README.md

---
license: apache-2.0
language:
- en
- hi
- de
task_categories:
- text-generation
tags:
- sft
- instruction-tuning
- multilingual
- qwen
- llama-factory
size_categories:
- 10K<n<100K
---

# Aya Multilingual SFT Dataset

This dataset is derived from the [CohereForAI/aya_collection](https://huggingface.co/datasets/CohereForAI/aya_collection) 
(translated_dolly subset) and formatted for supervised fine-tuning (SFT) with LLaMA-Factory.

## Dataset Subsets

| Subset | Languages | Train | Test | Description |
|--------|-----------|-------|------|-------------|
| `eng` | ENG | ✓ | ✓ | English only subset |
| `hin` | HIN | ✓ | ✓ | Hindi only subset |
| `deu` | DEU | ✓ | ✓ | German only subset |
| `eng_hin_deu` | ENG, HIN, DEU | ✓ | ✓ | Combined English, Hindi, and German subset |
| `eng_hin_deu_sampled` | ENG, HIN, DEU, SAMPLED | ✓ | ✓ | Combined English, Hindi, and German sampled subset |
| `ben` | BEN | ✓ | ✓ | Bengali only subset |
| `ben_eng_hin_deu` | BEN, ENG, HIN, DEU | ✓ | ✓ | Combined Bengali, English, Hindi, and German subset |
| `ben_eng_hin_deu_sampled` | BEN, ENG, HIN, DEU, SAMPLED | ✓ | ✓ | Combined Bengali English, Hindi, and German sampled subset |

## Usage

### Load a specific subset

```python
from datasets import load_dataset

# Load English subset
dataset = load_dataset("Mayank6255/translated_dolly_spec_decode", "eng")

# Load combined subset
dataset = load_dataset("Mayank6255/translated_dolly_spec_decode", "eng_hin_deu")

# Access splits
train_data = dataset["train"]
test_data = dataset["test"]
```

### Data Format

Each example contains:

```json
{
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "..."},
    {"role": "assistant", "content": "..."}
  ],
  "language": "eng"
}
```

## LLaMA-Factory Integration

Add to `data/dataset_info.json`:

```json
{
  "aya_eng_hin_deu_train": {
    "hf_hub_url": "Mayank6255/translated_dolly_spec_decode",
    "subset": "eng_hin_deu",
    "split": "train",
    "formatting": "sharegpt",
    "columns": {"messages": "messages"},
    "tags": {
      "role_tag": "role",
      "content_tag": "content",
      "user_tag": "user", 
      "assistant_tag": "assistant",
      "system_tag": "system"
    }
  }
}
```

## License

This dataset follows the licensing terms of the original Aya Collection dataset.