NucEngram / configuration_nucengram.py
FreakingPotato's picture
Add NucEngram genomic language model (base, S123) + usage
cb634e7 verified
Raw
History Blame Contribute Delete
537 Bytes
"""HuggingFace config for the NucEngram genomic language model."""
from transformers import PretrainedConfig
class NucEngramConfig(PretrainedConfig):
model_type = "nucengram"
def __init__(self, backbone=None, engram=None, max_length=8192, pad_token_id=0, **kwargs):
# `backbone` = ModernBertGenomicConfig kwargs; `engram` = EngramConfig kwargs.
self.backbone = backbone or {}
self.engram = engram or {}
self.max_length = max_length
super().__init__(pad_token_id=pad_token_id, **kwargs)