GPT2LMHeadModelPytorch Lightning を使用してカジュアルな言語モデリング (与えられた一連の単語から次の単語を予測する) 用に Huggingface モデルを微調整しようとすると、トレーニング中にエラーが発生します。
AttributeError: 'str' オブジェクトには属性 'size' がありません
トレーニング コードの何が問題だったのでしょうか? DataCollatorForLanguageModelingこれは Pytorch での間違った使用によるものDataLoaderですか?
再現可能な例:
import os
from pathlib import Path
import torch
import pytorch_lightning as pl
from transformers import (
GPT2Config,
GPT2LMHeadModel,
GPT2Tokenizer,
DataCollatorForLanguageModeling,
)
from transformers.optimization import AdamW
from tokenizers import ByteLevelBPETokenizer
from torch.utils.data import (
DataLoader,
Dataset,
)
TOKENIZER_DIRPATH = os.path.join("..", "data")
def tokenize_data():
tokenizer = ByteLevelBPETokenizer()
tokenizer.train(
files=os.path.join(TOKENIZER_DIRPATH, "words.txt"),
vocab_size=50000,
min_frequency=2,
special_tokens=["<s>", "</s>", "<unk>", "<mask>", "<pad>",],
)
tokenizer.save_model("../data")
class MyDataset(Dataset):
def __init__(self):
tokenizer = GPT2Tokenizer(
os.path.join(TOKENIZER_DIRPATH, "vocab.json"),
os.path.join(TOKENIZER_DIRPATH, "merges.txt"),
)
src_file = Path(os.path.join(TOKENIZER_DIRPATH, "words.txt"))
lines = src_file.read_text(encoding="utf-8").splitlines()
self.examples = [tokenizer.encode(line) for line in lines]
def __len__(self):
return len(self.examples)
def __getitem__(self, i):
return torch.tensor(self.examples[i])
class MyDataModule(pl.LightningDataModule):
def __init__(self):
super().__init__()
self.tokenizer = GPT2Tokenizer(
os.path.join(TOKENIZER_DIRPATH, "vocab.json"),
os.path.join(TOKENIZER_DIRPATH, "merges.txt"),
)
def setup(self, stage):
self.train_dataset = MyDataset()
def train_dataloader(self):
data_collator = DataCollatorForLanguageModeling(
tokenizer=self.tokenizer, mlm=False
)
train_dataloader = DataLoader(self.train_dataset, collate_fn=data_collator)
return train_dataloader
class MyModel(pl.LightningModule):
def __init__(self, learning_rate, adam_beta1, adam_beta2, adam_epsilon):
super().__init__()
self.save_hyperparameters()
config = GPT2Config()
self.model = GPT2LMHeadModel(config)
def forward(self, x):
return self.model(x).logits
def training_step(self, batch, batch_idx):
input_ids, labels = batch
loss = self.model(input_ids, labels=labels).loss
self.log("train_loss", loss, on_epoch=True)
return loss
def configure_optimizers(self):
optimizer = AdamW(
self.parameters(),
self.hparams.learning_rate,
betas=(self.hparams.adam_beta1, self.hparams.adam_beta2),
eps=self.hparams.adam_epsilon,
)
return optimizer
tokenize_data()
dm = MyDataModule()
model = MyModel(
learning_rate=5e-5, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-8,
)
trainer = pl.Trainer()
trainer.fit(model, dm)
エラー トレースバック:
Epoch 0: 0%| | 0/9 [00:00<?, ?it/s]
Traceback (most recent call last):
File "test_gpt.py", line 102, in <module>
trainer.fit(model, dm)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 499, in fit
self.dispatch()
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 546, in dispatch
self.accelerator.start_training(self)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/accelerators/accelerator.py", line 73, in start_training
self.training_type_plugin.start_training(trainer)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/plugins/training_type/training_type_plugin.py", line 114, in start_training
self._results = trainer.run_train()
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 637, in run_train
self.train_loop.run_training_epoch()
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/training_loop.py", line 493, in run_training_epoch
batch_output = self.run_training_batch(batch, batch_idx, dataloader_idx)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/training_loop.py", line 655, in run_training_batch
self.optimizer_step(optimizer, opt_idx, batch_idx, train_step_and_backward_closure)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/training_loop.py", line 426, in optimizer_step
model_ref.optimizer_step(
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/core/lightning.py", line 1387, in optimizer_step
optimizer.step(closure=optimizer_closure)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/core/optimizer.py", line 214, in step
self.__optimizer_step(*args, closure=closure, profiler_name=profiler_name, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/core/optimizer.py", line 134, in __optimizer_step
trainer.accelerator.optimizer_step(optimizer, self._optimizer_idx, lambda_closure=closure, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/accelerators/accelerator.py", line 277, in optimizer_step
self.run_optimizer_step(optimizer, opt_idx, lambda_closure, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/accelerators/accelerator.py", line 282, in run_optimizer_step
self.training_type_plugin.optimizer_step(optimizer, lambda_closure=lambda_closure, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/plugins/training_type/training_type_plugin.py", line 163, in optimizer_step
optimizer.step(closure=lambda_closure, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/transformers/optimization.py", line 318, in step
loss = closure()
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/training_loop.py", line 649, in train_step_and_backward_closure
result = self.training_step_and_backward(
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/training_loop.py", line 743, in training_step_and_backward
result = self.training_step(split_batch, batch_idx, opt_idx, hiddens)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/trainer/training_loop.py", line 293, in training_step
training_step_output = self.trainer.accelerator.training_step(args)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/accelerators/accelerator.py", line 156, in training_step
return self.training_type_plugin.training_step(*args)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/pytorch_lightning/plugins/training_type/training_type_plugin.py", line 125, in training_step
return self.lightning_module.training_step(*args, **kwargs)
File "test_gpt.py", line 81, in training_step
loss = self.model(input_ids, labels=labels).loss
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/transformers/models/gpt2/modeling_gpt2.py", line 904, in forward
transformer_outputs = self.transformer(
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/opt/anaconda3/envs/test_huggingface/lib/python3.8/site-packages/transformers/models/gpt2/modeling_gpt2.py", line 633, in forward
input_shape = input_ids.size()
AttributeError: 'str' object has no attribute 'size'
コンダパッケージ:
pytorch 1.7.0 py3.8_cuda10.2.89_cudnn7.6.5_0 pytorch
pytorch-lightning 1.2.5 pyhd8ed1ab_0 conda-forge
tokenizers 0.10.1 pypi_0 pypi
transformers 4.4.2 pypi_0 pypi