dadashzadeh's picture
Update README.md
5405081
|
raw
history blame
No virus
1.73 kB
metadata
license: mit
base_model: xlm-roberta-base
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: test-trainer
    results: []
language:
  - en

test-trainer

This model is a fine-tuned version of xlm-roberta-base on the cryptocurrency dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2337
  • Accuracy: 0.9169

Model description

intent search detection :

Navigational: Users want to find a specific page (e.g., “reddit login”) Informational: Users want to learn more about something (e.g., “what is seo”) Commercial: Users want to do research before making a purchase decision (e.g., “best coffee maker”) Transactional: Users want to complete a specific action, usually a purchase (e.g., “buy subaru forester”)

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-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: 3

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.3629 1.0 14391 0.3249 0.8866
0.313 2.0 28782 0.2640 0.9067
0.2723 3.0 43173 0.2337 0.9169

Framework versions

  • Transformers 4.34.1
  • Pytorch 2.1.0+cu118
  • Datasets 2.14.6
  • Tokenizers 0.14.1