Dataset · MineAITechnology
Embedding model training data (EN/UR)
MineAITechnology/mine-embed-v1
An experimental English and Urdu bilingual corpus for MLM pre-training and contrastive embedding tuning.
What's inside
A dataset of fewer than 1,000 records in English and Urdu used to create the experimental mine-embed-v1 model. It consists of two parts: a tokenized corpus (GPT-2 BPE) for masked language modeling (MLM) pre-training, and a set of contrastive triplets (anchor, positive, and hard negative examples).
What you can do with it
- Experiments with contrastive embedding model training on an English-Urdu language pair
- Testing fine-tuning pipelines for semantic search tasks
- Tasks
- Embeddings, Sentence similarity
- Languages
- English, Urdu
- Size
- up to 1K rows
- Downloads this month
- 217
- Likes
- 7
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("MineAITechnology/mine-embed-v1")