281 lines
8.8 KiB
Python
281 lines
8.8 KiB
Python
import os
|
|
|
|
os.environ["TOKENIZERS_PARALLELISM"] = "true"
|
|
|
|
from collections import defaultdict
|
|
|
|
import numpy as np
|
|
import torch
|
|
import torch.nn as nn
|
|
import transformers
|
|
from transformers import MT5EncoderModel
|
|
from transformers import AutoModelForSequenceClassification, AutoTokenizer
|
|
from transformers.modeling_outputs import ModelOutput
|
|
|
|
from gfun.vgfs.commons import Trainer
|
|
from gfun.vgfs.transformerGen import TransformerGen
|
|
from gfun.vgfs.viewGen import ViewGen
|
|
from dataManager.torchDataset import MultilingualDatasetTorch
|
|
|
|
transformers.logging.set_verbosity_error()
|
|
|
|
|
|
class MT5ForSequenceClassification(nn.Module):
|
|
def __init__(self, model_name, num_labels, output_hidden_states):
|
|
super().__init__()
|
|
|
|
self.output_hidden_states = output_hidden_states
|
|
self.mt5encoder = MT5EncoderModel.from_pretrained(
|
|
model_name, output_hidden_states=True
|
|
)
|
|
self.dropout = nn.Dropout(0.1)
|
|
self.linear = nn.Linear(512, num_labels)
|
|
|
|
def forward(self, input_ids):
|
|
embed = self.mt5encoder(input_ids=input_ids)
|
|
pooled = torch.mean(embed.last_hidden_state, dim=1)
|
|
outputs = self.dropout(pooled)
|
|
logits = self.linear(outputs)
|
|
if self.output_hidden_states:
|
|
return ModelOutput(
|
|
logits=logits,
|
|
pooled=pooled,
|
|
)
|
|
return ModelOutput(logits=logits)
|
|
|
|
def save_pretrained(self, checkpoint_dir):
|
|
torch.save(self.state_dict(), checkpoint_dir + ".pt")
|
|
return
|
|
|
|
def from_pretrained(self, checkpoint_dir):
|
|
checkpoint_dir += ".pt"
|
|
return self.load_state_dict(torch.load(checkpoint_dir))
|
|
|
|
|
|
class TextualTransformerGen(ViewGen, TransformerGen):
|
|
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,
|
|
classification_type="multilabel",
|
|
scheduler="ReduceLROnPlateau",
|
|
):
|
|
super().__init__(
|
|
self._validate_model_name(model_name),
|
|
dataset_name,
|
|
epochs=epochs,
|
|
lr=lr,
|
|
scheduler=scheduler,
|
|
batch_size=batch_size,
|
|
batch_size_eval=batch_size_eval,
|
|
device=device,
|
|
evaluate_step=evaluate_step,
|
|
patience=patience,
|
|
probabilistic=probabilistic,
|
|
max_length=max_length,
|
|
print_steps=print_steps,
|
|
n_jobs=n_jobs,
|
|
verbose=verbose,
|
|
)
|
|
self.clf_type = classification_type
|
|
self.fitted = False
|
|
print(
|
|
f"- init Textual TransformerModel model_name: {self.model_name}, device: {self.device}]"
|
|
)
|
|
|
|
def _validate_model_name(self, model_name):
|
|
if "bert" == model_name:
|
|
return "bert-base-uncased"
|
|
elif "mbert" == model_name:
|
|
return "bert-base-multilingual-uncased"
|
|
elif "xlm-roberta" == model_name:
|
|
return "xlm-roberta-base"
|
|
elif "mt5" == model_name:
|
|
return "google/mt5-small"
|
|
else:
|
|
raise NotImplementedError
|
|
|
|
def load_pretrained_model(self, model_name, num_labels):
|
|
if model_name == "google/mt5-small":
|
|
return MT5ForSequenceClassification(
|
|
model_name, num_labels=num_labels, output_hidden_states=True
|
|
)
|
|
else:
|
|
return AutoModelForSequenceClassification.from_pretrained(
|
|
model_name, num_labels=num_labels, output_hidden_states=True
|
|
)
|
|
|
|
def load_tokenizer(self, model_name):
|
|
return AutoTokenizer.from_pretrained(model_name)
|
|
|
|
def init_model(self, model_name, num_labels):
|
|
return self.load_pretrained_model(model_name, num_labels), self.load_tokenizer(
|
|
model_name
|
|
)
|
|
|
|
def _tokenize(self, X):
|
|
return self.tokenizer(
|
|
X,
|
|
return_tensors="pt",
|
|
padding="max_length",
|
|
truncation=True,
|
|
max_length=self.max_length,
|
|
)
|
|
|
|
def fit(self, lX, lY):
|
|
if self.fitted:
|
|
return self
|
|
print("- fitting Textual Transformer View Generating Function")
|
|
_l = list(lX.keys())[0]
|
|
self.num_labels = lY[_l].shape[-1]
|
|
self.model, self.tokenizer = self.init_model(
|
|
self.model_name, num_labels=self.num_labels
|
|
)
|
|
|
|
tr_lX, tr_lY, val_lX, val_lY = self.get_train_val_data(
|
|
lX, lY, split=0.2, seed=42, modality="text"
|
|
)
|
|
|
|
tra_dataloader = self.build_dataloader(
|
|
tr_lX,
|
|
tr_lY,
|
|
processor_fn=self._tokenize,
|
|
torchDataset=MultilingualDatasetTorch,
|
|
batch_size=self.batch_size,
|
|
split="train",
|
|
shuffle=True,
|
|
)
|
|
|
|
val_dataloader = self.build_dataloader(
|
|
val_lX,
|
|
val_lY,
|
|
processor_fn=self._tokenize,
|
|
torchDataset=MultilingualDatasetTorch,
|
|
batch_size=self.batch_size_eval,
|
|
split="val",
|
|
shuffle=False,
|
|
)
|
|
|
|
experiment_name = f"{self.model_name.replace('/', '-')}-{self.epochs}-{self.batch_size}-{self.dataset_name}"
|
|
|
|
trainer = Trainer(
|
|
model=self.model,
|
|
optimizer_name="adamW",
|
|
lr=self.lr,
|
|
device=self.device,
|
|
loss_fn=torch.nn.CrossEntropyLoss(),
|
|
print_steps=self.print_steps,
|
|
evaluate_step=self.evaluate_step,
|
|
patience=self.patience,
|
|
experiment_name=experiment_name,
|
|
checkpoint_path=os.path.join(
|
|
"models",
|
|
"vgfs",
|
|
"transformer",
|
|
self._format_model_name(self.model_name),
|
|
),
|
|
vgf_name="textual_trf",
|
|
classification_type=self.clf_type,
|
|
n_jobs=self.n_jobs,
|
|
scheduler_name=self.scheduler,
|
|
)
|
|
trainer.train(
|
|
train_dataloader=tra_dataloader,
|
|
eval_dataloader=val_dataloader,
|
|
epochs=self.epochs,
|
|
)
|
|
|
|
if self.probabilistic:
|
|
self.feature2posterior_projector.fit(self.transform(lX), lY)
|
|
|
|
self.fitted = True
|
|
|
|
return self
|
|
|
|
def transform(self, lX):
|
|
# forcing to only text modality
|
|
lX = {lang: data["text"] for lang, data in lX.items()}
|
|
_embeds = []
|
|
l_embeds = defaultdict(list)
|
|
|
|
dataloader = self.build_dataloader(
|
|
lX,
|
|
lY=None,
|
|
processor_fn=self._tokenize,
|
|
torchDataset=MultilingualDatasetTorch,
|
|
batch_size=self.batch_size_eval,
|
|
split="whole",
|
|
shuffle=False,
|
|
)
|
|
|
|
self.model.eval()
|
|
with torch.no_grad():
|
|
for input_ids, lang in dataloader:
|
|
input_ids = input_ids.to(self.device)
|
|
# TODO: check this
|
|
if isinstance(self.model, MT5ForSequenceClassification):
|
|
batch_embeddings = self.model(input_ids).pooled.cpu().numpy()
|
|
else:
|
|
out = self.model(input_ids).hidden_states[-1]
|
|
batch_embeddings = out[:, 0, :].cpu().numpy()
|
|
_embeds.append((batch_embeddings, lang))
|
|
|
|
for embed, lang in _embeds:
|
|
for sample_embed, sample_lang in zip(embed, lang):
|
|
l_embeds[sample_lang].append(sample_embed)
|
|
|
|
if self.probabilistic and self.fitted:
|
|
l_embeds = self.feature2posterior_projector.transform(l_embeds)
|
|
elif not self.probabilistic and self.fitted:
|
|
l_embeds = {lang: np.array(preds) for lang, preds in l_embeds.items()}
|
|
|
|
return l_embeds
|
|
|
|
def fit_transform(self, lX, lY):
|
|
return self.fit(lX, lY).transform(lX)
|
|
|
|
def save_vgf(self, model_id):
|
|
import pickle
|
|
from os import makedirs
|
|
from os.path import join
|
|
|
|
vgf_name = "textualTransformerGen"
|
|
_basedir = join("models", "vgfs", "textual_transformer")
|
|
makedirs(_basedir, exist_ok=True)
|
|
_path = join(_basedir, f"{vgf_name}_{model_id}.pkl")
|
|
with open(_path, "wb") as f:
|
|
pickle.dump(self, f)
|
|
return self
|
|
|
|
def freeze_model(self):
|
|
# TODO: up to n-layers? or all? avoid freezing head ovb...
|
|
for param in self.model.parameters():
|
|
param.requires_grad = False
|
|
|
|
def _format_model_name(self, model_name):
|
|
if "mt5" in model_name:
|
|
return "google-mt5"
|
|
elif "bert" in model_name:
|
|
if "multilingual" in model_name:
|
|
return "mbert"
|
|
elif "xlm-roberta" in model_name:
|
|
return "xlm-roberta"
|
|
else:
|
|
return model_name
|
|
|
|
def get_config(self):
|
|
c = super().get_config()
|
|
return {"textual_trf": c}
|