require_torch_gpu,
require_torch_multi_gpu,
slow,
)
from transformers.trainer_utils import set_seed
bindir = os.path.abspath(os.path.dirname(__file__))
with ExtendSysPath(f"{bindir}/.."):
from test_trainer import TrainerIntegrationCommon # noqa
if is_torch_available():
from test_trainer import get_regression_trainer # noqa
set_seed(42)
MBART_TINY = "sshleifer/tiny-mbart"
T5_SMALL = "t5-small"
def load_json(path):
with open(path) as f:
return json.load(f)
# a candidate for testing_utils
def require_deepspeed(test_case):
"""
Decorator marking a test that requires deepspeed
"""
if not is_deepspeed_available():
def __init__(
self,
args: TFTrainingArguments,
model,
num_train_examples,
total_train_batch_size: Optional[int] = 8,
train_dataset: Optional[tf.data.Dataset] = None,
eval_dataset: Optional[tf.data.Dataset] = None,
compute_metrics: Optional[Callable[[EvalPrediction], Dict]] = None,
tb_writer: Optional[tf.summary.SummaryWriter] = None,
optimizers: Tuple[
tf.keras.optimizers.Optimizer,
tf.keras.optimizers.schedules.LearningRateSchedule] = (
None,
None,
),
**kwargs,
):
assert parse(tf.__version__).release >= (2, 2, 0), (
"You need to run the TensorFlow trainer with at least the version 2.2.0, your version is %r "
% tf.__version__)
self.num_train_examples = num_train_examples
self.total_train_batch_size = total_train_batch_size
self.model = model
self.args = args
self.train_dataset = train_dataset
self.eval_dataset = eval_dataset
self.compute_metrics = compute_metrics
self.optimizer, self.lr_scheduler = optimizers
self.gradient_accumulator = GradientAccumulator()
self.global_step = 0
self.epoch_logging = 0
if "prediction_loss_only" in kwargs:
warnings.warn(
"Passing `prediction_loss_only` as a keyword argument is deprecated and won't be possible in a future version. Use `args.prediction_loss_only` instead.",
FutureWarning,
)
self.args.prediction_loss_only = kwargs.pop("prediction_loss_only")
assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
# if tb_writer is not None:
# self.tb_writer = tb_writer
# else:
# self.tb_writer = tf.summary.create_file_writer(self.args.logging_dir)
# if is_wandb_available():
# self.setup_wandb()
# elif os.environ.get("WANDB_DISABLED") != "true":
# logger.info(
# "You are instantiating a Trainer but W&B is not installed. To use wandb logging, "
# "run `pip install wandb; wandb login` see https://docs.wandb.com/huggingface."
# )
# if is_comet_available():
# self.setup_comet()
# elif os.environ.get("COMET_MODE") != "DISABLED":
# logger.info(
# "To use comet_ml logging, run `pip/conda install comet_ml` "
# "see https://www.comet.ml/docs/python-sdk/huggingface/"
# )
set_seed(self.args.seed)
def main():
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
parser = HfArgumentParser(
(ModelArguments, DataTrainingArguments, TrainingArguments))
if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
model_args, data_args, training_args = parser.parse_json_file(
json_file=os.path.abspath(sys.argv[1]))
else:
model_args, data_args, training_args = parser.parse_args_into_dataclasses(
)
if (os.path.exists(training_args.output_dir)
and os.listdir(training_args.output_dir) and training_args.do_train
and not training_args.overwrite_output_dir):
raise ValueError(
f"Output directory ({training_args.output_dir}) already exists and is not empty."
"Use --overwrite_output_dir to overcome.")
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO
if is_main_process(training_args.local_rank) else logging.WARN,
)
# Log on each process the small summary:
logger.warning(
f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+
f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
)
# Set the verbosity to info of the Transformers logger (on main process only):
if is_main_process(training_args.local_rank):
transformers.utils.logging.set_verbosity_info()
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
logger.info("Training/evaluation parameters %s", training_args)
# Set seed before initializing model.
set_seed(training_args.seed)
# Get the datasets: you can either provide your own CSV/JSON/TXT training and evaluation files (see below)
# or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/
# (the dataset will be downloaded automatically from the datasets Hub
#
# For CSV/JSON files, this script will use the column called 'text' or the first column. You can easily tweak this
# behavior (see below)
#
# In distributed training, the load_dataset function guarantee that only one local process can concurrently
# download the dataset.
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
datasets = load_dataset(data_args.dataset_name,
data_args.dataset_config_name)
if "validation" not in datasets.keys():
datasets["validation"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[:{data_args.validation_split_percentage}%]",
)
datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[{data_args.validation_split_percentage}%:]",
)
else:
data_files = {}
if data_args.train_file is not None:
data_files["train"] = data_args.train_file
if data_args.validation_file is not None:
data_files["validation"] = data_args.validation_file
extension = data_args.train_file.split(".")[-1]
if extension == "txt":
extension = "text"
datasets = load_dataset(extension, data_files=data_files)
