repo_id
stringlengths 15
89
| file_path
stringlengths 27
180
| content
stringlengths 1
2.23M
| __index_level_0__
int64 0
0
|
|---|---|---|---|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/summarization/run_summarization_flax.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Fine-tuning the library models for summarization.
"""
# You can also adapt this script on your own sequence to sequence task. Pointers for this are left as comments.
import json
import logging
import math
import os
import sys
import time
import warnings
from dataclasses import asdict, dataclass, field
from enum import Enum
from functools import partial
from pathlib import Path
from typing import Callable, Optional
import datasets
import evaluate
import jax
import jax.numpy as jnp
import nltk # Here to have a nice missing dependency error message early on
import numpy as np
import optax
from datasets import Dataset, load_dataset
from filelock import FileLock
from flax import jax_utils, traverse_util
from flax.jax_utils import pad_shard_unpad, unreplicate
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard, shard_prng_key
from huggingface_hub import Repository, create_repo
from tqdm import tqdm
import transformers
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
AutoConfig,
AutoTokenizer,
FlaxAutoModelForSeq2SeqLM,
HfArgumentParser,
is_tensorboard_available,
)
from transformers.utils import is_offline_mode, send_example_telemetry
logger = logging.getLogger(__name__)
try:
nltk.data.find("tokenizers/punkt")
except (LookupError, OSError):
if is_offline_mode():
raise LookupError(
"Offline mode: run this script without TRANSFORMERS_OFFLINE first to download nltk data files"
)
with FileLock(".lock") as lock:
nltk.download("punkt", quiet=True)
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class TrainingArguments:
output_dir: str = field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."},
)
overwrite_output_dir: bool = field(
default=False,
metadata={
"help": (
"Overwrite the content of the output directory. "
"Use this to continue training if output_dir points to a checkpoint directory."
)
},
)
do_train: bool = field(default=False, metadata={"help": "Whether to run training."})
do_eval: bool = field(default=False, metadata={"help": "Whether to run eval on the dev set."})
do_predict: bool = field(default=False, metadata={"help": "Whether to run predictions on the test set."})
per_device_train_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for training."}
)
per_device_eval_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for evaluation."}
)
learning_rate: float = field(default=5e-5, metadata={"help": "The initial learning rate for AdamW."})
weight_decay: float = field(default=0.0, metadata={"help": "Weight decay for AdamW if we apply some."})
adam_beta1: float = field(default=0.9, metadata={"help": "Beta1 for AdamW optimizer"})
adam_beta2: float = field(default=0.999, metadata={"help": "Beta2 for AdamW optimizer"})
adam_epsilon: float = field(default=1e-8, metadata={"help": "Epsilon for AdamW optimizer."})
label_smoothing_factor: float = field(
default=0.0, metadata={"help": "The label smoothing epsilon to apply (zero means no label smoothing)."}
)
adafactor: bool = field(default=False, metadata={"help": "Whether or not to replace AdamW by Adafactor."})
num_train_epochs: float = field(default=3.0, metadata={"help": "Total number of training epochs to perform."})
warmup_steps: int = field(default=0, metadata={"help": "Linear warmup over warmup_steps."})
logging_steps: int = field(default=500, metadata={"help": "Log every X updates steps."})
save_steps: int = field(default=500, metadata={"help": "Save checkpoint every X updates steps."})
eval_steps: int = field(default=None, metadata={"help": "Run an evaluation every X steps."})
seed: int = field(default=42, metadata={"help": "Random seed that will be set at the beginning of training."})
push_to_hub: bool = field(
default=False, metadata={"help": "Whether or not to upload the trained model to the model hub after training."}
)
hub_model_id: str = field(
default=None, metadata={"help": "The name of the repository to keep in sync with the local `output_dir`."}
)
hub_token: str = field(default=None, metadata={"help": "The token to use to push to the Model Hub."})
gradient_checkpointing: bool = field(
default=False,
metadata={
"help": "If True, use gradient checkpointing to save memory at the expense of slower backward pass."
