kinyarwanda-sentiment-afroxlmr-large

This model is a fine-tuned version of Davlan/afro-xlmr-large-76L on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.5923
• F1: 0.6921
• Precision: 0.6974
• Recall: 0.6959
• Accuracy: 0.6959

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Precision	Recall	Accuracy
1.6432	1.0	104	0.7593	0.6536	0.6592	0.6578	0.6578
1.4880	2.0	208	0.8054	0.6833	0.7140	0.6880	0.6880
1.0657	3.0	312	0.8070	0.6929	0.7027	0.6989	0.6989
0.8187	4.0	416	0.8961	0.6813	0.6857	0.6844	0.6844
0.5049	5.0	520	1.1145	0.6847	0.6936	0.6868	0.6868
0.4957	6.0	624	1.2119	0.6683	0.6889	0.6711	0.6711
0.2885	7.0	728	1.3918	0.6708	0.6852	0.6723	0.6723
0.2039	8.0	832	1.4049	0.6958	0.6966	0.6953	0.6953
0.2251	9.0	936	1.7742	0.6768	0.6852	0.6832	0.6832
0.2390	10.0	1040	1.7947	0.6971	0.7043	0.6965	0.6965
0.1849	11.0	1144	2.2932	0.6795	0.6851	0.6796	0.6796
0.1306	12.0	1248	2.3704	0.6911	0.6944	0.6917	0.6917
0.1611	13.0	1352	2.2066	0.6985	0.7060	0.6977	0.6977
0.0464	14.0	1456	2.3864	0.6887	0.6890	0.6892	0.6892
0.0953	15.0	1560	2.5195	0.6822	0.6839	0.6820	0.6820
0.0013	16.0	1664	2.4388	0.6928	0.6944	0.6929	0.6929
0.0797	17.0	1768	2.4965	0.6997	0.7021	0.7025	0.7025
0.0576	18.0	1872	2.5354	0.6905	0.6945	0.6904	0.6904
0.0490	19.0	1976	2.6234	0.6905	0.6955	0.6904	0.6904
0.0445	20.0	2080	2.5570	0.6905	0.6942	0.6904	0.6904

Framework versions

• Transformers 5.5.4
• Pytorch 2.11.0+cu130
• Datasets 2.21.0
• Tokenizers 0.22.2