---
library_name: transformers
tags:
- generated_from_trainer
model-index:
- name: ModernBERT-ar-base-small4
  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. -->

# ModernBERT-ar-base-tiny

This model was trained on [Fineweb2 Ar sample](https://huggingface.co/datasets/akhooli/fineweb2_ar_24_sample) dataset. 
The tokenizer was also trained using the same dataset.  
See [sample code](https://colab.research.google.com/drive/1CUsUsJQV4ZzJar2987yAzCTn8ve4hR5b?usp=sharing) 
(usage and training) and [initial post](https://www.linkedin.com/posts/akhooli_a-micro-arabic-modern-bert-a-couple-weeks-activity-7282005813357875202-SAGk)   
Updated: Jan. 12, 2025.

## Model description

ModernBERT Arabic (MLM) experiment.

## Intended uses & limitations

Educational and explorational uses only. Limited data, not fully trained. 

## Training and evaluation data

Evaluation on 5% of the data, uses 2 GPUs.

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 12
- eval_batch_size: 12
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 24
- total_eval_batch_size: 24
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 50000
- mixed_precision_training: Native AMP


### Framework versions

- Transformers 4.49.0.dev0
- Pytorch 2.5.1+cu121
- Datasets 3.2.0
- Tokenizers 0.21.0