Model Card for Mistral-DNA-v1-422M-Athaliana (Mistral for DNA)

The Mistral-DNA-v1-422M-Athaliana Large Language Model (LLM) is a pretrained generative DNA sequence model with 422M parameters. It is derived from Mixtral-8x7B-v0.1 model, which was simplified for DNA: the number of layers and the hidden size were reduced. The model was pretrained using 10kb DNA sequences from 7 A. thaliana genome assemblies (from https://1001genomes.org/data/MPIPZ/MPIPZJiao2020/releases/current/full_set/).

Model Architecture

Like Mixtral-8x7B-v0.1, it is a transformer model, with the following architecture choices:

• Grouped-Query Attention
• Sliding-Window Attention
• Byte-fallback BPE tokenizer
• Mixture of Experts

Load the model from huggingface:

import torch
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("RaphaelMourad/Mistral-DNA-v1-422M-Athaliana", trust_remote_code=True) 
model = AutoModel.from_pretrained("RaphaelMourad/Mistral-DNA-v1-422M-Athaliana", trust_remote_code=True)

Calculate the embedding of a protein sequence

dna = "TGATGATTGGCGCGGCTAGGATCGGCT"
inputs = tokenizer(dna, return_tensors = 'pt')["input_ids"]
hidden_states = model(inputs)[0] # [1, sequence_length, 256]

# embedding with max pooling
embedding_max = torch.max(hidden_states[0], dim=0)[0]
print(embedding_max.shape) # expect to be 256

Troubleshooting

Ensure you are utilizing a stable version of Transformers, 4.34.0 or newer.

Notice

Mistral-DNA-v1-422M-Athaliana is a pretrained base model for DNA.

Contact

Raphaël Mourad. raphael.mourad@univ-tlse3.fr