YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

BAAR2-130M

BAAR2-130M is a lightweight multilingual language model designed for fast, resource-efficient inference across multiple languages. Optimized for low latency and minimal compute environments, it enables responsive real-time generation on edge devices and standard cloud GPUs.

BAAR2-130M์€ ์ €์‚ฌ์–‘ ํ•˜๋“œ์›จ์–ด ๋ฐ ๋ฆฌ์†Œ์Šค ์ œ์•ฝ ํ™˜๊ฒฝ์—์„œ๋„ ๋น ๋ฅด๊ณ  ์•ˆ์ •์ ์ธ ๋™์ž‘์ด ๊ฐ€๋Šฅํ•˜๋„๋ก ๊ฒฝ๋Ÿ‰ํ™”๋œ ์ดˆ๊ฒฝ๋Ÿ‰ ๋‹ค๊ตญ์–ด ์–ธ์–ด ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. ์ง€์—ฐ ์‹œ๊ฐ„(Latency)์„ ์ตœ์†Œํ™”ํ•˜์—ฌ ์‹ค์‹œ๊ฐ„ ๋Œ€ํ™” ๋ฐ ํ…์ŠคํŠธ ์ƒ์„ฑ ํ™˜๊ฒฝ์— ์ตœ์ ํ™”๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.

โš ๏ธ Experimental Test Release / ์‹คํ—˜์  ํ…Œ์ŠคํŠธ ๋ชจ๋ธ
This checkpoint is an experimental test build (Proof-of-Concept) released strictly for latency evaluation, research, and multilingual inference testing. Outputs may exhibit incomplete phrasing or factual errors and are not intended for production deployments.
๋ณธ ๊ฐ€์ค‘์น˜๋Š” ๋‹ค๊ตญ์–ด ์ถ”๋ก  ์ง€์—ฐ ์‹œ๊ฐ„ ๊ฒ€์ฆ ๋ฐ ๊ธฐ์ˆ  ์‹ค์ฆ์„ ์œ„ํ•œ ์‹คํ—˜์šฉ ํ…Œ์ŠคํŠธ ์ฒดํฌํฌ์ธํŠธ(PoC)์ž…๋‹ˆ๋‹ค. ํ”„๋กœํ† ํƒ€์ž… ๋‹จ๊ณ„์ด๋ฏ€๋กœ ์‚ฌ์‹ค ๊ด€๊ณ„ ์™œ๊ณก์ด๋‚˜ ๋ฌธ์žฅ ๋ฏธ์™„์„ฑ์ด ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ƒ์šฉ ๋ฐฐํฌ์šฉ์ด ์•„๋‹Œ ์—ฐ๊ตฌ ๋ฐ ์„ฑ๋Šฅ ๊ฒ€์ฆ ๋ชฉ์ ์œผ๋กœ ๊ณต๊ฐœ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.


