Datasets:

neuralbioinfo
/

model-registry

Dataset Viewer

The dataset viewer is not available for this dataset.

Unexpected token '<', "<html> <h"... is not valid JSON

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

ProkBERT Training Registry

JSON-backed training registry for ProkBERT and related nucleotide foundation models.

Files:

basemodels.json: model metadata and tokenizer dispatch.
default_training_parameters.json: default finetuning parameters by model and sequence-length range.
finetuning_task.json: placeholder task mapping from the workbook.
task.json: empty sheet placeholder preserved from the workbook.
manifest.json: simple schema and file manifest.

Conventions:

model_id is the canonical short internal model identifier.
basemodel is preserved from the original workbook.
seq_length_min / seq_length_max are raw sequence-length ranges.
max_token_length is the model-facing token-length cap used by training defaults.

This dataset is intended to be consumed directly from the Hugging Face Hub by the ProkBERT helper utilities.