},
)
def __post_init__(self):
if self.output_dir is not None:
self.output_dir = os.path.expanduser(self.output_dir)
def to_dict(self):
"""
Serializes this instance while replace `Enum` by their values (for JSON serialization support). It obfuscates
the token values by removing their value.
"""
d = asdict(self)
for k, v in d.items():
if isinstance(v, Enum):
d[k] = v.value
if isinstance(v, list) and len(v) > 0 and isinstance(v[0], Enum):
d[k] = [x.value for x in v]
if k.endswith("_token"):
d[k] = f"<{k.upper()}>"
return d
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
)
},
)
use_auth_token: bool = field(
default=None,
metadata={
"help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
},
)
trust_remote_code: bool = field(
default=False,
metadata={
"help": (
"Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
"should only be set to `True` for repositories you trust and in which you have read the code, as it will "
"execute code present on the Hub on your local machine."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
text_column: Optional[str] = field(
default=None,
metadata={"help": "The name of the column in the datasets containing the full texts (for summarization)."},
)
summary_column: Optional[str] = field(
default=None,
metadata={"help": "The name of the column in the datasets containing the summaries (for summarization)."},
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
test_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input predict data file to do prediction on (a text file)."},
)
max_source_length: Optional[int] = field(
default=1024,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
max_target_length: Optional[int] = field(
default=128,
metadata={
"help": (
"The maximum total sequence length for target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
val_max_target_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total sequence length for validation target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
"This argument is also used to override the `max_length` param of `model.generate`, which is used "
"during evaluation."
)
},
)
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."
)
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
source_prefix: Optional[str] = field(
default=None, metadata={"help": "A prefix to add before every source text (useful for T5 models)."}
)
predict_with_generate: bool = field(
default=False, metadata={"help": "Whether to use generate to calculate generative metrics (ROUGE, BLEU)."}
)
num_beams: Optional[int] = field(
default=1,
metadata={
"help": (
"Number of beams to use for evaluation. This argument will be passed to `model.generate`, "
"which is used during evaluation."
)
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
def __post_init__(self):
if (
self.dataset_name is None
and self.train_file is None
and self.validation_file is None
and self.test_file is None
):
raise ValueError("Need either a dataset name or a training, validation, or test file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
if self.test_file is not None:
extension = self.test_file.split(".")[-1]
assert extension in ["csv", "json"], "`test_file` should be a csv or a json file."
if self.val_max_target_length is None:
self.val_max_target_length = self.max_target_length
summarization_name_mapping = {
"amazon_reviews_multi": ("review_body", "review_title"),
"big_patent": ("description", "abstract"),
"cnn_dailymail": ("article", "highlights"),
"orange_sum": ("text", "summary"),
"pn_summary": ("article", "summary"),
"psc": ("extract_text", "summary_text"),
"samsum": ("dialogue", "summary"),
"thaisum": ("body", "summary"),
"xglue": ("news_body", "news_title"),
"xsum": ("document", "summary"),
"wiki_summary": ("article", "highlights"),
}
class TrainState(train_state.TrainState):
dropout_rng: jnp.ndarray
def replicate(self):
return jax_utils.replicate(self).replace(dropout_rng=shard_prng_key(self.dropout_rng))
def data_loader(rng: jax.random.PRNGKey, dataset: Dataset, batch_size: int, shuffle: bool = False, drop_last=True):
"""
Returns batches of size `batch_size` from `dataset`. If `drop_last` is set to `False`, the final batch may be incomplete,
and range in size from 1 to `batch_size`. Shuffle batches if `shuffle` is `True`.
