logo

Orion-14B

๐Ÿ‡ฐ๐Ÿ‡ทํ•œ๊ตญ์–ด | ๐ŸŒ่‹ฑ่ชž | ๐Ÿ‡จ๐Ÿ‡ณไธญๆ–‡ | ๐Ÿ‡ฏ๐Ÿ‡ตๆ—ฅๆœฌ่ชž

๐Ÿค— HuggingFaceํ™ˆํŽ˜์ด์ง€ | ๐Ÿค– ModelScopeํ™ˆํŽ˜์ด์ง€
๐ŸŽฌ HuggingFace์˜จ๋ผ์ธ ์‹œ์šฉ | ๐ŸŽซ ModelScopeๅœจ็บฟ่ฏ•็”จ
๐Ÿ˜บ GitHub
๐Ÿ“– ๊ธฐ์ˆ  ๋ฆฌํฌํŠธ

# ๋ชฉ๋ก - [๐Ÿ“– ๋ชจํ˜• ์†Œ๊ฐœ](#model-introduction) - [๐Ÿ”— ๋‹ค์šด๋กœ๋“œ ๊ฒฝ๋กœ](#model-download) - [๐Ÿ”– ํ‰๊ฐ€๊ฒฐ๊ณผ](#model-benchmark) - [๐Ÿ“Š ๋ชจํ˜• ์ถ”๋ฆฌ](#model-inference)[vllm](#vllm) [llamacpp](#llama-cpp) - [๐Ÿ“œ ์„ฑ๋ช… ํ•ฉ์˜](#declarations-license) - [๐Ÿฅ‡ ๊ธฐ์—… ์†Œ๊ฐœ](#company-introduction)
# 1. ๋ชจ๋ธ์†Œ๊ฒŒ -Orion-14B-Base๋Š” 2.5์กฐ ํ† ํฐ์˜ ๋‹ค์–‘ํ•œ ๋ฐ์ดํ„ฐ ์ง‘ํ•ฉ์œผ๋กœ ํ›ˆ๋ จ๋œ 140์–ต ๊ฐœ์˜ ํŒŒ๋ผ๋ฉ”ํ„ฐ๋ฅผ ๊ฐ€์ง„ ๋‹ค์ค‘ ์–ธ์–ด ๋ชจ๋ธ์ด๋‹ค. ์ค‘๊ตญ์–ด, ์˜์–ด, ์ผ๋ณธ์–ด, ํ•œ๊ตญ์–ด ๋ฐ ๊ธฐํƒ€ ์–ธ์–ด๋ฅผ ํฌํ•จํ•œ๋‹ค.๋‹ค์ค‘ ์–ธ์–ด ํ™˜๊ฒฝ์—์„œ ์ผ๋ จ์˜ ์—…๋ฌด์—์„œ ํƒ์›”ํ•œ ์„ฑ๋Šฅ์„ ๋ณด์ธ๋‹ค. Orion-14B ์‹œ๋ฆฌ์ฆˆ์˜ ๋ชจ๋ธ๋“ค์€ ์ฃผ์š” ๊ณต๊ฐœ ๊ธฐ์ค€ ์ธก์ •์—์„œ ์šฐ์ˆ˜ํ•œ ์„ฑ์ ์„ ๊ฑฐ๋‘์—ˆ์œผ๋ฉฐ ์—ฌ๋Ÿฌ๊ฐ€์ง€ ์ง€ํ‘œ๊ฐ€ ๋™์ผํ•œ ํŒŒ๋ผ๋ฉ”ํ„ฐ๋ฅผ ๊ฐ€์ง„ ๋‹ค๋ฅธ ๋ชจ๋ธ๋“ค์„ ํ˜„์ €ํžˆ ์ดˆ์›”ํ•œ๋‹ค. ๊ตฌ์ฒด์ ์ธ ๊ธฐ์ˆ  ๋””ํ…Œ์ผ์€ [๊ธฐ์ˆ ๋ณด๊ณ ์„œ]๋ฅผ ์ฐธ๊ณ ํ•˜์„ธ์š”. (https://github.com/OrionStarAI/Orion/blob/master/doc/Orion14B_v3.pdf)ใ€‚ - Orion-14B์‹œ๋ฆฌ์ฆˆ ๋Œ€ํ˜• ๋ชจ๋ธ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ํŠน์ง•์ด ์žˆ๋‹ค. - ๋ฒ ์ด์Šค20B ํŒŒ๋ผ๋ฉ”ํ„ฐ ๋ ˆ๋ฒจ์ธ ๋Œ€ํ˜• ๋ชจ๋ธ์˜ ์ข…ํ•ฉ์ ์ธ ํ‰๊ฐ€ ๊ฒฐ๊ณผ๊ฐ€ ์šฐ์ˆ˜ํ•˜๋‹ค - ๋‹ค๊ตญ์–ด ๋Šฅ๋ ฅ์ด ๋›ฐ์–ด๋‚˜๊ณ  ์ผ๋ณธ์–ด์™€ ํ•œ๊ตญ์–ด ํ…Œ์ŠคํŠธ ์„ธํŠธ์—์„œ ํ˜„์ €ํžˆ ์•ž์„ ๋‹ค - ๋ฏธ์„ธ์กฐ์ • ๋ชจ๋ธ์€ ์ ์‘์„ฑ์ด ๊ฐ•ํ•˜๋ฉฐ ์ธ์œ„ ํ‘œ์‹œ์˜ ๋ธ”๋ผ์ธ๋“œ ํ…Œ์ŠคํŠธ์—์„œ ํ™œ์•ฝ์ด ๋‘๋“œ๋Ÿฌ์ง„๋‹ค - ๊ธด ์ปจํ…์ŠคํŠธ ๋ฒ„์ „์€ ์ตœ๋Œ€ 320k๊นŒ์ง€ ์ง€์›ํ•˜๋Š” 200k ํ† ํฐ์— ๋›ฐ์–ด๋‚œ ๊ธด ํ…์ŠคํŠธ๋ฅผ ์ง€์ง€ํ•œ๋‹ค - ์ •๋Ÿ‰ํ™” ๋ฒ„์ „ ๋ชจ๋ธ ํฌ๊ธฐ๋ฅผ 70% ์ค„์ด๊ณ  ์ถ”๋ก  ์†๋„๋ฅผ 30% ๋†’์ด๋ฉฐ ์„ฑ๋Šฅ ์†์‹ค์„ 1% ๋ฏธ๋งŒํ•˜๋‹ค
opencompass modelcap
- ๊ตฌ์ฒด์ ์œผ๋กœ ๋งํ•˜๋ฉด Orion-14B์‹œ๋ฆฌ์ฆˆ ๋Œ€ํ˜• ์–ธ์–ด ๋ชจ๋ธ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋‚ด์šฉ์„ ํฌํ•จํ•œ๋‹ค: - **Orion-14B-Base:** 2.5์–ต ํ† ์ผ„์Šค ๋‹ค์–‘ํ™” ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ 140์–ต ํŒŒ๋ผ๋ฉ”ํ„ฐ ๊ทœ๋ชจ์˜ ๋‹ค์–ธ์–ด ๊ธฐ๋ฐ˜ ๋ชจ๋ธ. - **Orion-14B-Chat:** ๊ณ ํ€„๋ฆฌํ‹ฐ ์ฝ”ํผ์Šค ๋ฏธ์„ธ์กฐ์ •์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ ๋Œ€ํ™”ํ˜• ๋ชจ๋ธ. ๋Œ€ํ˜• ๋ชจ๋ธ ์ปค๋ฎค๋‹ˆํ‹ฐ๋ฅผ ์œ„ํ•ด ๋” ๋‚˜์€ ์‚ฌ์šฉ์ž ์ธํ„ฐ๋ž™์…˜ ๊ฒฝํ—˜์„ ์ œ๊ณตํ•˜๋„๋ก ํ•œ๋‹ค. - **Orion-14B-LongChat:** 200k ํ† ํฐ ๊ธธ์ด์— ํšจ๊ณผ์ ์ด๋ฉฐ ์ตœ๋Œ€ 320k๊นŒ์ง€ ์ง€์›ํ•˜๋ฉฐ ๊ธด ํ…์ŠคํŠธ ํ‰๊ฐ€ ์„ธํŠธ์—์„œ ๋…์  ๋ชจ๋ธ๊ณผ ๋น„๊ตํ•  ์ˆ˜ ์žˆ๋‹ค. - **Orion-14B-Chat-RAG:** ๋งž์ถฐ ์ œ์ •๋œ ๊ฒ€์ƒ‰ ํ–ฅ์ƒ ์ƒ์„ฑ ๋ฐ์ดํ„ฐ ์„ธํŠธ์—์„œ ๋ฏธ์„ธ์กฐ์ •ํ•˜์—ฌ ๊ฒ€์ƒ‰ ํ–ฅ์ƒ ์ƒ์„ฑ ์ž‘์—…์—์„œ ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์„ ์ œ๊ณตํ•œ ์ฑ„ํŒ… ๋ชจ๋ธ. - **Orion-14B-Chat-Plugin:** ํ”Œ๋Ÿฌ๊ทธ์ธ ๋ฐ ํ•จ์ˆ˜ ์ „์šฉ ์ž‘์—…์— ๋งž์ถฐ ์ œ์ •๋œ ์ฑ„ํŒ… ๋ชจ๋ธ. ์—์ด์ „ํŠธ์™€ ๊ด€๋ จ๋œ ์ƒํ™ฉ์— ์•„์ฃผ ์ž˜ ์ ์šฉ๋˜์–ด ๋Œ€ํ˜• ์–ธ์–ด ๋ชจ๋ธ์ด ํ”Œ๋Ÿฌ๊ทธ์ธ ๋ฐ ํ•จ์ˆ˜ ์ „์šฉ ์‹œ์Šคํ…œ์˜ ์—ญํ• ์„ ํ•œ๋‹ค. - **Orion-14B-Base-Int4:** int4๋กœ ๊ณ„๋Ÿ‰ํ™”ํ•˜๋Š” ๋ฒ ์ด์Šค ๋ชจ๋ธ. ๋ชจ๋ธ ํฌ๊ธฐ๋ฅผ 70%๋ฅผ ์ค„์ด๋ฉฐ ์ถ”๋ฆฌ ์†๋„๋ฅผ 30% ๋†’์—ฌ 1%์˜ ์ตœ์†Œํ•œ์˜ ์„ฑ๋Šฅ ์†์‹ค๋งŒ ๊ฐ€์ ธ์™”๋‹ค. - **Orion-14B-Chat-Int4:** int4๋กœ ๊ณ„๋Ÿ‰ํ™”ํ•˜๋Š” ๋Œ€ํ™” ๋ชจ๋ธ.
# 2. ๋‹ค์šด๋กœ๋“œ ๊ฒฝ๋กœ ๋ฐœํ‘œ๋œ ๋ชจ๋ธ ๋ฐ ๋‹ค์šด๋กœ๋“œ ๋งํฌ๋Š” ๋‹ค์Œ ํ‘œ๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”: | ๋ชจ๋ธ ๋ช…์นญ | HuggingFace๋‹ค์šด๋กœ๋“œ ๋งํฌ | ModelScope๋‹ค์šด๋กœ๋“œ ๋งํฌ | |---------------------|-----------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------| | โšพ ๋ฒ ์ด์Šค ๋ชจ๋ธ | [Orion-14B-Base](https://huggingface.co/OrionStarAI/Orion-14B-Base) | [Orion-14B-Base](https://modelscope.cn/models/OrionStarAI/Orion-14B-Base/summary) | | ๐Ÿ˜› ๋Œ€ํ™” ๋ชจ๋ธ | [Orion-14B-Chat](https://huggingface.co/OrionStarAI/Orion-14B-Chat) | [Orion-14B-Chat](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat/summary) | | ๐Ÿ“ƒ ๊ธด ์ปจํ…์ŠคํŠธ ๋ชจ๋ธ | [Orion-14B-LongChat](https://huggingface.co/OrionStarAI/Orion-14B-LongChat) | [Orion-14B-LongChat](https://modelscope.cn/models/OrionStarAI/Orion-14B-LongChat/summary) | | ๐Ÿ”Ž ๊ฒ€์ƒ‰ ํ–ฅ์ƒ ๋ชจ๋ธ | [Orion-14B-Chat-RAG](https://huggingface.co/OrionStarAI/Orion-14B-Chat-RAG) | [Orion-14B-Chat-RAG](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat-RAG/summary) | | ๐Ÿ”Œ ํ”Œ๋Ÿฌ๊ทธ์ธ ๋ชจ๋ธ | [Orion-14B-Chat-Plugin](https://huggingface.co/OrionStarAI/Orion-14B-Chat-Plugin) | [Orion-14B-Chat-Plugin](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat-Plugin/summary)| | ๐Ÿ’ผ ๋ฒ ์ด์ŠคInt4๊ณ„๋Ÿ‰ํ™” ๋ชจ๋ธ | [Orion-14B-Base-Int4](https://huggingface.co/OrionStarAI/Orion-14B-Base-Int4) | [Orion-14B-Base-Int4](https://modelscope.cn/models/OrionStarAI/Orion-14B-Base-Int4/summary) | | ๐Ÿ“ฆ ๋Œ€ํ™”Int4๊ณ„๋Ÿ‰ํ™” ๋ชจ๋ธ | [Orion-14B-Chat-Int4](https://huggingface.co/OrionStarAI/Orion-14B-Chat-Int4) | [Orion-14B-Chat-Int4](https://modelscope.cn/models/OrionStarAI/Orion-14B-Chat-Int4/summary) |
# 3. ํ‰๊ฐ€ ๊ฒฐ๊ณผ ## 3.1. ๋ฒ ์ด์Šค ๋ชจ๋ธOrion-14B-Baseํ‰๊ฐ€ ### 3.1.1. ์ „๋ฌธ ์ง€์‹ ๋ฐ ์‹œํ—˜๋ฌธ์ œ ํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ | C-Eval | CMMLU | MMLU | AGIEval | Gaokao | BBH | |--------------------|----------|----------|----------|----------|----------|----------| | LLaMA2-13B | 41.4 | 38.4 | 55.0 | 30.9 | 18.2 | 45.6 | | Skywork-13B | 59.1 | 61.4 | 62.7 | 43.6 | 56.1 | 48.3 | | Baichuan2-13B | 59.0 | 61.3 | 59.5 | 37.4 | 45.6 | 49.0 | | QWEN-14B | 71.7 | 70.2 | 67.9 | 51.9 | **62.5** | 53.7 | | InternLM-20B | 58.8 | 59.0 | 62.1 | 44.6 | 45.5 | 52.5 | | **Orion-14B-Base** | **72.9** | **70.6** | **69.9** | **54.7** | 