SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Add MultipleChoice to TFTrainer [WIP] (#4270)

Browse Source 
* catch gpu len 1 set to gpu0

* Add mpc to trainer

* Add MPC for TF

* fix TF automodel for MPC and add Albert

* Apply style

* Fix import

* Note to self: double check

* Make shape None, None for datasetgenerator output shapes

* Add from_pt bool which doesnt seem to work

* Original checkpoint dir

* Fix docstrings for automodel

* Update readme and apply style

* Colab should probably not be from users

* Colabs should probably not be from users

* Add colab

* Update README.md

* Update README.md

* Cleanup __intit__

* Cleanup flake8 trailing comma

* Update src/transformers/training_args_tf.py

* Update src/transformers/modeling_tf_auto.py

Co-authored-by: Viktor Alm <viktoralm@pop-os.localdomain>
Co-authored-by: Julien Chaumond <chaumond@gmail.com>
This commit is contained in:
Viktor Alm
2020-05-12 14:48:48 +02:00
committed by GitHub
parent 65be574aec
commit e4512aab3b
9 changed files with 730 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
examples/multiple-choice/README.md
View File

@@ -29,3 +29,28 @@ Training with the defined hyper-parameters yields the following results:
eval_acc = 0.8338998300509847
eval_loss = 0.44457291918821606
```
## Tensorflow
```bash
export SWAG_DIR=/path/to/swag_data_dir
python ./examples/multiple-choice/run_tf_multiple_choice.py \
--task_name swag \
--model_name_or_path bert-base-cased \
--do_train \
--do_eval \
--data_dir $SWAG_DIR \
--learning_rate 5e-5 \
--num_train_epochs 3 \
--max_seq_length 80 \
--output_dir models_bert/swag_base \
--per_gpu_eval_batch_size=16 \
--per_gpu_train_batch_size=16 \
--logging-dir logs \
--gradient_accumulation_steps 2 \
--overwrite_output
```
# Run it in colab
[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/ViktorAlm/notebooks/blob/master/MPC_GPU_Demo_for_TF_and_PT.ipynb)

211
examples/multiple-choice/run_tf_multiple_choice.py Normal file
View File

@@ -0,0 +1,211 @@
# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
# Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
#
# 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 multiple choice (Bert, Roberta, XLNet)."""
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, Optional
import numpy as np
from transformers import (
AutoConfig,
AutoTokenizer,
EvalPrediction,
HfArgumentParser,
TFAutoModelForMultipleChoice,
TFTrainer,
TFTrainingArguments,
set_seed,
)
from utils_multiple_choice import Split, TFMultipleChoiceDataset, processors
logger = logging.getLogger(__name__)
def simple_accuracy(preds, labels):
return (preds == labels).mean()
@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 s3"}
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
task_name: str = field(metadata={"help": "The name of the task to train on: " + ", ".join(processors.keys())})
data_dir: str = field(metadata={"help": "Should contain the data files for the task."})
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 training and evaluation sets"}
)
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, TFTrainingArguments))
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,
)
logger.warning(
"device: %s, n_gpu: %s, 16-bits training: %s", training_args.device, training_args.n_gpu, training_args.fp16,
)
logger.info("Training/evaluation parameters %s", training_args)
# Set seed
set_seed(training_args.seed)
try:
processor = processors[data_args.task_name]()
label_list = processor.get_labels()
num_labels = len(label_list)
except KeyError:
raise ValueError("Task not found: %s" % (data_args.task_name))
# Load pretrained model and tokenizer
#
# 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,
finetuning_task=data_args.task_name,
cache_dir=model_args.cache_dir,
)
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,
)
with training_args.strategy.scope():
model = TFAutoModelForMultipleChoice.from_pretrained(
model_args.model_name_or_path,
from_pt=bool(".bin" in model_args.model_name_or_path),
config=config,
cache_dir=model_args.cache_dir,
)
# Get datasets
train_dataset = (
TFMultipleChoiceDataset(
data_dir=data_args.data_dir,
tokenizer=tokenizer,
task=data_args.task_name,
max_seq_length=data_args.max_seq_length,
overwrite_cache=data_args.overwrite_cache,
mode=Split.train,
)
if training_args.do_train
else None
)
eval_dataset = (
TFMultipleChoiceDataset(
data_dir=data_args.data_dir,
tokenizer=tokenizer,
task=data_args.task_name,
max_seq_length=data_args.max_seq_length,