๐Ÿ“Š Live Sample Inference

Hardware: Single NVIDIA Tesla T4 | FP16 Inference

Task / Category Prompt Generated Output Latency
Korean General QA (ํ•œ๊ตญ์–ด ์ผ๋ฐ˜ QA) ๋Œ€ํ•œ๋ฏผ๊ตญ์˜ ์ˆ˜๋„๋Š” ์–ด๋””์ธ๊ฐ€์š”? ๋Œ€ํ•œ๋ฏผ๊ตญ์˜ ์ˆ˜๋„๋Š” ์ ์ฐจ ์•ฝ 100๋งŒ ๋…„ ~468.6ms
English General QA (์˜์–ด ์ผ๋ฐ˜ QA) What is artificial intelligence? A major intelligence that is a maj ~502.9ms
KO-to-EN Translation 1 (์ผ์ƒ์–ด ๋ฒˆ์—ญ ์š”์ฒญ) ์•ˆ๋…•์ด ์˜์–ด๋กœ ๋ญ์•ผ? ์ €๋Š” ์•ˆ๋…•์ด ์˜์–ด๋กœ Hello ์ž…๋‹ˆ๋‹ค. ~615.7ms
KO-to-EN Translation 2 (์ผ์ƒ์–ด ๋ฒˆ์—ญ ์š”์ฒญ 2) ์‚ฌ๊ณผ๊ฐ€ ์˜์–ด๋กœ ๋ญ์•ผ? ์‚ฌ๊ณผ๋Š” ์˜์–ด๋กœ apple ์ž…๋‹ˆ๋‹ค. ~439.0ms
EN-to-KO Translation (์˜->ํ•œ ๋ฒˆ์—ญ ์š”์ฒญ) What is 'Seoul' in Korean? A major member of the Constitutiona ~620.4ms
Japanese General QA (์ผ๋ณธ์–ด ์ผ๋ฐ˜ QA) ๆ—ฅๆœฌใฎ้ฆ–้ƒฝใฏใฉใ“ใงใ™ใ‹๏ผŸ ๆ—ฅๆœฌใฎ้ฆ–้ƒฝใฏใ€ ๆ—ฅๆœฌใฎ้ฆ–้ƒฝใงใ™ ~622.4ms
Chinese General QA (์ค‘๊ตญ์–ด ์ผ๋ฐ˜ QA) ไบบๅทฅๆ™บ่ƒฝ็š„ไธป่ฆ็‰น็‚นๆ˜ฏไป€ไนˆ๏ผŸ ไบบๅทฅๆ™บ่ƒฝ็š„ไธป่ฆ็‰น็‚นๆ˜ฏ ๏ผŒ ~457.8ms
Spanish General QA (์ŠคํŽ˜์ธ์–ด ์ผ๋ฐ˜ QA) ยฟCuรกl es la capital de Espaรฑa? La capital de Espaรฑa is un capi ~438.9ms
Russian General QA (๋Ÿฌ์‹œ์•„์–ด ์ผ๋ฐ˜ QA) ะšะฐะบะฐั ัั‚ะพะปะธั†ะฐ ะ ะพััะธะธ? ะกั‚ะพะปะธั†ะฐ ะ ะพััะธะธ โ€” ัั‚ะพ ัั‚ะพะปะธั†ะฐ ~463.3ms
Hindi General QA (ํžŒ๋””์–ด ์ผ๋ฐ˜ QA) เคญเคพเคฐเคค เค•เฅ€ เคฐเคพเคœเคงเคพเคจเฅ€ เค•เฅเคฏเคพ เคนเฅˆ? เคญเคพเคฐเคค เค•เฅ€ เคฐเคพเคœเคงเคพเคจเฅ€ เคนเฅˆ, เคœเฅ‹ 1960 เคฎเฅ‡เค‚ เคฐ ~438.0ms

๐Ÿ“„ License & Commercial Terms (๋ผ์ด์„ ์Šค ๋ฐ ๋„์ž… ์•ˆ๋‚ด)

  • Free Tier (๊ฐœ์ธ ๋ฌด๋ฃŒ ์ œ๊ณต): This model is available free of charge for individual developers and creators with an annual revenue under $1,000 USD.
    ์—ฐ๋งค์ถœ 1,000๋‹ฌ๋Ÿฌ(USD) ๋ฏธ๋งŒ์˜ ๊ฐœ์ธ์—๊ฒŒ๋Š” ๋ฌด๋ฃŒ๋กœ ์ œ๊ณต๋ฉ๋‹ˆ๋‹ค.
  • Enterprise & Commercial Adoption (๋„์ž… ๋ฐ ์ƒ์šฉ ๋ผ์ด์„ ์Šค ๋ฌธ์˜): If your annual revenue exceeds $1,000 USD or you require commercial integration, enterprise customization, and dedicated support, please contact us.
    ์—ฐ๋งค์ถœ 1,000๋‹ฌ๋Ÿฌ๋ฅผ ์ดˆ๊ณผํ•˜๋Š” ํ™˜๊ฒฝ์—์„œ์˜ ์ƒ์šฉ ๋„์ž…์ด๋‚˜ ๊ธฐ์—…์šฉ ์ปค์Šคํ…€ ์ ์šฉ ๋ฐ ๊ธฐ์ˆ  ์ง€์›์ด ํ•„์š”ํ•˜์‹  ๊ฒฝ์šฐ ์•„๋ž˜ ์ด๋ฉ”์ผ๋กœ ๋ฌธ์˜ํ•ด ์ฃผ์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.

