README.md

metadata

license: cc-by-nc-4.0
base_model: KT-AI/midm-bitext-S-7B-inst-v1
tags:
  - generated_from_trainer
model-index:
  - name: dataset_infos_midm
    results: []

dataset_infos_midm

This model is a fine-tuned version of KT-AI/midm-bitext-S-7B-inst-v1 on an unknown dataset.

Model description

Midm은 KT가 개발한 사전학습 한국어-영어 언어모델 입니다. 문자열을 입력으로 하며, 문자열을 생성합니다. 해당 모델(KT-AI/midm-bitext-S-7B-inst-v1)을 베이스 모델로 하여 미세튜닝을 진행하였습니다.

Midm is a pre-trained Korean-English language model developed by KT. It takes text as input and creates text. We fine-tuned the model based on KT-AI/midm-bitext-S-7B-inst-v1.

Intended uses & limitations

nsmc 데이터셋의 사용자가 입력한 리뷰 문장을 분류하는 에이전트이다. 사용자 리뷰 문장으로부터 '긍정' 또는 '부정'을 판단합니다.

This is an agent that classifies user-input review sentences from NSMC dataset. It determines whether the user review sentences are 'positive' or 'negative'.

Training and test data

Training 및 test 데이터는 nsmc 데이터 셋에서 로딩해 사용합니다. (elvaluation 데이터는 사용하지 않습니다.)

We load and use training and test data from the NSMC dataset. (We do not use an evaluation data.)

Training procedure

사용자의 영화 리뷰 문장을 입력으로 받아 문장을 '긍정(1)' 또는 '부정(0)'으로 분류합니다.

Accepts movie review sentences from the user as input and classifies the sentences as 'Positive (1)' or 'Negative (0)'.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• training_steps: 300
• mixed_precision_training: Native AMP

Training results

• The following are the results considering incorrectly generated words(e.g., 정, ' ').

  • Binary Confusion Matrix

    	TP	TN
    PP	443	49
    PN	57	451

  • Accuracy: 0.894

• The following are the results without considering incorrectly generated words as wrong(e.g., 정, ' ').

  • Binary Confusion Matrix

    	TP	TN
    PP	443	38
    PN	44	451

  • Accuracy: 0.916

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0