overwrite_cache=data_args.overwrite_cache,
mode=Split.dev,
)
if training_args.do_eval
else None
)
def compute_metrics(p: EvalPrediction) -> Dict:
preds = np.argmax(p.predictions, axis=1)
return {"acc": simple_accuracy(preds, p.label_ids)}
# Initialize our Trainer
trainer = TFTrainer(
model=model,
args=training_args,
train_dataset=train_dataset.get_dataset() if train_dataset else None,
eval_dataset=eval_dataset.get_dataset() if eval_dataset else None,
compute_metrics=compute_metrics,
)
# Training
if training_args.do_train:
trainer.train()
trainer.save_model()
tokenizer.save_pretrained(training_args.output_dir)
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
result = trainer.evaluate()
output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in result.items():
logger.info(" %s = %s", key, value)
writer.write("%s = %s\n" % (key, value))
results.update(result)
return results
if __name__ == "__main__":
main()

256
examples/multiple-choice/utils_multiple_choice.py
View File

@@ -25,11 +25,9 @@ from dataclasses import dataclass
from enum import Enum
from typing import List, Optional
import torch
import tqdm
from torch.utils.data.dataset import Dataset
from transformers import PreTrainedTokenizer, torch_distributed_zero_first
from transformers import PreTrainedTokenizer, is_tf_available, is_torch_available
logger = logging.getLogger(__name__)
@@ -76,66 +74,160 @@ class Split(Enum):
test = "test"
class MultipleChoiceDataset(Dataset):
"""
This will be superseded by a framework-agnostic approach
soon.
"""
if is_torch_available():
import torch
from torch.utils.data.dataset import Dataset
from transformers import torch_distributed_zero_first
features: List[InputFeatures]
class MultipleChoiceDataset(Dataset):
"""
This will be superseded by a framework-agnostic approach
soon.
"""
def __init__(
self,
data_dir: str,
tokenizer: PreTrainedTokenizer,
task: str,
max_seq_length: Optional[int] = None,
overwrite_cache=False,
mode: Split = Split.train,
local_rank=-1,
):
processor = processors[task]()
features: List[InputFeatures]
cached_features_file = os.path.join(
data_dir,
"cached_{}_{}_{}_{}".format(mode.value, tokenizer.__class__.__name__, str(max_seq_length), task,),
)
with torch_distributed_zero_first(local_rank):
# Make sure only the first process in distributed training processes the dataset,
# and the others will use the cache.
def __init__(
self,
data_dir: str,
tokenizer: PreTrainedTokenizer,
task: str,
max_seq_length: Optional[int] = None,
overwrite_cache=False,
mode: Split = Split.train,
local_rank=-1,
):
processor = processors[task]()
if os.path.exists(cached_features_file) and not overwrite_cache:
logger.info(f"Loading features from cached file {cached_features_file}")
self.features = torch.load(cached_features_file)
else:
logger.info(f"Creating features from dataset file at {data_dir}")
label_list = processor.get_labels()
if mode == Split.dev:
examples = processor.get_dev_examples(data_dir)
elif mode == Split.test:
examples = processor.get_test_examples(data_dir)
cached_features_file = os.path.join(
data_dir,
"cached_{}_{}_{}_{}".format(mode.value, tokenizer.__class__.__name__, str(max_seq_length), task,),
)
with torch_distributed_zero_first(local_rank):
# Make sure only the first process in distributed training processes the dataset,
# and the others will use the cache.
if os.path.exists(cached_features_file) and not overwrite_cache:
logger.info(f"Loading features from cached file {cached_features_file}")
self.features = torch.load(cached_features_file)
else:
examples = processor.get_train_examples(data_dir)
logger.info("Training examples: %s", len(examples))
# TODO clean up all this to leverage built-in features of tokenizers
self.features = convert_examples_to_features(
examples,
label_list,
max_seq_length,
tokenizer,
pad_on_left=bool(tokenizer.padding_side == "left"),
pad_token=tokenizer.pad_token_id,
pad_token_segment_id=tokenizer.pad_token_type_id,
)
if local_rank in [-1, 0]:
logger.info("Saving features into cached file %s", cached_features_file)
torch.save(self.features, cached_features_file)
logger.info(f"Creating features from dataset file at {data_dir}")
label_list = processor.get_labels()
if mode == Split.dev:
examples = processor.get_dev_examples(data_dir)
elif mode == Split.test:
examples = processor.get_test_examples(data_dir)
else:
examples = processor.get_train_examples(data_dir)
logger.info("Training examples: %s", len(examples))
