results

This model is a fine-tuned version of kingkim/kodialogpt_v1.1_SecurityManual on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.9083

Model description

ν•΄λ‹Ή λͺ¨λΈμ€ λΉŒλ”© λ³΄μ•ˆ 맀뉴얼을 ν•™μŠ΅ν•œ νŒŒμΈνŠœλ‹λœ λͺ¨λΈλ‘œ, λ³΄μ•ˆ μƒν™©μ—μ„œμ˜ λŒ€μ²˜ 방법을 질문과 λ‹΅λ³€ ν˜•μ‹μœΌλ‘œ ν•™μŠ΅ν–ˆμŠ΅λ‹ˆλ‹€. μ €μΈ΅λΆ€ 및 κ³ μΈ΅λΆ€ ν™”μž¬ λŒ€μ‘, μΉ¨μž…μž λ°œμƒ μ‹œ λŒ€μ‘ 방법 λ“± λ‹€μ–‘ν•œ λ³΄μ•ˆ μƒν™©μ—μ„œμ˜ 맀뉴얼을 λ°”νƒ•μœΌλ‘œ ν›ˆλ ¨λ˜μ—ˆμŠ΅λ‹ˆλ‹€.


Intended uses & limitations

μš©λ„:

  • λ³΄μ•ˆ μš”μ› ν›ˆλ ¨ μ‹œμŠ€ν…œ
  • λΉŒλ”© λ³΄μ•ˆ κ΄€λ ¨ 응닡 μ‹œμŠ€ν…œ
  • 맀뉴얼 기반 챗봇 μ‘μš© ν”„λ‘œκ·Έλž¨

μ œν•œ 사항:

  • λͺ¨λΈμ€ λΉŒλ”© λ³΄μ•ˆ 맀뉴얼 λ°μ΄ν„°λ§Œμ„ 기반으둜 ν•™μŠ΅λ˜μ—ˆμœΌλ©°, λ‹€λ₯Έ λ³΄μ•ˆ μ‹œλ‚˜λ¦¬μ˜€μ— λŒ€ν•œ μΌλ°˜ν™” λŠ₯λ ₯은 μ œν•œμ μΌ 수 μžˆμŠ΅λ‹ˆλ‹€.

Training and evaluation data

λ³Έ λͺ¨λΈμ€ λΉŒλ”© λ³΄μ•ˆ 맀뉴얼을 λ°”νƒ•μœΌλ‘œ λ§Œλ“€μ–΄μ§„ 데이터셋을 기반으둜 ν•™μŠ΅λ˜μ—ˆμŠ΅λ‹ˆλ‹€. kingkim/DS_Building_SecurityManual kingkim/DS_Building_SecurityManual_V3 kingkim/DS_Building_SecurityManual_V5 데이터셋은 μ‹€μ œ 맀뉴얼을 λ°”νƒ•μœΌλ‘œ μž‘μ„±λ˜μ—ˆμœΌλ©°, 200개 μ΄μƒμ˜ λ³΄μ•ˆ μ‹œλ‚˜λ¦¬μ˜€λ₯Ό ν¬ν•¨ν•˜κ³  μžˆμŠ΅λ‹ˆλ‹€.


Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0002
  • train_batch_size: 2
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss
2.2114 0.9565 11 1.5838
0.8439 2.0 23 1.2342
0.6033 2.9565 34 0.9828
0.3294 4.0 46 0.9062
0.2423 4.7826 55 0.9083

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.4.1+cu121
  • Datasets 3.0.1
  • Tokenizers 0.19.1
Downloads last month
127
Safetensors
Model size
125M params
Tensor type
F32
Β·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for kingkim/kodialogpt_v3.0_SecurityManual