Clickbait1 / README.md
caush's picture
Update README.md
8d65c33
|
raw
history blame
4.1 kB
metadata
license: mit
tags:
  - generated_from_trainer
model-index:
  - name: Clickbait1
    results: []

Clickbait1

This model is a fine-tuned version of microsoft/Multilingual-MiniLM-L12-H384 on the Webis-Clickbait-17 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0260

Model description

MiniLM is a distilled model from the paper "MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers".

We fine tune this model to evaluate (regression) the clickbait level of title news.

Intended uses & limitations

Model was designed to work on Transformers (like in the paper "Predicting Clickbait Strength in Online Social Media" by Indurthi Vijayasaradhi, Syed Bakhtiyar, Gupta Manish, Varma Vasudeva).

The model wa trained with english titles.

Training and evaluation data

We train the model with the official training data of the chalenge, plus another set that was available after the end of the challenge.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-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
  • num_epochs: 4

Training results

Training Loss Epoch Step Validation Loss
No log 0.05 50 0.0390
No log 0.09 100 0.0365
No log 0.14 150 0.0321
No log 0.18 200 0.0319
No log 0.23 250 0.0350
No log 0.27 300 0.0317
No log 0.32 350 0.0304
No log 0.36 400 0.0283
No log 0.41 450 0.0300
0.0364 0.46 500 0.0282
0.0364 0.5 550 0.0285
0.0364 0.55 600 0.0276
0.0364 0.59 650 0.0278
0.0364 0.64 700 0.0293
0.0364 0.68 750 0.0280
0.0364 0.73 800 0.0320
0.0364 0.77 850 0.0269
0.0364 0.82 900 0.0269
0.0364 0.87 950 0.0271
0.0281 0.91 1000 0.0314
0.0281 0.96 1050 0.0265
0.0281 1.0 1100 0.0295
0.0281 1.05 1150 0.0295
0.0281 1.09 1200 0.0290
0.0281 1.14 1250 0.0281
0.0281 1.18 1300 0.0272
0.0281 1.23 1350 0.0273
0.0281 1.28 1400 0.0287
0.0281 1.32 1450 0.0267
0.026 1.37 1500 0.0284
0.026 1.41 1550 0.0264
0.026 1.46 1600 0.0273
0.026 1.5 1650 0.0280
0.026 1.55 1700 0.0266
0.026 1.59 1750 0.0260
0.026 1.64 1800 0.0266
0.026 1.68 1850 0.0268
0.026 1.73 1900 0.0269
0.026 1.78 1950 0.0260
0.0236 1.82 2000 0.0273
0.0236 1.87 2050 0.0272
0.0236 1.91 2100 0.0260
0.0236 1.96 2150 0.0269
0.0236 2.0 2200 0.0286
0.0236 2.05 2250 0.0266

Framework versions

  • Transformers 4.18.0
  • Pytorch 1.11.0a0+17540c5
  • Datasets 2.1.0
  • Tokenizers 0.12.1