Edit model card

Accuracy

TP TN
FP 492 62
FN 39 467

์ •ํ™•๋„: 89.9%(492+267/1000)

์ •ํ™•๋„๊ฐ€ ๊ฑฐ์˜ 90%์— ๊ทผ์ ‘ํ•ฉ๋‹ˆ๋‹ค.

-Train ๋ฐ์ดํ„ฐ๋Š” nsmc['train']์„ shuffleํ•œ ๋‹ค์Œ์— ์ƒ์œ„ 2000๊ฐœ๋ฅผ ํ™œ์šฉํ•˜์˜€์Šต๋‹ˆ๋‹ค

-Test ๋ฐ์ดํ„ฐ๋Š” nsmc['train']์„ shuffleํ•œ ๋‹ค์Œ์— ์ƒ์œ„ 1000๊ฐœ๋ฅผ ํ™œ์šฉํ•˜์˜€์Šต๋‹ˆ๋‹ค.

-1000๊ฐœ์˜ train ๋ฐ์ดํ„ฐ ์ค‘ 890๊ฐœ๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ๋ถ„๋ฅ˜ํ•˜์˜€์Šต๋‹ˆ๋‹ค.

-Base Model์€ KT-AI/midm-bitext-S-7B-inst-v1๋กœ base model์— Lora ์–ด๋Œ‘ํ„ฐ๋ฅผ ๋ถ™์—ฌ์„œ SFTtrainer๋ฅผ ํ†ตํ•˜์—ฌ ์ƒˆ๋กœ์šด ๋ฐ์ดํ„ฐ์…‹ nsmc์— ๋Œ€ํ•ด์„œ ๋ฏธ์„ธํŠœ๋‹์„ ํ•˜์˜€์Šต๋‹ˆ๋‹ค.

-๋ฏธ์„ธํŠœ๋‹ํ•œ ๋กœ๋ผ ์–ด๋Œ‘ํ„ฐ๋ฅผ ํ—ˆ๊น…ํŽ˜์ด์Šค์— ์—…๋กœ๋“œํ•œ ํ›„ 4๋น„ํŠธ๋กœ ์–‘์žํ™”๋œ base model์— ๋ถ™์ธ ํ›„ ์ถ”๋ก ์„ ํ•˜์˜€์Šต๋‹ˆ๋‹ค.

-์ƒ์œ„ 2000๊ฐœ์˜ train_dataset์— ๋Œ€ํ•ด์„œ ํ›ˆ๋ จ์„ ํ•œ ํ›„ ์ƒ์œ„ 1000๊ฐœ์˜ test_dataset์— ๋Œ€ํ•ด์„œ ์ถ”๋ก ์€ ํ•œ ๊ฒฐ๊ณผ๋Š” ์œ„์˜ ํ‘œ์™€ ๊ฐ™์Šต๋‹ˆ๋‹ค.

-์ •ํ™•๋„ ํ–ฅ์ƒ์„ ์œ„ํ•˜์—ฌ test ๋ฐ์ดํ„ฐ์…‹์˜ ํฌ๊ธฐ๋ฅผ ๋Š˜๋ฆฌ๋ ค ํ•˜์˜€์œผ๋‚˜, GPU ์šฉ๋Ÿ‰ ์ œํ•œ์œผ๋กœ ์ธํ•˜์—ฌ ์ง€์†์ ์œผ๋กœ ์˜ค๋ฅ˜๊ฐ€ ๋ฐœ์ƒํ•˜์˜€์Šต๋‹ˆ๋‹ค.

-์ •ํ™•๋„๋ฅผ ์˜ฌ๋ฆฌ๋Š” ๋ฐฉ๋ฒ•: seq_length๋ฅผ ํ‚ค์šด๋‹ค. 512 ์ •๋„๋กœ

Downloads last month
3
Inference API
Unable to determine this modelโ€™s pipeline type. Check the docs .

Model tree for kayla0913/hw-midm-7B-nsmc

Adapter
(16)
this model