kin-sentiC

This model is a fine-tuned version of RogerB/afro-xlmr-large-finetuned-kintweetsD on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8401
• F1: 0.7066

Model description

The model was trained and evaluated on a Kinyarwanda sentiment analysis dataset of tweets created by Muhammad et al. It classifies Kinyarwanda sentences into three categories: positive (0), neutral (1), and negative (2).

Intended uses & limitations

The model is specifically designed for classifying Kinyarwanda sentences, with a focus on Kinyarwanda tweets.

Training and evaluation data

The training data used for training the model were a combination of the train set from Muhammad et al and the val set from Muhammad et al , which served as the validation data during the training process. For evaluating the model's performance, the test data used were sourced from the test set from Muhammad et al

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 4
• eval_batch_size: 4
• seed: 100000
• 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	F1
0.913	1.0	1013	0.6933	0.7054
0.737	2.0	2026	0.5614	0.7854
0.646	3.0	3039	0.5357	0.8039

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3