PiccoviralesGPT

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

output_v3

This model is a fine-tuned version of avuhong/ParvoGPT2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4775
• Accuracy: 0.9290

Model description

This model is a GPT2-like model for generating capsid amino acid sequences. It was trained exclusively on capsid aa_seqs of Piccovirales members.

Intended uses & limitations

As a typical GPT model, it can be used to generate new sequences or used to evaluate the perplexity of given sequences.

Generate novel sequences for viral capsid proteins

from transformers import pipeline
protgpt2 = pipeline('text-generation', model="avuhong/PiccoviralesGPT")
sequences = protgpt2("<|endoftext|>", max_length=750, do_sample=True, top_k=950, repetition_penalty=1.2, num_return_sequences=10, eos_token_id=0)

Calculate the perplexity of a protein sequence

def calculatePerplexity(sequence, model, tokenizer):
    input_ids = torch.tensor(tokenizer.encode(sequence)).unsqueeze(0) 
    input_ids = input_ids.to(device)
    with torch.no_grad():
        outputs = model(input_ids, labels=input_ids)
    loss, logits = outputs[:2]
    return math.exp(loss)

def split_sequence(sequence):
    chunks = []
    max_i = 0
    for i in range(0, len(sequence), 60):
        chunk = sequence[i:i+60]
        
        if i == 0:
            chunk = '<|endoftext|>' + chunk
        chunks.append(chunk)
        max_i = i
    
    chunks = '\n'.join(chunks)
    
    if max_i+61==len(sequence):
        chunks = chunks+"\n<|endoftext|>"
    else:
        chunks = chunks+"<|endoftext|>"
    return chunks

seq = "MAADGYLPDWLEDNLSEGIREWWALKPGAPQPKANQQHQDNARGLVLPGYKYLGPGNGLDKGEPVNAADAAALEHDKAYDQQLKAGDNPYLKYNHADAEFQERLKEDTSFGGNLGRAVFQAKKRLLEPLGLVEEAAKTAPGKKRPVEQSPQEPDSSAGIGKSGAQPAKKRLNFGQTGDTESVPDPQPIGEPPAAPSGVGSLTMASGGGAPVADNNEGADGVGSSSGNWHCDSQWLGDRVITTSTRTWALPTYNNHLYKQISNSTSGGSSNDNAYFGYSTPWGYFDFNRFHCHFSPRDWQRLINNNWGFRPKRLNFKLFNIQVKEVTDNNGVKTIANNLTSTVQVFTDSDYQLPYVLGSAHEGCLPPFPADVFMIPQYGYLTLNDGSQAVGRSSFYCLEYFPSQMLRTGNNFQFSYEFENVPFHSSYAHSQSLDRLMNPLIDQYLYYLSKTINGSGQNQQTLKFSVAGPSNMAVQGRNYIPGPSYRQQRVSTTVTQNNNSEFAWPGASSWALNGRNSLMNPGPAMASHKEGEDRFFPLSGSLIFGKQGTGRDNVDADKVMITNEEEIKTTNPVATESYGQVATNHQSAQAQAQTGWVQNQGILPGMVWQDRDVYLQGPIWAKIPHTDGNFHPSPLMGGFGMKHPPPQILIKNTPVPADPPTAFNKDKLNSFITQYSTGQVSVEIEWELQKENSKRWNPEIQYTSNYYKSNNVEFAVNTEGVYSEPRPIGTRYLTRNL"
seq = split_sequence(seq)
print(f"{calculatePerplexity(seq, model, tokenizer):.2f}")

Training and evaluation data

Traning script is included in bash file in this repository.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• total_eval_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 32.0
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu117
• Datasets 2.9.0
• Tokenizers 0.13.2