metadata
library_name: transformers
license: mit
base_model: EthioNLP/EthioLLM-b-250K
tags:
- generated_from_trainer
metrics:
- f1
model-index:
- name: STL-ethiollm-b-250K-sentiment
results: []
STL-ethiollm-b-250K-sentiment
This model is a fine-tuned version of EthioNLP/EthioLLM-b-250K on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.9272
- Sentiment Precision: 0.6650
- Sentiment Recall: 0.6514
- F1: 0.6451
- Sentiment Acc: 0.6484
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: 1e-05
- train_batch_size: 16
- eval_batch_size: 32
- seed: 42
- 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: cosine
- lr_scheduler_warmup_steps: 300
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Sentiment Precision | Sentiment Recall | F1 | Sentiment Acc |
|---|---|---|---|---|---|---|---|
| 0.9908 | 1.0 | 377 | 0.8728 | 0.6072 | 0.6070 | 0.6071 | 0.6110 |
| 0.8654 | 2.0 | 754 | 0.8423 | 0.6625 | 0.6281 | 0.6123 | 0.6222 |
| 0.7344 | 3.0 | 1131 | 0.7865 | 0.6662 | 0.6595 | 0.6568 | 0.6584 |
| 0.6174 | 4.0 | 1508 | 0.7960 | 0.6592 | 0.6565 | 0.6532 | 0.6571 |
| 0.5512 | 5.0 | 1885 | 0.8839 | 0.6608 | 0.6438 | 0.6313 | 0.6434 |
| 0.4653 | 6.0 | 2262 | 0.9272 | 0.6650 | 0.6514 | 0.6451 | 0.6484 |
Framework versions
- Transformers 5.16.1
- Pytorch 2.11.0+cu128
- Datasets 4.0.0
- Tokenizers 0.23.1