---
library_name: transformers
base_model: rasyosef/bert-medium-amharic
tags:
- generated_from_trainer
metrics:
- f1
model-index:
- name: STL-rasyosef-bert-medium-amharic-sentiment
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# STL-rasyosef-bert-medium-amharic-sentiment

This model is a fine-tuned version of [rasyosef/bert-medium-amharic](https://huggingface.co/rasyosef/bert-medium-amharic) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.7579
- Sentiment Precision: 0.6921
- Sentiment Recall: 0.6954
- F1: 0.6918
- Sentiment Acc: 0.6995

## 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 |
|:-------------:|:-----:|:----:|:---------------:|:-------------------:|:----------------:|:------:|:-------------:|
| 0.8272        | 1.0   | 377  | 0.7678          | 0.6527              | 0.6522           | 0.6523 | 0.6571        |
| 0.6971        | 2.0   | 754  | 0.7301          | 0.6812              | 0.6759           | 0.6668 | 0.6783        |
| 0.6232        | 3.0   | 1131 | 0.7282          | 0.6891              | 0.6876           | 0.6782 | 0.6908        |
| 0.5508        | 4.0   | 1508 | 0.7142          | 0.6995              | 0.6990           | 0.6981 | 0.7007        |
| 0.5353        | 5.0   | 1885 | 0.7224          | 0.6848              | 0.6885           | 0.6844 | 0.6933        |
| 0.4981        | 6.0   | 2262 | 0.7598          | 0.6897              | 0.6913           | 0.6860 | 0.6933        |
| 0.4942        | 7.0   | 2639 | 0.7579          | 0.6921              | 0.6954           | 0.6918 | 0.6995        |


### Framework versions

- Transformers 5.16.1
- Pytorch 2.11.0+cu128
- Datasets 4.0.0
- Tokenizers 0.23.1