#!/usr/bin/env python
# coding=utf-8
# Copyright 2020 The HuggingFace Inc. team. All rights reserved.
# Modifications copyright (C) 2022 IBM
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
""" Finetuning the library models for sequence classification tasks."""
# You can also adapt this script on your own text classification task. Pointers for this are left as comments.
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import numpy as np
from sklearn.metrics import f1_score
import datasets
import evaluate
from datasets import load_dataset
from datasets import features, ClassLabel
from aim.hugging_face import AimCallback
import transformers
from transformers import (
AutoConfig,
AutoModelForSequenceClassification,
AutoTokenizer,
EarlyStoppingCallback,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
PretrainedConfig,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
IntervalStrategy,
PreTrainedModel,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
import shutil
import torch
from torch import nn
os.environ["TOKENIZERS_PARALLELISM"] = "false"
os.environ["AIM_UI_TELEMETRY_ENABLED"] = "0"
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.10.0.dev0")
require_version(
"datasets>=1.8.0",
"To fix: pip install -r examples/pytorch/text-classification/requirements.txt",
)
task_to_keys = {
"cola": ("sentence", None),
"mnli": ("premise", "hypothesis"),
"mrpc": ("sentence1", "sentence2"),
"qnli": ("question", "sentence"),
"qqp": ("question1", "question2"),
"rte": ("sentence1", "sentence2"),
"sst2": ("sentence", None),
"stsb": ("sentence1", "sentence2"),
"wnli": ("sentence1", "sentence2"),
"sentiment": ("sentence", None),
"generic": ("sentence", None),
"rotten_tomatoes": ("text", None)
}
logger = logging.getLogger(__name__)
[docs]@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
Using `HfArgumentParser` we can turn this class
into argparse arguments to be able to specify them on
the command line.
"""
task_name: Optional[str] = field(
default=None,
metadata={
"help": "The name of the task to train on: "
+ ", ".join(task_to_keys.keys())
},
)
dataset_name: Optional[str] = field(
default=None,
metadata={"help": "The name of the dataset to use (via the datasets library)."},
)
dataset_config_name: Optional[str] = field(
default=None,
metadata={
"help": "The configuration name of the dataset to use (via the datasets library)."
},
)
max_seq_length: int = field(
default=128,
metadata={
"help": "The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
},
)
overwrite_cache: bool = field(
default=False,
metadata={"help": "Overwrite the cached preprocessed datasets or not."},
)
pad_to_max_length: bool = field(
default=True,
metadata={
"help": "Whether to pad all samples to `max_seq_length`. "
"If False, will pad the samples dynamically when batching to the maximum length in the batch."
},
)
max_train_samples: Optional[int] = field(
default=None,
metadata={
"help": "For debugging purposes or quicker training, truncate the number of training examples to this "
"value if set."
},
)
max_eval_samples: Optional[int] = field(
default=None,
metadata={
"help": "For debugging purposes or quicker training, truncate the number of evaluation examples to this "
"value if set."
},
)
max_predict_samples: Optional[int] = field(
default=None,
metadata={
"help": "For debugging purposes or quicker training, truncate the number of prediction examples to this "
"value if set."
},
)
train_file: Optional[str] = field(
default=None,
metadata={"help": "A csv or a json file containing the training data."},
)
validation_file: Optional[str] = field(
default=None,
metadata={"help": "A csv or a json file containing the validation data."},
)
test_file: Optional[str] = field(
default=None,
metadata={"help": "A csv or a json file containing the test data."},
)
def __post_init__(self):
if self.task_name is not None:
self.task_name = self.task_name.lower()
if self.task_name not in task_to_keys.keys():
raise ValueError(
"Unknown task, you should pick one in "
+ ",".join(task_to_keys.keys())
)
elif self.dataset_name is not None:
pass
elif self.train_file is None or self.validation_file is None:
raise ValueError(
"Need either a GLUE task, a training/validation file or a dataset name."
)
else:
train_extension = self.train_file.split(".")[-1]
assert train_extension in [
"csv",
"json",
"txt",
], "`train_file` should be a csv or a json file."
validation_extension = self.validation_file.split(".")[-1]
assert (
validation_extension == train_extension
), "`validation_file` should have the same extension (csv or json) as `train_file`."
[docs]@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
model_name_or_path: str = field(
metadata={
"help": "Path to pretrained model or model identifier from huggingface.co/models"
}
)
config_name: Optional[str] = field(
default=None,
metadata={
"help": "Pretrained config name or path if not the same as model_name"
},
)
tokenizer_name: Optional[str] = field(
default=None,
metadata={
"help": "Pretrained tokenizer name or path if not the same as model_name"
},
)
cache_dir: Optional[str] = field(
default=None,
metadata={
"help": "Where do you want to store the pretrained models downloaded from huggingface.co"
},
)
use_fast_tokenizer: bool = field(
default=True,
metadata={
"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."
