144M_H_MLM_last / README.md

pantoniadis

Update README.md

b4df2d6 verified 2 months ago

preview code

raw

history blame contribute delete

621 Bytes

metadata

license: mit
language:
  - en

Load the model:

import torch
from transformers import AutoModel, AutoTokenizer

model_name = "rnalm/144M_H_MLM_last"

tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModel.from_pretrained(model_name, trust_remote_code=True)

# Move model to GPU
model = model.cuda()

Get embeddings:

inputs = tokenizer("ACGTACGT", return_tensors="pt")

with torch.no_grad():
    outputs = model(input_ids=inputs["input_ids"].cuda())

outputs.last_hidden_state.shape
# torch.Size([1, 8, 768])

outputs.seq_logits.shape
# torch.Size([1, 8, 11])