SriRamanaAtmic/AtmicContrastive

---
language:
- en
- hi
- ta
- te
- bn
- gu
license: cc-by-4.0
task_categories:
- sentence-similarity
tags:
- advaita-vedanta
- ramana-maharshi
- multilingual
- embeddings
- spiritual-texts
pretty_name: AtmicContrastive
---

# AtmicContrastive

Contrastive learning dataset with hard negatives for Atmic embedding model

Trained model: [SriRamanaAtmic/AtmicEmbeddingv1](https://huggingface.co/SriRamanaAtmic/AtmicEmbeddingv1)

## Splits

| Split | Records |
|-------|---------|
| train | 15,006 |
| validation | 303 |
| test | 200 |
| hard_negatives | 1,308 |

## Schema

All columns are `string` type. Columns not applicable to a split are filled with `"Unknown"`.
Columns: `anchor`, `collection`, `lang`, `negative`, `positive`, `source`, `synthetic`, `type`

## Usage

```python
from datasets import load_dataset
ds = load_dataset("SriRamanaAtmic/AtmicContrastive")
print(ds["train"][0])
```

## Related Resources

| Resource | Link |
|----------|------|
| Embedding Model | [SriRamanaAtmic/AtmicEmbeddingv1](https://huggingface.co/SriRamanaAtmic/AtmicEmbeddingv1) |
| AtmicMLM | [datasets/SriRamanaAtmic/AtmicMLM](https://huggingface.co/datasets/SriRamanaAtmic/AtmicMLM) |
| AtmicContrastive | [datasets/SriRamanaAtmic/AtmicContrastive](https://huggingface.co/datasets/SriRamanaAtmic/AtmicContrastive) |
| AtmicCrosslingual | [datasets/SriRamanaAtmic/AtmicCrosslingual](https://huggingface.co/datasets/SriRamanaAtmic/AtmicCrosslingual) |