},
)
model_revision: str = field(
default="main",
metadata={
"help": "The specific model version to use (can be a branch name, tag name or commit id)."
},
)
use_auth_token: bool = field(
default=False,
metadata={
"help": "Will use the token generated when running `transformers-cli login` (necessary to use this script "
"with private models)."
},
)
log_dir: Optional[str] = field(
default=None,
metadata={
"help": "Where do you want to store the log files"
},
)
[docs]@dataclass
class TaskArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
task: str = field(metadata={"help": "name of taks: pos or ner"})
early_stop: bool = field(default=False, metadata={"help": "Use early stopping "})
[docs]def main(args):
# 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, TaskArguments)
)
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, task_args = parser.parse_json_file(
json_file=os.path.abspath(sys.argv[1])
)
else:
(
model_args,
data_args,
training_args,
task_args,
) = parser.parse_args_into_dataclasses(args)
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
handlers=[logging.StreamHandler(sys.stdout)],
)
log_level = training_args.get_process_log_level()
logger.setLevel(log_level)
datasets.utils.logging.set_verbosity(log_level)
transformers.utils.logging.set_verbosity(log_level)
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# 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}"
)
logger.info(f"Training/evaluation parameters {training_args}")
# Detecting last checkpoint.
last_checkpoint = None
if (
os.path.isdir(training_args.output_dir)
and training_args.do_train
and not training_args.overwrite_output_dir
):
last_checkpoint = get_last_checkpoint(training_args.output_dir)
if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
raise ValueError(
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif (
last_checkpoint is not None and training_args.resume_from_checkpoint is None
):
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
)
# Set seed before initializing model.
set_seed(training_args.seed)
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
raw_datasets = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
cache_dir=model_args.cache_dir,
use_auth_token=True if model_args.use_auth_token else None,
)
else:
# Loading a dataset from your local files.
# CSV/JSON training and evaluation files are needed.
data_files = {
"train": data_args.train_file,
"validation": data_args.validation_file,
"test": data_args.test_file,
}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
if training_args.do_predict:
if data_args.test_file is not None:
train_extension = data_args.train_file.split(".")[-1]
test_extension = data_args.test_file.split(".")[-1]
assert (
test_extension == train_extension
), "`test_file` should have the same extension (csv or json) as `train_file`."
data_files["test"] = data_args.test_file
else:
raise ValueError("Need either a GLUE task or a test file for `do_predict`.")
for key in data_files.keys():
logger.info(f"load a local file for {key}: {data_files[key]}")
if data_args.train_file.endswith(".csv"):
# Loading a dataset from local csv files
raw_datasets = load_dataset(
"csv",
data_files=data_files,
cache_dir=model_args.cache_dir,
delimiter="\t",
)
elif data_args.train_file.endswith(".json"):
raw_datasets = load_dataset(
"json", data_files=data_files, cache_dir=model_args.cache_dir
)
else:
# Loading a dataset from local json files
raw_datasets = load_dataset(
"csv",
data_files=data_files,
cache_dir=model_args.cache_dir,
delimiter="\t",
)
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.html.
for key in raw_datasets:
if "label" in raw_datasets[key].column_names:
logger.info("We found \"label\" column in the dataset. We need to rename it to \"labels\"")
raw_datasets[key] = raw_datasets[key].rename_column("label", "labels")
if not isinstance(raw_datasets[key].features['labels'], ClassLabel):
raw_datasets[key] = raw_datasets[key].class_encode_column("labels")
# Labels
if data_args.task_name is not None:
is_regression = data_args.task_name == "stsb"
if not is_regression:
label_list = raw_datasets["train"].features["labels"].names
num_labels = len(label_list)
else:
num_labels = 1
else:
