79 lines
2.3 KiB
Python
79 lines
2.3 KiB
Python
from sklearn.model_selection import train_test_split
|
|
from torch.utils.data import Dataset, DataLoader
|
|
|
|
|
|
class TransformerGen:
|
|
"""Base class for all transformers. It implements the basic methods for
|
|
the creation of the datasets, datalaoders and the train-val split method.
|
|
It is designed to be used with MultilingualDataset in the
|
|
form of dictioanries {lang: data}
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
model_name,
|
|
dataset_name,
|
|
epochs=10,
|
|
lr=1e-5,
|
|
batch_size=4,
|
|
batch_size_eval=32,
|
|
max_length=512,
|
|
print_steps=50,
|
|
device="cpu",
|
|
probabilistic=False,
|
|
n_jobs=-1,
|
|
evaluate_step=10,
|
|
verbose=False,
|
|
patience=5,
|
|
):
|
|
self.model_name = model_name
|
|
self.dataset_name = dataset_name
|
|
self.device = device
|
|
self.model = None
|
|
self.lr = lr
|
|
self.epochs = epochs
|
|
self.tokenizer = None
|
|
self.max_length = max_length
|
|
self.batch_size = batch_size
|
|
self.batch_size_eval = batch_size_eval
|
|
self.print_steps = print_steps
|
|
self.probabilistic = probabilistic
|
|
self.n_jobs = n_jobs
|
|
self.fitted = False
|
|
self.datasets = {}
|
|
self.evaluate_step = evaluate_step
|
|
self.verbose = verbose
|
|
self.patience = patience
|
|
self.datasets = {}
|
|
|
|
def make_probabilistic(self):
|
|
raise NotImplementedError
|
|
|
|
def build_dataloader(
|
|
self,
|
|
lX,
|
|
lY,
|
|
torchDataset,
|
|
processor_fn,
|
|
batch_size,
|
|
split="train",
|
|
shuffle=False,
|
|
):
|
|
l_tokenized = {lang: processor_fn(data) for lang, data in lX.items()}
|
|
self.datasets[split] = torchDataset(l_tokenized, lY, split=split)
|
|
return DataLoader(self.datasets[split], batch_size=batch_size, shuffle=shuffle)
|
|
|
|
def get_train_val_data(self, lX, lY, split=0.2, seed=42):
|
|
tr_lX, tr_lY, val_lX, val_lY = {}, {}, {}, {}
|
|
|
|
for lang in lX.keys():
|
|
tr_X, val_X, tr_Y, val_Y = train_test_split(
|
|
lX[lang], lY[lang], test_size=split, random_state=seed, shuffle=False
|
|
)
|
|
tr_lX[lang] = tr_X
|
|
tr_lY[lang] = tr_Y
|
|
val_lX[lang] = val_X
|
|
val_lY[lang] = val_Y
|
|
|
|
return tr_lX, tr_lY, val_lX, val_lY
|