trumancai/Revela-1b

Model Summary

Revela is a self-supervised bi-encoder retrieval model, trained on raw text with an in-batch attention mechanism. This version, Revela-1b was trained on a corpus of 320K batches in the size of 16 by chunking Wikipedia. See the paper for more details.

• Repository: TRUMANCFY/Revela
• Training Dataset: trumancai/revela_training_corpus

Other Links

Binary	Description
trumancai/Revela-code-3b	3 B-parameter code-retriever.
trumancai/Revela-code-1b	1 B-parameter code-retriever.
trumancai/Revela-code-500M	500 M-parameter code-retriever.
trumancai/Revela-3b	3 B-parameter Wikipedia retriever.
trumancai/Revela-1b	1 B-parameter Wikipedia retriever.
trumancai/Revela-500M	500 M-parameter Wikipedia retriever.
trumancai/revela_code_training_corpus	Code training corpus.
trumancai/revela_training_corpus	Wikipedia training corpus.

Usage

We can evaluate the trained models with customized mteb.

from mteb.model_meta import ModelMeta
from mteb.models.repllama_models import RepLLaMAWrapper, _loader

revela_llama_3b = ModelMeta(
    loader=_loader(
        RepLLaMAWrapper,
        base_model_name_or_path="meta-llama/Llama-3.2-3B",
        peft_model_name_or_path="trumancai/Revela-3b",
        device_map="auto",
        torch_dtype=torch.bfloat16,
    ),
    name="trumancai/Revela-3b",
    languages=["eng_Latn"],
    open_source=True,
    revision="2b31c92f23acc46762587ea37cb55032da788561",  # base-peft revision
    release_date="2025-04-13",
)
revela_llama_3b_model = revela_llama_3b.loader()

evaluation = mteb.MTEB(tasks=["SciFact", "NFCorpus"])
evaluation.run(model=revela_llama_3b_model, output_folder="results/Revela-3b")

License

Citation