62.1 | **56.5** | ### 3.1.2. ์ดํ•ด ๋ฐ ํ†ต์‹ ํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ |RACE-middle|RACE-high| HellaSwag| PIQA | Lambada | WSC | |--------------------|----------|----------|----------|----------|----------|----------| | LLaMA 2-13B | 63.0 | 58.9 | 77.5 | 79.8 | 76.5 | 66.3 | | Skywork-13B | 87.6 | 84.1 | 73.7 | 78.3 | 71.8 | 66.3 | | Baichuan 2-13B | 68.9 | 67.2 | 70.8 | 78.1 | 74.1 | 66.3 | | QWEN-14B | 93.0 | 90.3 | **80.2** | 79.8 | 71.4 | 66.3 | | InternLM-20B | 86.4 | 83.3 | 78.1 | **80.3** | 71.8 | 68.3 | | **Orion-14B-Base** | **93.2** | **91.3** | 78.5 | 79.5 | **78.8** | **70.2** | ### 3.1.3. OpenCompassํ‰๊ฐ€ ์„ธํŠธ ํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ | Average | Examination | Language | Knowledge | Understanding | Reasoning | |------------------|----------|----------|----------|----------|----------|----------| | LLaMA 2-13B | 47.3 | 45.2 | 47.0 | 58.3 | 50.9 | 43.6 | | Skywork-13B | 53.6 | 61.1 | 51.3 | 52.7 | 64.5 | 45.2 | | Baichuan 2-13B | 49.4 | 51.8 | 47.5 | 48.9 | 58.1 | 44.2 | | QWEN-14B | 62.4 | 71.3 | 52.67 | 56.1 | 68.8 | 60.1 | | InternLM-20B | 59.4 | 62.5 | 55.0 | **60.1** | 67.3 | 54.9 | |**Orion-14B-Base**| **64.3** | **71.4** | **55.0** | 60.0 | **71.9** | **61.6** | ### 3.1.4. ์ผ๋ณธ์–ด ํ…Œ์ŠคํŠธ ์„ธํŠธ ํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ |**Average**| JCQA | JNLI | MARC | JSQD | JQK | XLS | XWN | MGSM | |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------| | PLaMo-13B | 52.3 | 56.7 | 42.8 | 95.8 | 70.6 | 71.0 | 8.70 | 70.5 | 2.40 | | WebLab-10B | 50.7 | 66.6 | 53.7 | 82.1 | 62.9 | 56.2 | 10.0 | 72.0 | 2.40 | | ELYZA-jp-7B | 48.8 | 71.7 | 25.3 | 86.6 | 70.8 | 64.1 | 2.50 | 62.1 | 7.20 | | StableLM-jp-7B | 51.1 | 33.4 | 43.3 | **96.7** | 70.6 | 78.1 | 10.7 | 72.8 | 2.80 | | LLaMA 2-13B | 46.3 | 75.0 | 47.6 | 38.8 | 76.1 | 67.7 | 18.1 | 63.2 | 10.4 | | Baichuan 2-13B | 57.1 | 73.7 | 31.3 | 91.6 | 80.5 | 63.3 | 18.6 | 72.2 | 25.2 | | QWEN-14B | 65.8 | 85.9 | 60.7 | 97.0 | 83.3 | 71.8 | 18.8 | 70.6 | 38.0 | | Yi-34B | 67.1 | 83.8 | 61.2 | 95.2 | **86.1** | 78.5 | **27.2** | 69.2 | 35.2 | | **Orion-14B-Base** | **69.1** | **88.2** | **75.8** | 94.1 | 75.7 | **85.1** | 17.3 | **78.8** | **38.0** | ### 3.1.5. ํ•œ๊ตญ์–ด ํ…Œ์ŠคํŠธ ์„ธํŠธn-shotํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ | **Average**
n=0  n=5 | HellaSwag
n=0  n=5 | COPA
n=0  n=5 | BooIQ
n=0  n=5 | SentiNeg
