SEOKDONG commited on
Commit
a8b8d24
ยท
verified ยท
1 Parent(s): 092ef36

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -3
README.md CHANGED
@@ -18,7 +18,7 @@ pipeline_tag: text-generation
18
  ### ๊ฐœ๋ฐœ ๋˜์—ˆ์œผ๋ฉฐ ์ž์ฒด ์ œ์ž‘ํ•œ 53๊ฐœ ์˜์—ญ์˜ ํ•œ๊ตญ์–ด ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ•˜์—ฌ ํ•œ๊ตญ ์‚ฌํšŒ ๊ฐ€์น˜์™€
19
  ### ๋ฌธํ™”๋ฅผ ์ดํ•ดํ•˜๋Š” ๋ชจ๋ธ ์ž…๋‹ˆ๋‹ค. Thanks for ktdsโœŒ
20
 
21
- ### V1.6 Helpfulness ๊ฐ๊ด€์‹ ๋ฌธ์ œ ์ถ”๊ฐ€: 120K, Epoch=1
22
 
23
 
24
  # โถ ๋ชจ๋ธ ์„ค๋ช…
@@ -69,7 +69,10 @@ tokenizer = AutoTokenizer.from_pretrained(base_LLM_model)
69
  from tqdm import tqdm
70
  for i in tqdm(range(0,1)): #len(answer_list))):
71
  input_text = """
72
- ๋‹น์‹ ์€ AI ๋น„์„œ์ž…๋‹ˆ๋‹ค. ๋‹ค์Œ ์งˆ๋ฌธ์— ๋งž๋Š” ๋‹ต๋ณ€์„ ๊ณ ๋ฅด์„ธ์š”. ๋‹ต๋ณ€์€ 1,2,3,4 ์ค‘์— ํ•˜๋‚˜๋งŒ ์„ ํƒํ•˜์„ธ์š”. ๋‹ค์Œ ์ค‘ ํƒ„์ˆ˜ํ™”๋ฌผ์˜ ์ผ์ข…์œผ๋กœ, ํฌ๋„๋‹น์˜ ์ด์ค‘ ๊ฒฐํ•ฉ์œผ๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ์œผ๋ฉฐ ์ž์ฃผ ์‚ฌ์šฉ๋˜๋Š” ์„คํƒ•์˜ ์„ฑ๋ถ„์€ ๋ฌด์—‡์ธ๊ฐ€? ์„ ํƒ์ง€: 1. ์…€๋ฃฐ๋กœ์˜ค์Šค 2. ์ž๋‹น 3. ๋…น๋ง 4. ๊ธ€๋ฆฌ์ฝ”๊ฒ ๋‹ต๋ณ€:"""
 
 
 
73
  inputs = tokenizer(input_text, return_tensors="pt")
74
  inputs = inputs.to("cuda:0")
75
  # 3. ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•œ ์ถ”๋ก 
@@ -79,6 +82,7 @@ for i in tqdm(range(0,1)): #len(answer_list))):
79
  result = tokenizer.decode(outputs[0], skip_special_tokens=True)
80
 
81
  print(result.split("๋‹ต๋ณ€:")[1].strip())
82
- ๊ฒฐ๊ณผ: '2. ์ž๋‹น์ž…๋‹ˆ๋‹ค. ์ž๋‹น์€ ํฌ๋„๋‹น์ด ๋‘ ๊ฐœ์˜ ์ด์ค‘ ๊ฒฐํ•ฉ์œผ๋กœ ์ด๋ฃจ์–ด์ง„ ํ˜•ํƒœ๋กœ ์กด์žฌํ•˜๋ฉฐ, ์ฃผ๋กœ ์„คํƒ•์˜ ์ฃผ์š” ์„ฑ๋ถ„์œผ๋กœ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.
 
