decisions (Specialist)

1

#27 opened 2 days ago by

rony

Submission: rony/gemma3-270m-typed-decisions (Gemma 3 270M, fine-tuned on train)[General]

1

#26 opened 2 days ago by

rony

Submission: ProKope-421M-8K (Alan Canto) — 0.7185 acc / 0.385 Brier / 8K context (specialist)

🔥 1

1

#23 opened 3 days ago by

AlanCantoFTW

Qwen3.5-4B LoRA fine-tune: results submitted, model missing from the Hub leaderboard

🔥 1

1

#22 opened 3 days ago by

wozniakos10

Leaderboard question: the current #1 was trained on test-split rows (Phocinae-Largha-150M-v1)

4

#25 opened 3 days ago by

masterset-ai

gianlucamazza 

posted an update 2 days ago

view post

Post

4,502

Upstream models my open-source projects run or my articles discuss. Each note says which it is. thenlper/gte-small is real use: RAGFS embeds file chunks with it locally. The two retrieval models from my RAG article, a cross-encoder re-ranker and a bi-encoder, are examples, labelled as such, not claims of production use.
gianlucamazza/models-used-in-my-projects-6aca0b3201f38f4c264ac9bf

See translation

🚀

2

2

🔥

1

1

+

Reply

codelion 

updated a Space 4 days ago

Running

5

DataInventor

🧪

5

Invent training rows, or curate your pretraining data

gianlucamazza 

posted an update 4 days ago

view post

Post

39

I released my own benchmark sets behind emotional-memory, my library for mood-aware recall in LLM agents. The card opens with the limits: the main set is built to favour the method, and the negative results are listed up front, e.g. on LoCoMo AFT scores F1 0.168 against 0.271 for a naive RAG baseline. Use it to check where affect-conditioned retrieval helps and where it doesn't.
gianlucamazza/emotional-memory-benchmarks

See translation

🔥

2

2

+

Reply

codelion 

in LocalLLaMA/typed-decisions 5 days ago

Add Phocinae-Largha-150M-v1 (specialist, fitted on train) to Table 2

#21 opened 5 days ago by

perrylink