n=0  n=5| |------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------| | KoGPT | 53.0    70.1 | 55.9    58.3 | 73.5    72.9 | 45.1    59.8 | 37.5    89.4 | | Polyglot-ko-13B | 69.6    73.7 |**59.5**    **63.1**|**79.4**    **81.1**| 48.2    60.4 | 91.2    90.2 | | LLaMA 2-13B | 46.7    63.7 | 41.3    44.0 | 59.3    63.8 | 34.9    73.8 | 51.5    73.4 | | Baichuan 2-13B | 52.1    58.7 | 39.2    39.6 | 60.6    60.6 | 58.4    61.5 | 50.3    72.9 | | QWEN-14B | 53.8    73.7 | 45.3    46.8 | 64.9    68.9 | 33.4    83.5 | 71.5    95.7 | | Yi-34B | 54.2    72.1 | 44.6    44.7 | 58.0    60.6 | 65.9    90.2 | 48.3    92.9 | |**Orion-14B-Base**|**74.5**    **79.6**| 47.0    49.6 | 77.7    79.4 |**81.6**    **90.7**|**92.4**    **98.7**| ### 3.1.6. ๋‹ค๊ตญ์–ด ํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ | Train Lang | Japanese | Korean | Chinese | English | |--------------------|------------|----------|----------|----------|----------| | PLaMo-13B | En,Jp | 52.3 | * | * | * | | Weblab-10B | En,Jp | 50.7 | * | * | * | | ELYZA-jp-7B | En,Jp | 48.8 | * | * | * | | StableLM-jp-7B | En,Jp | 51.1 | * | * | * | | KoGPT-6B | En,Ko | * | 70.1 | * | * | | Polyglot-ko-13B | En,Ko | * | 70.7 | * | * | | Baichuan2-13B | Multi | 57.1 | 58.7 | 50.8 | 57.1 | | Qwen-14B | Multi | 65.8 | 73.7 | 64.5 | 65.4 | | Llama2-13B | Multi | 46.3 | 63.7 | 41.4 | 55.3 | | Yi-34B | Multi | 67.1 | 72.2 | 58.7 | **68.8** | | **Orion-14B-Base** | Multi | **69.1** | **79.5** | **67.9** | 67.3 | ## 3.2. ๋Œ€ํ™” ๋ชจ๋ธOrion-14B-Chatํ‰๊ฐ€ ### 3.2.1. ๋Œ€ํ™” ๋ชจ๋ธMTBench์ฃผ๊ด€์  ํ‰๊ฐ€ | ๋ชจ๋ธ ๋ช…์นญ | 1๋ผ์šด๋“œ | 2๋ผ์šด๋“œ | **ํ‰๊ท ** | |----------------------|----------|----------|----------| | Baichuan2-13B-Chat | 7.05 | 6.47 | 6.76 | | Qwen-14B-Chat | 7.30 | 6.62 | 6.96 | | Llama2-13B-Chat | 7.10 | 6.20 | 6.65 | | InternLM-20B-Chat | 7.03 | 5.93 | 6.48 | | **Orion-14B-Chat** | **7.68** | **7.07** | **7.37** | \*์ด ํ‰๊ฐ€๋Š” vllm์„ ์ด์šฉํ•˜์—ฌ ์ถ”๋ฆฌํ•œ๋‹ค ### 3.2.2. ๋Œ€ํ™” ๋ชจ๋ธAlignBench์ฃผ๊ด€์  ํ‰๊ฐ€ | ๋ชจ๋ธ ๋ช…์นญ | ์ˆ˜ํ•™ ๋Šฅ๋ ฅ | ๋…ผ๋ฆฌ์  ์ถ”๋ฆฌ | ๊ธฐ๋ณธ ๋Šฅ๋ ฅ | ์ค‘๊ตญ์–ด ์ดํ•ด | ์ข…ํ•ฉ์  ๋ฌธ๋‹ต | ๊ธ€์“ฐ๊ธฐ ๋Šฅ๋ ฅ | ๋กค ํ”Œ๋ ˆ์ด | ์ „๋ฌธ ์ง€์‹ | **ํ‰๊ท ** | |--------------------|----------|----------|----------|----------|----------|----------|----------|----------|----------| | Baichuan2-13B-Chat | 3.76 | 4.07 | 6.22 | 6.05 | 7.11 | 6.97 | 6.75 | 6.43 | 5.25 | | Qwen-14B-Chat | **4.91** | **4.71** | **6.90** | 6.36 | 6.74 | 6.64 | 6.59 | 6.56 | **5.72** | | Llama2-13B-Chat | 3.05 | 3.79 | 5.43 | 4.40 | 6.76 | 6.63 | 6.99 | 5.65 | 4.70 | | InternLM-20B-Chat | 3.39 | 3.92 | 5.96 | 5.50 | **7.18** | 6.19 | 6.49 | 6.22 | 4.96 | | **Orion-14B-Chat** | 4.00 | 4.24 | 6.18 | **6.57** | 7.16 | **7.36** | **7.16** | **6.99** | 5.51 | \*์ด ํ‰๊ฐ€๋Š” vllm์„ ์ด์šฉํ•˜์—ฌ ์ถ”๋ฆฌํ•œ๋‹ค ## 3.3. ๊ธด ์ปจํ…์ŠคํŠธ ๋ชจ๋ธOrion-14B-LongChatํ‰๊ฐ€ ### 3.3.1. ๊ธด ์ปจํ…์ŠคํŠธ ๋ชจ๋ธLongBenchํ‰๊ฐ€ | ๋ชจ๋ธ ๋ช…์นญ | NarrativeQA| MultiFieldQA-en| MultiFieldQA-zh | DuReader | QMSum | VCSUM | TREC | TriviaQA | LSHT | RepoBench-P | |--------------------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|-----------| | GPT-3.5-Turbo-16k | **23.60** | **52.30** | **61.20** | 28.70 | 23.40 | **16.00** | 68.00 | **91.40** | 29.20 | 53.60 | | LongChat-v1.5-7B-32k | 16.90 | 41.40 | 29.10 | 19.50 | 22.70 | 9.90 | 63.50 | 82.30 | 23.20 | 55.30 | | Vicuna-v1.5-7B-16k | 19.40 | 38.50 | 43.00 | 19.30 | 22.80 | 15.10 | 71.50 | 86.20 | 28.80 | 43.50 | | Yi-6B-200K | 14.11 | 36.74 | 22.68 | 14.01 | 20.44 | 8.08 | 72.00 | 86.61 | 38.00 | **63.29** | | Orion-14B-LongChat | 19.47 | 48.11 | 55.84 | **37.02** | **24.87** | 15.44 | **77.00** | 89.12 | **45.50** | 54.31 | ## 3.4. ๊ฒ€์ƒ‰ ํ–ฅ์ƒ ๋ชจ๋ธOrion-14B-Chat-RAGํ‰๊ฐ€ ### 3.4.1. ์ž๊ธฐ ๋งŒ๋“  ๊ฒ€์ƒ‰ ํ–ฅ์ƒ ํ…Œ์ŠคํŠธ ์„ธํŠธ ํ‰๊ฐ€ ๊ฒฐ๊ณผ |๋ชจ๋ธ ๋ช…์นญ|์‘๋‹ต ํšจ๊ณผ(ํ‚ค์›Œ๋“œ)|*์‘๋‹ต ํšจ๊ณผ(์ฃผ๊ด€์  ์ ์ˆ˜)|์ธ์šฉ ๋Šฅ๋ ฅ|๊ธฐ๋ณธ ๋– ๋งก๋Š” ๋Šฅ๋ ฅ|*AutoQA|*๋ฐ์ดํ„ฐ ์ถ”์ถœ| |---------------------|------|------|------|------|------|------| | Baichuan2-13B-Chat | 85 | 76 | 1 | 0 | 69 | 51 | | Qwen-14B-Chat | 79 | 77 | 75 | 47 | 68 | 72 | | Qwen-72B-Chat(Int4) | 87 | 89 | 90 | 32 | 67 | 76 | | GPT-4 | 91 | 94 | 96 | 95 | 75 | 86 | | Orion-14B-Chat-RAG | 86 | 87 | 91 | 97 | 73 | 71 | \* ์‚ฌ๋žŒ ํ‰๊ฐ€ ๊ฒฐ๊ณผ๋ฅผ ๊ฐ€๋ฆฌํ‚จ๋‹ค ## 3.5. ํ”Œ๋Ÿฌ๊ทธ์ธ ๋ชจ๋ธOrion-14B-Chat-Pluginํ‰๊ฐ€ ### 3.5.1. ์ž๊ธฐ ๋งŒ๋“ ํ”Œ๋Ÿฌ๊ทธ์ธ ํ…Œ์ŠคํŠธ ์„ธํŠธ ํ‰๊ฐ€ ๊ฒฐ๊ณผ | ๋ชจ๋ธ ๋ช…์นญ | ํ’€ ํŒŒ๋ผ๋ฉ”ํ„ฐ ์˜๋„ ์‹๋ณ„ | ๋ถˆ์™„์ „ ํŒŒ๋ผ๋ฉ”ํ„ฐ ์˜๋„ ์‹๋ณ„ | ๋น„ ํ”Œ๋Ÿฌ๊ทธ์ธ ์ „์šฉ ์‹๋ณ„ | |-----------------------|--------|-----------|--------| | Baichuan2-13B-Chat | 25 | 0 | 0 | | Qwen-14B-Chat | 55 | 0 | 50 | | GPT-4 | **95** | 52.38 | 70 | | Orion-14B-Chat-Plugin | 92.5 | **60.32** | **90** | ## 3.6. ๊ณ„๋Ÿ‰ํ™” ๋ชจ๋ธOrion-14B-Base-Int4ํ‰๊ฐ€ ### 3.6.1. ๊ณ„๋Ÿ‰ํ™” ์ „ํ›„ ์ „๋ฐ˜์ ์ธ ๋น„๊ต |๋ชจ๋ธ ๋ช…์นญ|๋ชจ๋ธ ํฌ๊ธฐ(GB)|์ถ”๋ฆฌ ์†๋„(ํ† ํฐ ์ˆ˜/์ดˆ)|C-Eval |CMMLU |MMLU |RACE | HellaSwag| |-------------------------|------|-----|------|------|------|------|------| | OrionStar-14B-Base | 28.0 | 135 | 72.8 | 70.6 | 70.0 | 93.3 | 78.5 | | OrionStar-14B-Base-Int4 | 8.3 | 178 | 71.8 | 69.8 | 69.2 | 93.1 | 78.0 |
# 4. ๋ชจ๋ธ ์ถ”๋ฆฌ ์ถ”๋ฆฌ์— ํ•„์š”ํ•œ ๋ชจ๋ธ ๊ฐ€์ค‘์น˜, ์†Œ์Šค ์ฝ”๋“œ, ๋ฐฐ์น˜๋Š” Hugging Face์— ๊ฒŒ์‹œ๋˜์–ด ๋‹ค์šด๋กœ๋“œ ๋งํฌ๋Š” ์ด ํŒŒ์ผ ๋งจ ์ฒ˜์Œ์— ์žˆ๋Š” ํ‘œ๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”. ์ €ํฌ๋Š” ์—ฌ๊ธฐ์„œ ๋‹ค์–‘ํ•œ ์ถ”๋ฆฌ ๋ฐฉ์‹์„ ๋ณด์—ฌ ์ฃผ๊ณ  ํ”„๋กœ๊ทธ๋žจ์€ Hugging Face๋กœ๋ถ€ํ„ฐ ํ•„์š”ํ•œ ์ž๋ฃŒ๋ฅผ ์ž๋™์œผ๋กœ ๋‹ค์šด๋กœ๋“œ ํ•  ๊ฒƒ์ด๋‹ค. ## 4.1. Python ์ฝ”๋“œ ๋ฐฉ์‹ ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer from transformers.generation.utils import GenerationConfig tokenizer = AutoTokenizer.from_pretrained("OrionStarAI/Orion-14B", use_fast=False, trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("OrionStarAI/Orion-14B", device_map="auto", torch_dtype=torch.bfloat16, trust_remote_code=True) model.generation_config = GenerationConfig.from_pretrained("OrionStarAI/Orion-14B") messages = [{"role": "user", "content": "์•ˆ๋…•! ์ด๋ฆ„์ด ๋ญ์˜ˆ์š”!"