โš ๏ธ Notes (์•ˆ๋‚ด ์‚ฌํ•ญ)

  • Pre-trained Model: This model is a base pre-trained checkpoint. For specific tasks, factual alignment, or structured dialogue, task-specific fine-tuning (SFT) or Retrieval-Augmented Generation (RAG) is recommended.
  • ๊ธฐ์ดˆ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ: ๋ณธ ๊ฐ€์ค‘์น˜๋Š” ์‚ฌ์ „ํ•™์Šต ๋‹จ๊ณ„์˜ ๋ฒ ์ด์Šค ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. ํŠน์ • ๋„๋ฉ”์ธ ์ž‘์—…, ์‚ฌ์‹ค ๊ด€๊ณ„ ๊ฒ€์ฆ ๋ฐ ์ •๊ตํ•œ ๋Œ€ํ™” ์ƒ์„ฑ์ด ํ•„์š”ํ•œ ๊ฒฝ์šฐ ํŒŒ์ธํŠœ๋‹(SFT) ๋˜๋Š” ๊ฒ€์ƒ‰ ์ฆ๊ฐ• ์ƒ์„ฑ(RAG) ํŒŒ์ดํ”„๋ผ์ธ๊ณผ์˜ ์—ฐ๊ณ„๋ฅผ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ’ผ Opportunities & Contact (์ฑ„์šฉ ์ œ์•ˆ ๋ฐ ํˆฌ์ž ๋ฌธ์˜)

This project demonstrates practical competency in custom model architecture design, end-to-end distributed training optimization, and efficient multi-language serving under compute constraints.

I am actively seeking AI Engineering / Research opportunities, team recruitment offers, and project investment/partnerships.

  • Model Adoption & Licensing (๋„์ž… ๋ฐ ๋ผ์ด์„ ์Šค ๋ฌธ์˜): admin@099.kr
  • Recruitment & Hiring Inquiries (์ฑ„์šฉ ๋ฐ ํฌ์ง€์…˜ ์ œ์•ˆ): admin@099.kr
  • Investment & Technical Collaboration (ํˆฌ์ž, ํŒŒํŠธ๋„ˆ์‹ญ ๋ฐ ๊ธฐ์ˆ  ํ˜‘์—…): admin@099.kr

์ œํ•œ๋œ ํ•˜๋“œ์›จ์–ด ๋ฆฌ์†Œ์Šค ํ™˜๊ฒฝ์—์„œ ๊ณ ํšจ์œจ ๋‹ค๊ตญ์–ด ๋ชจ๋ธ ์•„ํ‚คํ…์ฒ˜๋ฅผ ์ง์ ‘ ์„ค๊ณ„ํ•˜๊ณ  ํ›ˆ๋ จ ํŒŒ์ดํ”„๋ผ์ธ์„ ์—”๋“œํˆฌ์—”๋“œ๋กœ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๋Š” AI ์—”์ง€๋‹ˆ์–ด์ž…๋‹ˆ๋‹ค. ์ €์˜ ๊ธฐ์ˆ ์  ์—ญ๋Ÿ‰ ์˜์ž…(์ฑ„์šฉ)์ด๋‚˜ ํ”„๋กœ์ ํŠธ ํ˜‘์—…/ํˆฌ์ž์— ๊ด€์‹ฌ์ด ์žˆ์œผ์‹  ๊ธฐ์—… ๋ฐ ํŒ€์˜ ์—ฐ๋ฝ์„ ๊ธฐ๋‹ค๋ฆฝ๋‹ˆ๋‹ค.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support