参考：
https://github.com/HazyResearch/hyena-dna

整体框架基本就是GPT模型架构
不一样的就是𝖧𝗒𝖾𝗇𝖺𝖣𝖭𝖠 block ，主要是GPT的多重自注意力层引入了cnn

特征向量提取

# python huggingface.py

#@title Single example
import json
import os
import subprocess
# import transformers
from transformers import PreTrainedModel

def inference_single():

    '''
    this selects which backbone to use, and grabs weights/ config from HF
    4 options:
      'hyenadna-tiny-1k-seqlen'   # fine-tune on colab ok
      'hyenadna-small-32k-seqlen'
      'hyenadna-medium-160k-seqlen'  # inference only on colab
      'hyenadna-medium-450k-seqlen'  # inference only on colab
      'hyenadna-large-1m-seqlen'  # inference only on colab
    '''

    # you only need to select which model to use here, we'll do the rest!
    pretrained_model_name = 'hyenadna-small-32k-seqlen'

    max_lengths = {
        'hyenadna-tiny-1k-seqlen': 1024,
        'hyenadna-small-32k-seqlen': 32768,
        'hyenadna-medium-160k-seqlen': 160000,
        'hyenadna-medium-450k-seqlen': 450000,  # T4 up to here
        'hyenadna-large-1m-seqlen': 1_000_000,  # only A100 (paid tier)
    }

    max_length = max_lengths[pretrained_model_name]  # auto selects

    # data settings:
    use_padding = True
    rc_aug = False  # reverse complement augmentation
    add_eos = False  # add end of sentence token

    # we need these for the decoder head, if using
    use_head = False
    n_classes = 2  # not used for embeddings only

    # you can override with your own backbone config here if you want,
    # otherwise we'll load the HF one in None
    backbone_cfg = None

    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    print("Using device:", device)

    # instantiate the model (pretrained here)
    if pretrained_model_name in ['hyenadna-tiny-1k-seqlen',
                                 'hyenadna-small-32k-seqlen',
                                 'hyenadna-medium-160k-seqlen',
                                 'hyenadna-medium-450k-seqlen',
                                 'hyenadna-large-1m-seqlen']:
        # use the pretrained Huggingface wrapper instead
        model = HyenaDNAPreTrainedModel.from_pretrained(
            './checkpoints',
            pretrained_model_name,
            download=True,
            config=backbone_cfg,
            device=device,
            use_head=use_head,
            n_classes=n_classes,
        )

    # from scratch
    elif pretrained_model_name is None:
        model = HyenaDNAModel(**backbone_cfg, use_head=use_head, n_classes=n_classes)

    # create tokenizer
    tokenizer = CharacterTokenizer(
        characters=['A', 'C', 'G', 'T', 'N'],  # add DNA characters, N is uncertain
        model_max_length=max_length + 2,  # to account for special tokens, like EOS
        add_special_tokens=False,  # we handle special tokens elsewhere
        padding_side='left', # since HyenaDNA is causal, we pad on the left
    )

    #### Single embedding example ####

    # create a sample 450k long, prepare
    sequence = 'ACTG' * int(max_length/4)
    tok_seq = tokenizer(sequence)
    tok_seq = tok_seq["input_ids"]  # grab ids

    # place on device, convert to tensor
    tok_seq = torch.LongTensor(tok_seq).unsqueeze(0)  # unsqueeze for batch dim
    tok_seq = tok_seq.to(device)

    # prep model and forward
    model.to(device)
    model.eval()
    with torch.inference_mode():
        embeddings = model(tok_seq)

    print(embeddings.shape)  # embeddings here!

# # uncomment to run! (to get embeddings)
inference_single()