}] response = model.chat(tokenizer, messages, streaming=Flase) print(response) ``` ์œ„์˜ ๋‘ ์ฝ”๋“œ์—์„œ ๋ชจ๋ธ์€ ์ง€์ •๋œ `device_map='auto'`๋กœ๋”ฉํ•˜๋ฉด ๋ชจ๋“  ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ๊ทธ๋ž˜ํ”ฝ ์นด๋“œ๋ฅผ ์‚ฌ์šฉํ•  ๊ฒƒ์ด๋‹ค. ์‚ฌ์šฉํ•  ์žฅ์น˜๋ฅผ ์ง€์ •ํ•˜๋ ค๋ฉด `export CUDA_VISIBLE_DEVICES=0,1`(๊ทธ๋ž˜ํ”ฝ ์นด๋“œ 0๊ณผ 1์„ ์‚ฌ์šฉ)๊ณผ ๊ฐ™์€ ๋ฐฉ์‹์œผ๋กœ ์ œ์–ดํ•  ์ˆ˜ ์žˆ๋‹ค. ## 4.2. ๋ช…๋ น์ค„ ํˆด ๋ฐฉ์‹ ```shell CUDA_VISIBLE_DEVICES=0 python cli_demo.py ``` ์ด ๋ช…๋ น์ค„ ํˆด์€ Chat ์‹œ๋‚˜๋ฆฌ์˜ค๋ฅผ ์œ„ํ•ด ์„ค๊ณ„๋˜์—ˆ์œผ๋ฏ€๋กœ ์ด ํˆด๋กœ ๋ฒ ์ด์Šค ๋ชจ๋ธ์„ ์ „์šฉํ•˜๋Š” ๊ฒƒ ์ง€์›ํ•˜์ง€ ์•Š๋Š”๋‹ค. ## 4.3. ์Šคํฌ๋ฆฝํŠธ ์ง์ ‘ ์ถ”๋ฆฌ ```shell # base model CUDA_VISIBLE_DEVICES=0 python demo/text_generation_base.py --model OrionStarAI/Orion-14B --tokenizer OrionStarAI/Orion-14B --prompt ์•ˆ๋…•. ์ด๋ฆ„์ด ๋ญ์˜ˆ์š” # chat model CUDA_VISIBLE_DEVICES=0 python demo/text_generation.py --model OrionStarAI/Orion-14B-Chat --tokenizer OrionStarAI/Orion-14B-Chat --prompt ์•ˆ๋…•. ์ด๋ฆ„์ด ๋ญ์˜ˆ์š” ```
## 4.4. vLLM ์ถ”๋ก ์„ ํ†ตํ•ด - ํ”„๋กœ์ ํŠธ ์ฃผ์†Œ
https://github.com/vllm-project/vllm - ํ’€ ๋ฆฌํ€˜์ŠคํŠธ
https://github.com/vllm-project/vllm/pull/2539
## 4.5. llama.cpp ์ถ”๋ก ์„ ํ†ตํ•ด - ํ”„๋กœ์ ํŠธ ์ฃผ์†Œ
https://github.com/ggerganov/llama.cpp - ํ’€ ๋ฆฌํ€˜์ŠคํŠธ
https://github.com/ggerganov/llama.cpp/pull/5118 - GGUF ํ˜•์‹์œผ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๋ฐฉ๋ฒ• ```shell python convert-hf-to-gguf.py path/to/Orion-14B-Chat --outfile chat.gguf ``` - ๋ชจ๋ธ ์ถ”๋ก  ๋ฐฉ๋ฒ• ```shell ./main --frequency-penalty 0.5 --frequency-penalty 0.5 --top-k 5 --top-p 0.9 -m chat.gguf -p "Building a website can be done in 10 simple steps:\nStep 1:" -n 400 -e ``` ## 4.6. ์˜ˆ์‹œ ๋…ธ์ถœ ### 4.6.1. ์žก๋‹ด ````` ์‚ฌ์šฉ์ž๏ผš์•ˆ๋…•,์ด๋ฆ„์ด ๋ญ์˜ˆ์š” Orion-14B๏ผš์•ˆ๋…•! ์ €๋Š” ์ธ๊ณต์ง€๋Šฅ ์กฐ์ˆ˜ ์ž…๋‹ˆ๋‹ค. ์ •ํ•ด์ง„ ์ด๋ฆ„์€ ์—†๊ณ  ๋งˆ์Œ๋Œ€๋กœ ๋ถˆ๋Ÿฌ๋„ ๋ฉ๋‹ˆ๋‹ค. ๋„์™€์ค„ ์ˆ˜ ์žˆ๋Š” ๊ฑฐ ์žˆ๋‚˜์š”? ์‚ฌ์šฉ์ž๏ผš์ด์•ผ๊ธฐ ํ•˜๋‚˜ ๋‚˜๋ˆ ์ค˜ Orion-14B๏ผš์˜ˆ์ „์— ์žญ์ด๋ผ๋Š” ์–ด๋ฆฐ ์†Œ๋…„์ด ์žˆ์—ˆ๋‹ค. ๊ทธ๋Š” ์ž‘์€ ๋งˆ์„์— ์‚ด์•˜๋‹ค. ์žญ์€ ๋ชจํ—˜๊ณผ ํƒ๊ตฌ๋ฅผ ๋งค์šฐ ์ข‹์•„ํ•ด์„œ ํ•ญ์ƒ ์‹ ๊ธฐํ•œ ๊ฒƒ์„ ๋ฐœ๊ฒฌํ•˜๋Š” ๊ฟˆ์„ ๊พผ๋‹ค. ์–ด๋Š๋‚  ์žญ์€ ๊ทผ์ฒ˜์˜ ์ˆฒ ๊นŠ์€ ๊ณณ์— ์ „์„ค์†์˜ ๋งˆ๋ฒ•์„ฑ์ด ์žˆ์–ด ๊ทธ ์•ˆ์— ์‚ฌ์•…ํ•œ ๊ฑฐ์ธ์ด ์‚ฐ๋‹ค๋Š” ์†Œ์‹์„ ๋“ ๋‹ค. ๊ทธ ๊ฑฐ์ธ์˜ ์ด๋ฆ„์€ ๊ทธ๋ฆผ์ด์—ˆ๋‹ค. ๊ทธ๋Š” ์„ฑ ์•ˆ์— ๊ฐ‡ํžŒ ์ฑ„ ๋– ๋‚  ์ˆ˜ ์—†์—ˆ๋‹ค. ์žญ์€ ๊ทธ๋ฆผ์ด ๊ทธ๊ณณ์— ๊ฐ‡ํ˜€์žˆ๋Š” ์ด์œ ๊ฐ€ ๊ผญ ์žˆ์„ ๊ฑฐ๋ผ๊ณ  ๋ฏฟ๊ธฐ ๋•Œ๋ฌธ์— ๊ฑฐ์ธ์„ ๊ตฌํ•˜๋Ÿฌ ๊ฐ€๊ธฐ๋กœ ๊ฒฐ์‹ฌํ•œ๋‹ค. ๊ทธ๋ฆฌํ•˜์—ฌ ๊ทธ๋Š” ์šฉ๊ธฐ์™€ ๊ฒฐ์‹ฌ์„ ๊ฐ€์ ธ ๋งˆ๋ฒ•์˜ ์„ฑ์œผ๋กœ ๊ฐ€๋Š” ์—ฌ์ •์„ ์‹œ์ž‘ํ•œ๋‹ค. ์˜ค๋žœ ๊ณ ์ƒ ๋์— ์žญ์€ ๋งˆ์นจ๋‚ด ์„ฑ๋ณด ๋Œ€๋ฌธ ์•ž์— ๋‹ค๋‹ค๋ž๋‹ค. ๊ทธ๋Š” ์กฐ์‹ฌ์Šค๋Ÿฝ๊ฒŒ ๋Œ€๋ฌธ์„ ์—ด๊ณ  ์‹ ๋น„ํ•œ ๊ธฐ์šด์ด ๊ฐ€๋“ํ•œ ๋ฐฉ์œผ๋กœ ๋“ค์–ด๊ฐ”๋‹ค. ๋ฐฉ์•ˆ์— ์–ด๋‘์šด ๊ธฐ์šด์ด ๊ฐ€๋“ํ–ˆ์ง€๋งŒ ์žญ์€ ๋†€๋ผ์ง€ ์•Š์•˜๋‹ค. ๊ทธ๋Š” ์šฉ๊ฐํ•˜๊ฒŒ ์„ฑ ์•ˆ์œผ๋กœ ๋“ค์–ด๊ฐ€ ๋งˆ์นจ๋‚ด ๊ทธ๋ฆผ์„ ์ฐพ์•„๋ƒˆ๋‹ค. ๊ทธ๋ฆผ์€ ์ž์‹ ์€ ์‚ฌ์•…ํ•œ ๋งˆ๋…€์˜ ์ €์ฃผ๋ฅผ ๋ฐ›์•„์„œ ์ €์ฃผ๋ฅผ ํ’€์–ด์•ผ๋งŒ ์ž์œ ๋กœ์›Œ์งˆ ์ˆ˜ ์žˆ๋‹ค๊ณ  ์žญ์—๊ฒŒ ๋งํ•œ๋‹ค. ์žญ์€ ๊ทธ๋ฆผ์„ ๋•๊ธฐ๋กœ ๊ฒฐ์ •ํ•˜์—ฌ ๋‹จ์„œ์™€ ์ฃผ๋ฌธ์„ ํ’€ ๋ฐฉ๋ฒ•์„ ์ฐพ๊ธฐ ์‹œ์ž‘ํ•œ๋‹ค. ํ•œ๋ฐ”ํƒ•์˜ ๋…ธ๋ ฅ ๋์— ์žญ์€ ๋“œ๋””์–ด ๋งˆ๋…€์˜ ๋น„๋ฐ€ ์‹คํ—˜์‹ค๋กœ ๊ฐ€๋Š” ๊ธธ์„ ์ฐพ์•„๋‚ธ๋‹ค. ๊ทธ๋Š” ๊ทธ๊ณณ์—์„œ ๋งˆ๋…€๊ฐ€ ๋‚จ๊ธด ๋งˆ๋ฒ•์„œ๋ฅผ ๋ฐœ๊ฒฌํ–ˆ๋Š”๋ฐ ๊ทธ ์ฑ…์—์„œ ์ฃผ๋ฌธ์„ ํ‘ธ๋Š” ๋ฐฉ๋ฒ•์ด ์ ํ˜€ ์žˆ์—ˆ๋‹ค. ์žญ์€ ์ฑ…์— ์“ฐ์ธ ์ ˆ์ฐจ๋ฅผ ๋”ฐ๋ผํ•˜์—ฌ ๊ทธ๋ฆผ์˜ ์ €์ฃผ๋ฅผ ์„ฑ๊ณต์ ์œผ๋กœ ํ’€์–ด๋ƒˆ๋‹ค. ๊ฑฐ์ธ์€ ๋„ˆ๋ฌด ๊ฐ์‚ฌํ•ด์„œ ์žญ์˜ ์†Œ์›์ด ์ด๋ฃจ์–ด์ง€๋„๋ก ๋•๊ฒ ๋‹ค๊ณ  ์•ฝ์†ํ–ˆ๋‹ค. ์ด๋•Œ๋กœ๋ถ€ํ„ฐ ์žญ๊ณผ ๊ทธ๋ฆผ์€ ์ข‹์€ ์นœ๊ตฌ๊ฐ€ ๋˜์–ด ๋งŽ์€ ์œ„ํ—˜ํ•˜๊ณ  ์งœ๋ฆฟํ•œ ๋ชจํ—˜์„ ํ•จ๊ป˜ ๊ฒช์—ˆ์œผ๋ฉฐ ๊ทธ๋“ค์˜ ์šฉ๊ธฐ์™€ ์ง€ํ˜œ๋กœ ๊ฐ์ข… ์–ด๋ ค์›€์„ ์ด๊ฒจ๋ƒˆ๋‹ค. ์ด ์ด์•ผ๊ธฐ๋Š” ์ €ํฌ์—๊ฒŒ ์šฉ๊ธฐ์™€ ๊ฒฐ์‹ฌ์ด ์žˆ๋‹ค๋ฉด ๋ชจ๋“  ์–ด๋ ค์›€์„ ๊ทน๋ณตํ•˜๊ณ  ์ž์‹ ์˜ ๊ฟˆ์„ ์ด๋ฃฐ ์ˆ˜ ์žˆ๋‹ค๋Š” ๊ฒƒ์„ ์•Œ๋ ค์ค€๋‹ค. ````` ### 4.6.2. ํ•œ์ผ ````` ็”จๆˆท๏ผš่‡ชๅทฑใ‚’็ดนไป‹ใ—ใฆใใ ใ•ใ„ Orion-14B๏ผšใ“ใ‚“ใซใกใฏใ€็งใฎๅๅ‰ใฏChatMaxใงใ€OrionStarใซใ‚ˆใฃใฆ้–‹็™บใ•ใ‚ŒใŸAIใ‚ขใ‚ทใ‚นใ‚ฟใƒณใƒˆใงใ™ใ€‚็งใฏใƒฆใƒผใ‚ถใƒผใซๆญฃ็ขบใงใ€ๅฎข่ฆณ็š„ใงใ€ๆœ‰็”จใชๅ›ž็ญ”ใ‚’ๆไพ›ใ—ใ€ๅฝผใ‚‰ใŒๅ•้กŒใ‚’่งฃๆฑบใ—ใ€ๆƒ…ๅ ฑใ‚’ๅ–ๅพ—ใ™ใ‚‹ใฎใ‚’ๅŠฉใ‘ใ‚‹ใ“ใจใŒใงใใพใ™ใ€‚ ็”จๆˆท๏ผš์ž๊ธฐ์†Œ๊ฐœ๋ฅผ ํ•ด์ฃผ์„ธ์š”. Orion-14B๏ผš์•ˆ๋…•ํ•˜์„ธ์š”, ์ œ ๋ช…์นญ์€ ChatMax์ž…๋‹ˆ๋‹ค. OrionStar์—์„œ ๊ฐœ๋ฐœํ•œ AI ์–ด์‹œ์Šคํ„ดํŠธ์ž…๋‹ˆ๋‹ค. ์ €๋Š” ์‚ฌ์šฉ์ž๋“ค์—๊ฒŒ ์ •ํ™•ํ•˜๊ณ  ๊ฐ๊ด€์ ์ด๋ฉฐ ์œ ์šฉํ•œ ๋‹ต๋ณ€์„ ์ œ๊ณตํ•˜์—ฌ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ณ  ์ •๋ณด๋ฅผ ์–ป๋Š” ๋ฐ ๋„์›€์„ ์ค„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. `````
# 5. ์„ฑ๋ช…, ํ˜‘์˜ ## 5.1. ์„ฑ๋ช… ์ €ํฌ๋Š” ๋ชจ๋“  ์‚ฌ์šฉ์ž๋“ค์—๊ฒŒ Orion-14B๋ชจ๋ธ์„ ์ด์šฉํ•˜์—ฌ ๊ตญ๊ฐ€ ์‚ฌํšŒ ์•ˆ์ „์— ํ•ด์น˜๊ฑฐ๋‚˜ ๋ถˆ๋ฒ•์ ์ธ ํ–‰์œ„๋ฅผ ํ•˜๋Š” ๊ฑฐ ํ•˜์ง€ ์•Š๋„๋ก ๊ฐ•๋ ฅํžˆ ํ˜ธ์†Œํ•œ๋‹ค. ๋˜ํ•œ, ์ €ํฌ๋Š” ์‚ฌ์šฉ์ž๋“ค์—๊ฒŒ Orion-14B ๋ชจ๋ธ์„ ์ ์ ˆํ•œ ๋ณด์•ˆ ๊ฒ€ํ† ๋ฅผ ํ•˜์ง€ ์•Š๊ฑฐ๋‚˜ ๋ฌธ์„œํ™”๋˜์ง€ ์•Š์€ ์ธํ„ฐ๋„ท ์„œ๋น„์Šค๋กœ ์ด์šฉํ•˜์ง€ ๋ง๋ผ๋Š” ๊ฒƒ์„ ์š”์ฒญํ•œ๋‹ค. ์ €ํฌ๋Š” ๋ชจ๋“  ์‚ฌ์šฉ์ž๊ฐ€ ์ด ์›์น™์„ ์ง€ํ‚ค๋ฉฐ ๊ธฐ์ˆ ์˜ ๋ฐœ์ „์ด ๊ทœ๋ฒ”์ ์ด๊ณ  ํ•ฉ๋ฒ•์ ์ธ ํ™˜๊ฒฝ์—์„œ ์ด๋ฃจ์–ด์งˆ ์ˆ˜ ์žˆ๊ธฐ๋ฅผ ๋ฐ”๋ž€๋‹ค. ์ €ํฌ๋Š” ์ด๋ฏธ ์ตœ์„ ์„ ๋‹คํ•ด ๋ชจ๋ธ ํ›ˆ๋ จ ๊ณผ์ •์—์„œ ์‚ฌ์šฉ๋œ ๋ฐ์ดํ„ฐ์˜ ์ค€์น™์„ฑ์„ ํ™•๋ณดํ•˜๋„๋ก ํ•˜์˜€๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ๋ง‰๋Œ€ํ•œ ๋…ธ๋ ฅ์„ ๊ธฐ์šธ์˜€์Œ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ  ๋ชจ๋ธ๊ณผ ๋ฐ์ดํ„ฐ์˜ ๋ณต์žก์„ฑ์œผ๋กœ ๋ง๋ฏธ์•”์•„ ์ผ๋ถ€ ์˜ˆ๊ฒฌํ•  ์ˆ˜ ์—†์„ ๋ฌธ์ œ๋“ค์ด ์—ฌ์ „ํžˆ ์กด์žฌํ•  ์ˆ˜ ์žˆ๋‹ค. ๋”ฐ๋ผ์„œ Orion-14B ์˜คํ”ˆ์†Œ์Šค ๋ชจ๋ธ์˜ ์‚ฌ์šฉ์œผ๋กœ ์•ผ๊ธฐ๋œ ๋ฌธ์ œ, ๋ฐ์ดํ„ฐ ๋ณด์•ˆ ๋ฌธ์ œ์™€ ๊ณต๋ก  ์œ„ํ—˜์ด๋‚˜ ๋ชจ๋ธ์˜ ์˜ค๋„, ๋‚จ์šฉ, ์ „ํŒŒ, ๋˜ํ•œ ๋ถˆ์ ๋‹นํ•œ ์‚ฌ์šฉ ๋“ฑ์œผ๋กœ ๊ฐ€์ ธ์˜จ ์œ„ํ—˜๊ณผ ๋ฌธ์ œ์— ๋Œ€ํ•ด ์ €ํฌ๋Š” ์ฑ…์ž„์„ ์ง€์ง€ ์•Š๊ฒ ๋‹ค. ## 5.2. ํ˜‘์˜ ์ปค๋ฎค๋‹ˆํ‹ฐ ์‚ฌ์šฉOrion-14B์‹œ๋ฆฌ์ฆˆ ๋ชจ๋ธ - ์ฝ”๋“œ๋Š” [Apache License Version 2.0](./LICENSE)
๋”ฐ๋ฅด์„ธ์š” - ๋ชจ๋ธ์€ [Orion-14B์‹œ๋ฆฌ์ฆˆ ๋ชจ๋ธ ์ปค๋ฎค๋‹ˆํ‹ฐ ํ—ˆ๊ฐ€ ํ˜‘์˜](./ModelsCommunityLicenseAgreement)๋”ฐ๋ฅด์„ธ์š”
# 6. ํšŒ์‚ฌ์†Œ๊ฐœ ์˜ค๋ฆฌ์˜จ ์Šคํƒ€๏ผˆOrionStar๏ผ‰๋Š” 2016๋…„ 9์›” ์„ค๋ฆฝ๋œ ์„ธ๊ณ„ ์ตœ๊ณ ์˜ ์„œ๋น„์Šค ๋กœ๋ด‡ ์†”๋ฃจ์…˜ ํšŒ์‚ฌ์ด๋‹ค. ์˜ค๋ฆฌ์˜จ ์Šคํƒ€๋Š” ์ธ๊ณต์ง€๋Šฅ ๊ธฐ์ˆ ์„ ๋ฐ”ํƒ•์œผ๋กœ ์ฐจ์„ธ๋Œ€ ํ˜๋ช…์  ๋กœ๋ด‡ ๋งŒ๋“ค์–ด ์‚ฌ๋žŒ๋“ค์ด ๋ฐ˜๋ณต๋˜๋Š” ์œก์ฒด๋…ธ๋™์—์„œ ๋ฒ—์–ด๋‚˜ ์ผ๊ณผ ์ƒํ™œ์„ ๋”์šฑ ์ง€๋Šฅ์ ์ด๊ณ  ์žฌ๋ฏธ์žˆ๊ฒŒ ๋งŒ๋“ค๊ณ  ๊ธฐ์ˆ ์„ ํ†ตํ•ด ์‚ฌํšŒ์™€ ์„ธ๊ณ„๋ฅผ ๋”์šฑ ์•„๋ฆ„๋‹ต๊ฒŒ ๋งŒ๋“  ๊ฒƒ์— ํž˜์„ ๊ธฐ์šธ์ธ๋‹ค. ์˜ค๋ฆฌ์˜จ ์Šคํƒ€๋Š” ์Œ์„ฑ ์ธํ„ฐ๋ ‰์…˜๊ณผ ์‹œ๊ฐ ๋„ค๋น„๊ฒŒ์ด์…˜ ๋“ฑ ์™„์ „ํžˆ ๋…์ž์ ์œผ๋กœ ๊ฐœ๋ฐœํ•œ ํ’€ ์ฒด์ธ ์ธ๊ณต์ง€๋Šฅ ๊ธฐ์ˆ ์„ ๊ฐ€์ง€๊ณ  ์žˆ๋‹ค. ์ €ํฌ๋Š” ํ”„๋กœ๋•ํŠธ ๊ฐœ๋ฐœ ๋Šฅ๋ ฅ๊ณผ ๊ธฐ์ˆ  ์‘์šฉ ๋Šฅ๋ ฅ์„ ํ†ตํ•ฉํ•˜์˜€๋‹ค. ์˜ค๋ฆฌ์˜จ ๋กœ๋ด‡ ํŒ” ํ”Œ๋žซํผ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ORIONSTAR AI Robot Greeting, AI Robot Greeting Mini, Lucki, CoffeeMaster ๋“ฑ์˜ ํ”„๋กœ๋•ํŠธ ์ถœ์‹œํ•˜์˜€์œผ๋ฉฐ ์˜ค๋ฆฌ์˜จ ๋กœ๋ด‡์˜ ์˜คํ”ˆ ํ”Œ๋žซํผ์ธ OrionOS๋ฅผ ์„ค๋ฆฝํ•˜์˜€๋‹ค. **์ง„์งœ ์œ ์šฉํ•œ ๋กœ๋ด‡์„ ์œ„ํ•ด ํƒœ์–ด๋‚˜๋ผ**์˜ ์ด๋…์„ ์œ„ํ•œ ์‹ค์ฒœํ•˜์—ฌ AI๊ธฐ์ˆ ์„ ํ†ตํ•ด ๋” ๋งŽ์€ ์‚ฌ๋žŒ๋“ค์—๊ฒŒ ๋Šฅ๋ ฅ์„ ๋ถ€์—ฌํ•œ๋‹ค. 7๋…„์˜ AI๊ฒฝํ—˜ ๋ˆ„์ ์„ ๋ฐ”ํƒ•์œผ๋กœ ์˜ค๋ฆฌ์˜จ ์Šคํƒ€๋Š” ๋Œ€ํ˜• ๋ชจ๋ธ ์‹ฌ์ธต ์‘์šฉ"์ฅ์–ธ(Chatmax)"์„ ์ถœ์‹œํ–ˆ๊ณ  ์—…๊ณ„ ๊ณ ๊ฐ์—๊ฒŒ ๋งž์ถคํ˜• AI๋Œ€ํ˜• ๋ชจ๋ธ ์ปจ์„คํŒ…๊ณผ ์„œ๋น„์Šค ์†”๋ฃจ์…˜์„ ์ง€์†์ ์œผ๋กœ ์ œ๊ณตํ•˜์—ฌ ์ง„์ •์œผ๋กœ ๊ธฐ์—… ๊ฒฝ์˜ ํšจ์œจ์ด ๋™์ข… ์—…๊ณ„์— ์•ž์„œ๋Š” ๋ชฉํ‘œ๋ฅผ ๋‹ฌ์„ฑํ•  ์ˆ˜ ์žˆ๋„๋ก ๊ณ ๊ฐ๋“ค์—๊ฒŒ ๋•๊ณ  ์žˆ๋‹ค. **์˜ค๋ฆฌ์˜จ ์Šคํƒ€๋Š” ํ’€ ์ฒด์ธ ๋Œ€ํ˜• ๋ชจ๋ธ ์‘์šฉ๋Šฅ๋ ฅ์ด๋ž€ ํ•ต์‹ฌ์  ์šฐ์„ธ๋ฅผ ๊ฐ–๊ณ  ์žˆ๋‹ค**, ๋Œ€๋Ÿ‰ ๋ฐ์ดํ„ฐ ์ฒ˜๋ฆฌ, ๋Œ€ํ˜• ๋ชจ๋ธ ์‚ฌ์ „ ํ›ˆ๋ จ, 2์ฐจ ์‚ฌ์ „ ํ›ˆ๋ จ, ๋ฏธ์„ธ ์กฐ์ •(Fine-tune), PromptEngineering, Agent๋“ฑ์—์„œ ๊ฐœ๋ฐœ๋œ ํ’€ ์ฒด์ธ ๋Šฅ๋ ฅ๊ณผ ๊ฒฝํ—˜ ๋ˆ„์ ์„ ๊ฐ€์ง€๋Š” ๊ฑฐ ํฌํ•จํ•œ๋‹ค. ์ฒด๊ณ„ํ™”๋œ ๋ฐ์ดํ„ฐ ์ฒ˜๋ฆฌ ์ ˆ์ฐจ์™€ ์ˆ˜๋ฐฑ ๊ฐœ์˜ GPU์˜ ๋ณ‘๋ ฌ ๋ชจ๋ธ ํ›ˆ๋ จ ๋Šฅ๋ ฅ์„ ํฌํ•จํ•œ ์™„์ •ํ•œ ์—”๋“œํˆฌ์—”๋“œ ๋ชจ๋ธ ํ›ˆ๋ จ ๋Šฅ๋ ฅ์„ ๊ฐ€์ง€๊ณ  ์žˆ์œผ๋ฉฐ ํ˜„์žฌ ๋Œ€ํ˜• ์ •๋ฌด, ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค, ์ถœํ•ด ์ „์ž์ƒ๊ฑฐ๋ž˜, ์พŒ์†์†Œ๋น„ํ’ˆ ๋“ฑ ์—ฌ๋Ÿฌ ์—…๊ณ„์—์„œ ๊ตฌํ˜„๋˜์—ˆ๋‹ค. ***๋Œ€ํ˜• ๋ชจ๋ธ ์‘์šฉ ๊ตฌํ˜„ ํ•„์š”๊ฐ€ ์žˆ์œผ์‹  ํšŒ์‚ฌ๊ป˜์„œ ์ €ํฌ์™€ ์—ฐ๋ฝํ•˜๋Š” ๊ฒƒ์„ ํ™˜์˜ํ•œ๋‹ค***
**๋ฌธ์˜ ์ „ํ™”:** 400-898-7779
**์ด๋ฉ”์ผ:** ai@orionstar.com
**Discord ์ปค๋ฎค๋‹ˆํ‹ฐ ๋งํฌ: https://discord.gg/zumjDWgdAs**
wechat