Biatron/few-shot-test

pip install git+https://github.com/Fazzioni/Biatron.git

from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer,  StoppingCriteria, StoppingCriteriaList
from biatron import BiatronForCausalLM, BiatronConfig
import torch
import datasets
import pandas as pd
from tqdm import tqdm
AutoConfig.register("Biatron", BiatronConfig)
AutoModelForCausalLM.register(BiatronConfig, BiatronForCausalLM)

model_id = 'Fazzioni/biatron-345m'


class StopOnNewLine(StoppingCriteria):
    def __init__(self, tokenizer):
        self.tokenizer = tokenizer
    def __call__(self, input_ids, scores, **kwargs):
        return self.tokenizer.decode(input_ids[0, :-1]).endswith("\n")




dataset = datasets.load_dataset("Biatron/few-shot-test", 'verbs-conjugation',split='test')
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained(model_id)


stopping_criteria = StoppingCriteriaList([ StopOnNewLine(tokenizer) ])
outputs = []
for sample in tqdm(dataset):
    text = sample['input']

    inputs = tokenizer([text], return_tensors="pt").to(model.device)

    if 'token_type_ids' in inputs:
        del inputs['token_type_ids']

    with torch.inference_mode():
        generation = model.generate(**inputs, max_new_tokens=10, do_sample=False, stopping_criteria=stopping_criteria)
        output = generation[0][inputs['input_ids'].shape[1]:]
        output = tokenizer.decode(output)
        output = output.split("\n")[0].strip(" ")

    sample['answer'] = output
    sample['accurate'] = output == sample['output']
    outputs.append(sample)


df = pd.DataFrame(outputs)
print('model  accuracy :', df['accurate'].mean())