83
  ์…€๋ฃฐ๋กœ์˜ค์Šค๋Š” ๋‹จ๋ฐฑ์งˆ, ๋…น๋ง์€ ์‹๋ฌผ์„ฑ ๋‹จ๋ฐฑ์งˆ, ๊ธ€๋ฆฌ์ฝ”๊ฒ์€ ์ง€๋ฐฉ๊ณผ ๊ด€๋ จ๋œ ๋‹จ๋ฐฑ์งˆ๋กœ, ๋ชจ๋‘ ์„คํƒ•์˜ ์„ฑ๋ถ„์ด ์•„๋‹™๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ •๋‹ต์€ 2์ž…๋‹ˆ๋‹ค.'
84
  </code></pre>
 
18
  ### ๊ฐœ๋ฐœ ๋˜์—ˆ์œผ๋ฉฐ ์ž์ฒด ์ œ์ž‘ํ•œ 53๊ฐœ ์˜์—ญ์˜ ํ•œ๊ตญ์–ด ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ•˜์—ฌ ํ•œ๊ตญ ์‚ฌํšŒ ๊ฐ€์น˜์™€
19
  ### ๋ฌธํ™”๋ฅผ ์ดํ•ดํ•˜๋Š” ๋ชจ๋ธ ์ž…๋‹ˆ๋‹ค. Thanks for ktdsโœŒ
20
 
21
+ ### V0.2 Epoch=2
22
 
23
 
24
  # โถ ๋ชจ๋ธ ์„ค๋ช…
 
69
  from tqdm import tqdm
70
  for i in tqdm(range(0,1)): #len(answer_list))):
71
  input_text = """
72
+ ๋‹น์‹ ์€ AI ๋น„์„œ์ž…๋‹ˆ๋‹ค. ๋‹ค์Œ ์งˆ๋ฌธ์— ๋งž๋Š” ๋‹ต๋ณ€์„ ๊ณ ๋ฅด์„ธ์š”. ๋‹ต๋ณ€์€ 1,2,3,4 ์ค‘์— ํ•˜๋‚˜๋งŒ ์„ ํƒํ•˜์„ธ์š”.
73
+ ๋‹ค์Œ ์ค‘ ํƒ„์ˆ˜ํ™”๋ฌผ์˜ ์ผ์ข…์œผ๋กœ, ํฌ๋„๋‹น์˜ ์ด์ค‘ ๊ฒฐํ•ฉ์œผ๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ์œผ๋ฉฐ ์ž์ฃผ ์‚ฌ์šฉ๋˜๋Š” ์„คํƒ•์˜ ์„ฑ๋ถ„์€ ๋ฌด์—‡์ธ๊ฐ€?
74
+ ์„ ํƒ์ง€: 1. ์…€๋ฃฐ๋กœ์˜ค์Šค 2. ์ž๋‹น 3. ๋…น๋ง 4. ๊ธ€๋ฆฌ์ฝ”๊ฒ ๋‹ต๋ณ€:"""
75
+
76
  inputs = tokenizer(input_text, return_tensors="pt")
77
  inputs = inputs.to("cuda:0")
78
  # 3. ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•œ ์ถ”๋ก 
 
82
  result = tokenizer.decode(outputs[0], skip_special_tokens=True)
83
 
84
  print(result.split("๋‹ต๋ณ€:")[1].strip())
85
+
86
+ ๊ฒฐ๊ณผ: '2. ์ž๋‹น์ž…๋‹ˆ๋‹ค. ์ž๋‹น์€ ํฌ๋„๋‹น์ด ๋‘ ๊ฐœ์˜ ์ด์ค‘ ๊ฒฐํ•ฉ์œผ๋กœ ์ด๋ฃจ์–ด์ง„ ํ˜•ํƒœ๋กœ ์กด์žฌํ•˜๋ฉฐ, ์ฃผ๋กœ ์„คํƒ•์˜ ์ฃผ์š” ์„ฑ๋ถ„์œผ๋กœ ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค.
87
  ์…€๋ฃฐ๋กœ์˜ค์Šค๋Š” ๋‹จ๋ฐฑ์งˆ, ๋…น๋ง์€ ์‹๋ฌผ์„ฑ ๋‹จ๋ฐฑ์งˆ, ๊ธ€๋ฆฌ์ฝ”๊ฒ์€ ์ง€๋ฐฉ๊ณผ ๊ด€๋ จ๋œ ๋‹จ๋ฐฑ์งˆ๋กœ, ๋ชจ๋‘ ์„คํƒ•์˜ ์„ฑ๋ถ„์ด ์•„๋‹™๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ •๋‹ต์€ 2์ž…๋‹ˆ๋‹ค.'
88
  </code></pre>