qa_kor_math / README.md
idah4's picture
Update README.md
650d354 verified
|
raw
history blame
2.45 kB
metadata
license: mit
base_model: gogamza/kobart-base-v2
tags:
  - generated_from_trainer
model-index:
  - name: qa_kor_math
    results: []

qa_kor_math

This model is a fine-tuned version of gogamza/kobart-base-v2 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2586

Model description

ํ•œ๊ตญ์–ด ์ˆ˜ํ•™ ๋ฌธ์ œ๋ฅผ ์ž…๋ ฅํ•˜๋ฉด, ๋ฌธ์ œ ์œ ํ˜•๊ณผ ๋ฌธ์ œ ์œ ํ˜•์— ๋Œ€ํ•œ ์„ค๋ช…, ํ’€์ด(์ฝ”๋“œ), ์ •๋‹ต์ด ์ถœ๋ ฅ๋˜๋„๋ก fine tuning ํ–ˆ์Šต๋‹ˆ๋‹ค.
๋ฌธ์ œ ์œ ํ˜• ์ข…๋ฅ˜๋กœ๋Š” ์‚ฐ์ˆ ์—ฐ์‚ฐ, ์ˆœ์„œ์ •ํ•˜๊ธฐ, ์กฐํ•ฉํ•˜๊ธฐ, ์ˆ˜ ์ฐพ๊ธฐ, ํฌ๊ธฐ ๋น„๊ต, ๋„ํ˜•์ด ์žˆ์Šต๋‹ˆ๋‹ค.
๋ชจ๋ธ์ด ๊ฐ€๋ฒผ์šด ํƒ“์ธ์ง€ ์ •ํ™•๋„๊ฐ€ ๋†’์ง€๋Š” ์•Š์•„ ๋ณด์ž…๋‹ˆ๋‹ค.

Intended uses & limitations

Training and evaluation data

tunib-ai์—์„œ git์— ๊ณต๊ฐœํ•œ train ๋ฐ์ดํ„ฐ ์…‹์œผ๋กœ ํ•™์Šตํ•˜์˜€์Šต๋‹ˆ๋‹ค.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 1e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 400
  • num_epochs: 10

Training results

Training Loss Epoch Step Validation Loss
No log 0.63 100 2.9907
No log 1.26 200 0.9196
No log 1.89 300 0.5858
No log 2.52 400 0.4351
2.4889 3.14 500 0.3693
2.4889 3.77 600 0.3356
2.4889 4.4 700 0.3182
2.4889 5.03 800 0.3017
2.4889 5.66 900 0.2949
0.3483 6.29 1000 0.2798
0.3483 6.92 1100 0.2748
0.3483 7.55 1200 0.2695
0.3483 8.18 1300 0.2649
0.3483 8.81 1400 0.2610
0.2753 9.43 1500 0.2586

Framework versions

  • Transformers 4.38.2
  • Pytorch 2.2.1+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.2