DunnBC22's picture
Update README.md
6f3d747
---
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- accuracy
- f1
- recall
- precision
model-index:
- name: albert-base-v2-Tweet_About_Disaster_Or_Not
results: []
language:
- en
---
# albert-base-v2-Tweet_About_Disaster_Or_Not
This model is a fine-tuned version of [albert-base-v2](https://huggingface.co/albert-base-v2) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2899
- Accuracy: 0.8989
- F1: 0.7784
- Recall: 0.8523
- Precision: 0.7163
## Model description
This is a binary classification model to determine if tweet input samples are about a disaster or not.
For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/Binary%20Classification/Transformer%20Comparison/Is%20This%20Tweet%20Referring%20to%20a%20Disaster%20or%20Not%3F%20-%20ALBERT.ipynb
### Associated Projects
This project is part of a comparison of multiple transformers. The others can be found at the following links:
- https://huggingface.co/DunnBC22/roberta-base-Tweet_About_Disaster_Or_Not
- https://huggingface.co/DunnBC22/deberta-v3-small-Tweet_About_Disaster_Or_Not
- https://huggingface.co/DunnBC22/electra-base-emotion-Tweet_About_Disaster_Or_Not
- https://huggingface.co/DunnBC22/ernie-2.0-base-en-Tweet_About_Disaster_Or_Not
- https://huggingface.co/DunnBC22/distilbert-base-uncased-Tweet_About_Disaster_Or_Not
## Intended uses & limitations
This model is intended to demonstrate my ability to solve a complex problem using technology.
The main limitation is the quality of the data source.
## Training and evaluation data
Dataset Source: https://www.kaggle.com/datasets/vstepanenko/disaster-tweets
_Input Word Length By Class:_
![Length of Input Text (in Words) By Class](https://github.com/DunnBC22/NLP_Projects/raw/main/Binary%20Classification/Transformer%20Comparison/Images/Tweet%20Word%20Lengths%20By%20Class.png)
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- 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 | Accuracy | F1 | Recall | Precision |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:------:|:---------:|
| 0.3598 | 1.0 | 143 | 0.3025 | 0.8795 | 0.7495 | 0.8650 | 0.6613 |
| 0.234 | 2.0 | 286 | 0.2899 | 0.8989 | 0.7784 | 0.8523 | 0.7163 |
| 0.1557 | 3.0 | 429 | 0.3424 | 0.9156 | 0.7904 | 0.7637 | 0.8190 |
| 0.0871 | 4.0 | 572 | 0.4189 | 0.9182 | 0.7901 | 0.7384 | 0.8495 |
| 0.0517 | 5.0 | 715 | 0.4396 | 0.9200 | 0.8043 | 0.7890 | 0.8202 |
### Framework versions
- Transformers 4.26.1
- Pytorch 1.13.1
- Datasets 2.9.0
- Tokenizers 0.12.1