# See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at
# https://huggingface.co/docs/datasets/loading_datasets.html.
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
# if model_args.config_name:
# config = AutoConfig.from_pretrained(model_args.config_name, cache_dir=model_args.cache_dir)
# elif model_args.model_name_or_path:
# config = AutoConfig.from_pretrained(model_args.model_name_or_path, cache_dir=model_args.cache_dir)
# else:
# config = CONFIG_MAPPING[model_args.model_type]()
# logger.warning("You are instantiating a new config instance from scratch.")
pretrained_name = "bert-base-uncased"
config = BertConfig.from_pretrained(pretrained_name)
tokenizer = BertTokenizer.from_pretrained(pretrained_name)
# if model_args.model_name_or_path:
# model = AutoModelForMaskedLM.from_pretrained(
# model_args.model_name_or_path,
# from_tf=bool(".ckpt" in model_args.model_name_or_path),
# config=config,
# cache_dir=model_args.cache_dir,
# )
# else:
# logger.info("Training new model from scratch")
# model = AutoModelForMaskedLM.from_config(config)
model = LitOutputBertModel(name="test",
config=config,
tokenizer=tokenizer,
pretrained_name=pretrained_name)
model.bert.train()
print("Requiring grads for bert")
for param in model.bert.parameters(recurse=True):
param.requires_grad = True
for param in model.training_bert.parameters(recurse=True):
param.requires_grad = True
model.bert.resize_token_embeddings(len(tokenizer))
# Preprocessing the datasets.
# First we tokenize all the texts.
if training_args.do_train:
column_names = datasets["train"].column_names
else:
column_names = datasets["validation"].column_names
text_column_name = "text" if "text" in column_names else column_names[0]
if data_args.line_by_line:
# When using line_by_line, we just tokenize each nonempty line.
padding = "max_length" if data_args.pad_to_max_length else False
def tokenize_function(examples):
# Remove empty lines
examples["text"] = [
line for line in examples["text"]
if len(line) > 0 and not line.isspace()
]
return tokenizer(
examples["text"],
padding=padding,
truncation=True,
max_length=data_args.max_seq_length,
# We use this option because DataCollatorForLanguageModeling (see below) is more efficient when it
# receives the `special_tokens_mask`.
return_special_tokens_mask=True,
)
tokenized_datasets = datasets.map(
tokenize_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=[text_column_name],
load_from_cache_file=not data_args.overwrite_cache,
)
else:
# Otherwise, we tokenize every text, then concatenate them together before splitting them in smaller parts.
# We use `return_special_tokens_mask=True` because DataCollatorForLanguageModeling (see below) is more
# efficient when it receives the `special_tokens_mask`.
def tokenize_function(examples):
return tokenizer(examples[text_column_name],
return_special_tokens_mask=True)
tokenized_datasets = datasets.map(
tokenize_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
if data_args.max_seq_length is None:
max_seq_length = tokenizer.model_max_length
else:
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warn(
f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
)
max_seq_length = min(data_args.max_seq_length,
tokenizer.model_max_length)
# Main data processing function that will concatenate all texts from our dataset and generate chunks of
# max_seq_length.
def group_texts(examples):
# Concatenate all texts.
concatenated_examples = {
k: sum(examples[k], [])
for k in examples.keys()
}
total_length = len(concatenated_examples[list(examples.keys())[0]])
# We drop the small remainder, we could add padding if the model supported it instead of this drop, you can
# customize this part to your needs.
total_length = (total_length // max_seq_length) * max_seq_length
# Split by chunks of max_len.
result = {
k: [
t[i:i + max_seq_length]
for i in range(0, total_length, max_seq_length)
]
for k, t in concatenated_examples.items()
}
return result
# Note that with `batched=True`, this map processes 1,000 texts together, so group_texts throws away a
# remainder for each of those groups of 1,000 texts. You can adjust that batch_size here but a higher value
# might be slower to preprocess.
#
# To speed up this part, we use multiprocessing. See the documentation of the map method for more information:
# https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map
tokenized_datasets = tokenized_datasets.map(
group_texts,
batched=True,
num_proc=data_args.preprocessing_num_workers,
load_from_cache_file=not data_args.overwrite_cache,
)
# Data collator
# This one will take care of randomly masking the tokens.
tokenized_datasets["train"]._output_all_columns = True
data_collator = CustomDataCollatorForLanguageModeling(
tokenizer=tokenizer, mlm_probability=data_args.mlm_probability)
# Initialize our Trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=tokenized_datasets["train"]
if training_args.do_train else None,
eval_dataset=tokenized_datasets["validation"]
if training_args.do_eval else None,
tokenizer=tokenizer,
data_collator=data_collator,
)
# Training
if training_args.do_train:
model_path = (model_args.model_name_or_path if
(model_args.model_name_or_path is not None
and os.path.isdir(model_args.model_name_or_path)) else
None)
trainer.train(model_path=model_path)
trainer.save_model() # Saves the tokenizer too for easy upload
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
eval_output = trainer.evaluate()
perplexity = math.exp(eval_output["eval_loss"])
results["perplexity"] = perplexity
output_eval_file = os.path.join(training_args.output_dir,
"eval_results_mlm.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in results.items():
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
return results
import typing as t
from typing import TYPE_CHECKING
import pytest
import requests
import transformers.pipelines
from transformers.trainer_utils import set_seed
import bentoml
if TYPE_CHECKING:
from bentoml._internal.external_typing import transformers as ext
set_seed(123)
MODEL_ID = "julien-c/dummy-unknown"
REVISION_ID_INVALID = "e10"
model_name = "gpt2"
test_sentence = {"text": "A Bento box is a "}
batched_sentence = [
"I love you and I want to spend my whole life with you",
"I hate you, Lyon, you broke my heart.",
]
result = (
"A Bento box is a urn that is used to store the contents of a Bento box. "
+
"It is usually used to store the contents of a Bento box in a storage container."
+ "\n\nThe Bento box is a small, rectangular")