# Trying to have good defaults here, don't hesitate to tweak to your needs.
is_regression = raw_datasets["train"].features["labels"].dtype in [
"float32",
"float64",
]
if is_regression:
num_labels = 1
else:
# A useful fast method:
# https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.unique
label_list = raw_datasets["train"].unique("labels")
label_list.sort() # Let's sort it for determinism
num_labels = len(label_list)
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
config = AutoConfig.from_pretrained(
model_args.config_name
if model_args.config_name
else model_args.model_name_or_path,
num_labels=num_labels,
cache_dir=model_args.cache_dir,
revision=model_args.model_revision,
use_auth_token=True if model_args.use_auth_token else None,
)
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name
if model_args.tokenizer_name
else model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
)
model = AutoModelForSequenceClassification.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,
revision=model_args.model_revision,
use_auth_token=True if model_args.use_auth_token else None,
)
n_params = sum(
dict((p.data_ptr(), p.numel()) for p in model.parameters()).values()
)
logger.warning(
f"Fine-tuning existing model - Total size = {n_params/2**20:.2f}M params"
)
logger.warning(model)
sentence1_key, sentence2_key = task_to_keys[data_args.task_name]
logger.warning("Task name is {0}".format(data_args.task_name))
logger.warning("Sentence 1 key is {0}".format(sentence1_key))
logger.warning("Sentence 2 key is {0}".format(sentence2_key))
# Padding strategy
if data_args.pad_to_max_length:
padding = "max_length"
else:
# We will pad later, dynamically at batch creation, to the max sequence length in each batch
padding = False
if not is_regression:
label_to_id = None
label_to_id = {v: raw_datasets["train"].features["labels"].str2int(v) for v in label_list}
if label_to_id is not None:
model.config.label2id = label_to_id
model.config.id2label = {id: label for label, id in config.label2id.items()}
logger.warning('Label2id:')
logger.warning(model.config.label2id)
logger.warning('id2Label:')
logger.warning(model.config.id2label)
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warning(
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)
def preprocess_function(examples):
# Tokenize the texts
args = (
(examples[sentence1_key],)
if sentence2_key is None
else (examples[sentence1_key], examples[sentence2_key])
)
result = tokenizer(
*args, padding=padding, max_length=max_seq_length, truncation=True
)
if label_to_id is not None and "labels" in examples:
result["labels"] = [
l for l in examples["labels"]
]
return result
with training_args.main_process_first(desc="dataset map pre-processing"):
raw_datasets = raw_datasets.map(
preprocess_function,
batched=True,
load_from_cache_file=False,
desc="Running tokenizer on dataset",
)
if training_args.do_train:
if "train" not in raw_datasets:
raise ValueError("--do_train requires a train dataset")
train_dataset = raw_datasets["train"]
if data_args.max_train_samples is not None:
train_dataset = train_dataset.select(range(data_args.max_train_samples))
if training_args.do_eval:
if (
"validation" not in raw_datasets
and "validation_matched" not in raw_datasets
):
raise ValueError("--do_eval requires a validation dataset")
eval_dataset = raw_datasets[
"validation_matched" if data_args.task_name == "mnli" else "validation"
]
if data_args.max_eval_samples is not None:
eval_dataset = eval_dataset.select(range(data_args.max_eval_samples))
if (
training_args.do_predict
or data_args.task_name is not None
or data_args.test_file is not None
):
if "test" not in raw_datasets and "test_matched" not in raw_datasets:
raise ValueError("--do_predict requires a test dataset")
predict_dataset = raw_datasets[
"test_matched" if data_args.task_name == "mnli" else "test"
]
if data_args.max_predict_samples is not None:
predict_dataset = predict_dataset.select(
range(data_args.max_predict_samples)
)
# Log a few random samples from the training set:
if training_args.do_train:
for index in random.sample(range(len(train_dataset)), 3):
logger.info(f"Sample {index} of the training set: {train_dataset[index]}.")
# Metrics
process_id = os.path.basename(os.path.normpath(training_args.output_dir))
metric = evaluate.load("f1", experiment_id=process_id, cache_dir=model_args.cache_dir)
aim_callback = AimCallback(repo=model_args.log_dir, experiment=process_id)
# You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
# predictions and label_ids field) and has to return a dictionary string to float.
def compute_metrics(p: EvalPrediction):
preds = p.predictions[0] if isinstance(p.predictions, tuple) else p.predictions
preds = np.squeeze(preds) if is_regression else np.argmax(preds, axis=1)
if data_args.task_name is not None:
result = metric.compute(predictions=preds, references=p.label_ids)
return result
elif is_regression:
return {"mse": ((preds - p.label_ids) ** 2).mean().item()}
else:
logger.warning(
"Accuracy: "
+ str((preds == p.label_ids).astype(np.float32).mean().item())
)
logger.warning("F1: " + str(f1_score(p.label_ids, preds, average="macro")))