# TODO clean up all this to leverage built-in features of tokenizers
self.features = convert_examples_to_features(
examples,
label_list,
max_seq_length,
tokenizer,
pad_on_left=bool(tokenizer.padding_side == "left"),
pad_token=tokenizer.pad_token_id,
pad_token_segment_id=tokenizer.pad_token_type_id,
)
if local_rank in [-1, 0]:
logger.info("Saving features into cached file %s", cached_features_file)
torch.save(self.features, cached_features_file)
def __len__(self):
return len(self.features)
def __len__(self):
return len(self.features)
def __getitem__(self, i) -> InputFeatures:
return self.features[i]
def __getitem__(self, i) -> InputFeatures:
return self.features[i]
if is_tf_available():
import tensorflow as tf
class TFMultipleChoiceDataset:
"""
This will be superseded by a framework-agnostic approach
soon.
"""
features: List[InputFeatures]
def __init__(
self,
data_dir: str,
tokenizer: PreTrainedTokenizer,
task: str,
max_seq_length: Optional[int] = 128,
overwrite_cache=False,
mode: Split = Split.train,
):
processor = processors[task]()
logger.info(f"Creating features from dataset file at {data_dir}")
label_list = processor.get_labels()
if mode == Split.dev:
examples = processor.get_dev_examples(data_dir)
elif mode == Split.test:
examples = processor.get_test_examples(data_dir)
else:
examples = processor.get_train_examples(data_dir)
logger.info("Training examples: %s", len(examples))
# TODO clean up all this to leverage built-in features of tokenizers
self.features = convert_examples_to_features(
examples,
label_list,
max_seq_length,
tokenizer,
pad_on_left=bool(tokenizer.padding_side == "left"),
pad_token=tokenizer.pad_token_id,
pad_token_segment_id=tokenizer.pad_token_type_id,
)
def gen():
for (ex_index, ex) in tqdm.tqdm(enumerate(self.features), desc="convert examples to features"):
if ex_index % 10000 == 0:
logger.info("Writing example %d of %d" % (ex_index, len(examples)))
yield (
{
"example_id": 0,
"input_ids": ex.input_ids,
"attention_mask": ex.attention_mask,
"token_type_ids": ex.token_type_ids,
},
ex.label,
)
self.dataset = tf.data.Dataset.from_generator(
gen,
(
{
"example_id": tf.int32,
"input_ids": tf.int32,
"attention_mask": tf.int32,
"token_type_ids": tf.int32,
},
tf.int64,
),
(
{
"example_id": tf.TensorShape([]),
"input_ids": tf.TensorShape([None, None]),
"attention_mask": tf.TensorShape([None, None]),
"token_type_ids": tf.TensorShape([None, None]),
},
tf.TensorShape([]),
),
)
def get_dataset(self):
return self.dataset
def __len__(self):
return len(self.features)
def __getitem__(self, i) -> InputFeatures:
return self.features[i]
class DataProcessor:
@@ -225,6 +317,52 @@ class RaceProcessor(DataProcessor):
return examples
class SynonymProcessor(DataProcessor):
"""Processor for the Synonym data set."""
def get_train_examples(self, data_dir):
"""See base class."""
logger.info("LOOKING AT {} train".format(data_dir))
return self._create_examples(self._read_csv(os.path.join(data_dir, "mctrain.csv")), "train")
def get_dev_examples(self, data_dir):
"""See base class."""
logger.info("LOOKING AT {} dev".format(data_dir))
return self._create_examples(self._read_csv(os.path.join(data_dir, "mchp.csv")), "dev")
def get_test_examples(self, data_dir):
"""See base class."""
logger.info("LOOKING AT {} dev".format(data_dir))
return self._create_examples(self._read_csv(os.path.join(data_dir, "mctest.csv")), "test")
def get_labels(self):
"""See base class."""
return ["0", "1", "2", "3", "4"]
def _read_csv(self, input_file):
with open(input_file, "r", encoding="utf-8") as f:
return list(csv.reader(f))
def _create_examples(self, lines: List[List[str]], type: str):
"""Creates examples for the training and dev sets."""
examples = [
InputExample(
example_id=line[0],
question="", # in the swag dataset, the
# common beginning of each
# choice is stored in "sent2".
contexts=[line[1], line[1], line[1], line[1], line[1]],
endings=[line[2], line[3], line[4], line[5], line[6]],
label=line[7],
)
for line in lines # we skip the line with the column names
]
return examples
class SwagProcessor(DataProcessor):
"""Processor for the SWAG data set."""
@@ -435,7 +573,5 @@ def convert_examples_to_features(
return features
processors = {"race": RaceProcessor, "swag": SwagProcessor, "arc": ArcProcessor}
MULTIPLE_CHOICE_TASKS_NUM_LABELS = {"race", 4, "swag", 4, "arc", 4}
processors = {"race": RaceProcessor, "swag": SwagProcessor, "arc": ArcProcessor, "syn": SynonymProcessor}
MULTIPLE_CHOICE_TASKS_NUM_LABELS = {"race", 4, "swag", 4, "arc", 4, "syn", 5}