return {
"accuracy": (preds == p.label_ids).astype(np.float32).mean().item(),
"f1": f1_score(p.label_ids, preds, average="macro"),
"micro f1": f1_score(p.label_ids, preds, average="micro"),
}
# Data collator will default to DataCollatorWithPadding, so we change it if we already did the padding.
if data_args.pad_to_max_length:
data_collator = default_data_collator
elif training_args.fp16:
data_collator = DataCollatorWithPadding(tokenizer, pad_to_multiple_of=8)
else:
data_collator = None
# Initialize our Trainer
logger.warning("Initializing Trainer")
early_stopping_callback = EarlyStoppingCallback(early_stopping_patience=5)
training_args.metric_for_best_model = "eval_f1"
training_args.load_best_model_at_end = True
training_args.evaluation_strategy = IntervalStrategy.STEPS
training_args.eval_steps = training_args.save_steps
training_args.greater_is_better = True
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset if training_args.do_train else None,
eval_dataset=eval_dataset if training_args.do_eval else None,
compute_metrics=compute_metrics,
tokenizer=tokenizer,
data_collator=data_collator,
callbacks=[early_stopping_callback, aim_callback] if task_args.early_stop else [aim_callback],
)
# Training
if training_args.do_train:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
train_result = trainer.train(resume_from_checkpoint=checkpoint)
metrics = train_result.metrics
max_train_samples = (
data_args.max_train_samples
if data_args.max_train_samples is not None
else len(train_dataset)
)
metrics["train_samples"] = min(max_train_samples, len(train_dataset))
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
tokenizer.save_pretrained(training_args.output_dir)
# load the latest checkpoint
checkpoints = os.path.join(training_args.output_dir, "GOAT")
if os.path.exists(checkpoints):
dir = os.listdir(checkpoints)
if len(dir) > 0:
logger.warning("Loading from GOAT model {}".format(checkpoints))
temp_model = AutoModelForSequenceClassification.from_pretrained(checkpoints)
else:
logger.warning("Loading from model {}".format(training_args.output_dir))
temp_model = AutoModelForSequenceClassification.from_pretrained(
training_args.output_dir
)
else:
logger.warning("Loading from model {}".format(training_args.output_dir))
temp_model = AutoModelForSequenceClassification.from_pretrained(
training_args.output_dir
)
logger.warning("Copying weights from pre-trained model \n")
trainer.model.load_state_dict(temp_model.state_dict(), strict=True)
if training_args.do_predict:
logger.info("*** Predict ***")
# Loop to handle MNLI double evaluation (matched, mis-matched)
tasks = [data_args.task_name]
predict_datasets = [predict_dataset]
if data_args.task_name == "mnli":
tasks.append("mnli-mm")
predict_datasets.append(raw_datasets["test_mismatched"])
for predict_dataset, task in zip(predict_datasets, tasks):
predict_dataset = predict_dataset.remove_columns("labels")
predictions = trainer.predict(
predict_dataset, metric_key_prefix="predict"
).predictions
predictions = (
np.squeeze(predictions)
if is_regression
else np.argmax(predictions, axis=1)
)
output_predict_file = os.path.join(
training_args.output_dir, f"predict_results_{task}.txt"
)
if trainer.is_world_process_zero():
with open(output_predict_file, "w") as writer:
logger.info(f"***** Predict results {task} *****")
writer.write("index\tprediction\n")
for index, item in enumerate(predictions):
if is_regression:
writer.write(f"{index}\t{item:3.3f}\n")
else:
item = label_list[item]
writer.write(f"{index}\t{item}\n")
# Saving predictions
output_test_predictions_file = os.path.join(
training_args.output_dir, "test_predictions.txt"
)
logger.warning(predictions)
with open(output_test_predictions_file, "w") as writer:
for x in predictions:
writer.write(str(model.config.id2label[x]) + "\n")
metrics = trainer.evaluate(eval_dataset=predict_dataset)
max_eval_samples = (
data_args.max_eval_samples
if data_args.max_eval_samples is not None
else len(predict_dataset)
)
metrics["eval_samples"] = min(max_eval_samples, len(predict_dataset))
trainer.log_metrics("test", metrics)
trainer.save_metrics("test", metrics)
# Evaluation
if training_args.do_eval:
logger.info("*** Evaluate ***")
# Loop to handle MNLI double evaluation (matched, mis-matched)
tasks = [data_args.task_name]
eval_datasets = [eval_dataset]
if data_args.task_name == "mnli":
tasks.append("mnli-mm")
eval_datasets.append(raw_datasets["validation_mismatched"])
for eval_dataset, task in zip(eval_datasets, tasks):
metrics = trainer.evaluate(eval_dataset=eval_dataset)
max_eval_samples = (
data_args.max_eval_samples
if data_args.max_eval_samples is not None
else len(eval_dataset)
)
metrics["eval_samples"] = min(max_eval_samples, len(eval_dataset))
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
return metrics
def _mp_fn(index):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main(sys.argv[1:])