"""
if shuffle:
batch_idx = jax.random.permutation(rng, len(dataset))
batch_idx = np.asarray(batch_idx)
else:
batch_idx = np.arange(len(dataset))
if drop_last:
steps_per_epoch = len(dataset) // batch_size
batch_idx = batch_idx[: steps_per_epoch * batch_size] # Skip incomplete batch.
batch_idx = batch_idx.reshape((steps_per_epoch, batch_size))
else:
steps_per_epoch = math.ceil(len(dataset) / batch_size)
batch_idx = np.array_split(batch_idx, steps_per_epoch)
for idx in batch_idx:
batch = dataset[idx]
batch = {k: np.array(v) for k, v in batch.items()}
yield batch
def write_metric(summary_writer, train_metrics, eval_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
def create_learning_rate_fn(
train_ds_size: int, train_batch_size: int, num_train_epochs: int, num_warmup_steps: int, learning_rate: float
) -> Callable[[int], jnp.ndarray]:
"""Returns a linear warmup, linear_decay learning rate function."""
steps_per_epoch = train_ds_size // train_batch_size
num_train_steps = steps_per_epoch * num_train_epochs
warmup_fn = optax.linear_schedule(init_value=0.0, end_value=learning_rate, transition_steps=num_warmup_steps)
decay_fn = optax.linear_schedule(
init_value=learning_rate, end_value=0, transition_steps=num_train_steps - num_warmup_steps
)
schedule_fn = optax.join_schedules(schedules=[warmup_fn, decay_fn], boundaries=[num_warmup_steps])
return schedule_fn
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 model_args.use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
FutureWarning,
)
if model_args.token is not None:
raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
model_args.token = model_args.use_auth_token
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_summarization", model_args, data_args, framework="flax")
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."
)
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
# Setup logging, we only want one process per machine to log things on the screen.
logger.setLevel(logging.INFO if jax.process_index() == 0 else logging.ERROR)
if jax.process_index() == 0:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
# Handle the repository creation
if training_args.push_to_hub:
# Retrieve of infer repo_name
repo_name = training_args.hub_model_id
if repo_name is None:
repo_name = Path(training_args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=training_args.hub_token).repo_id
# Clone repo locally
repo = Repository(training_args.output_dir, clone_from=repo_id, token=training_args.hub_token)
# Get the datasets: you can either provide your own CSV/JSON 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 first column for the full texts and the second column for the
# summaries (unless you specify column names for this with the `text_column` and `summary_column` arguments).
#
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
dataset = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
cache_dir=model_args.cache_dir,
keep_in_memory=False,
token=model_args.token,
)
else:
data_files = {}
if data_args.train_file is not None:
data_files["train"] = data_args.train_file
extension = data_args.train_file.split(".")[-1]
if data_args.validation_file is not None:
data_files["validation"] = data_args.validation_file
extension = data_args.validation_file.split(".")[-1]
if data_args.test_file is not None:
data_files["test"] = data_args.test_file
extension = data_args.test_file.split(".")[-1]
dataset = load_dataset(
extension,
data_files=data_files,
cache_dir=model_args.cache_dir,
token=model_args.token,
)
# 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.
# Load pretrained model and tokenizer
if model_args.config_name:
config = AutoConfig.from_pretrained(
model_args.config_name,
cache_dir=model_args.cache_dir,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
elif model_args.model_name_or_path:
config = AutoConfig.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
config = CONFIG_MAPPING[model_args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
if model_args.model_name_or_path:
model = FlaxAutoModelForSeq2SeqLM.from_pretrained(
model_args.model_name_or_path,
config=config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
model = FlaxAutoModelForSeq2SeqLM.from_config(
config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
trust_remote_code=model_args.trust_remote_code,
)
if training_args.gradient_checkpointing:
model.enable_gradient_checkpointing()
if model.config.decoder_start_token_id is None:
raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")
prefix = data_args.source_prefix if data_args.source_prefix is not None else ""
# Preprocessing the datasets.
# We need to tokenize inputs and targets.
if training_args.do_train:
if "train" not in dataset:
raise ValueError("--do_train requires a train dataset")
column_names = dataset["train"].column_names
elif training_args.do_eval:
if "validation" not in dataset:
raise ValueError("--do_eval requires a validation dataset")
column_names = dataset["validation"].column_names
elif training_args.do_predict:
if "test" not in dataset:
raise ValueError("--do_predict requires a test dataset")
column_names = dataset["test"].column_names
else:
logger.info("There is nothing to do. Please pass `do_train`, `do_eval` and/or `do_predict`.")
return
# Get the column names for input/target.
dataset_columns = summarization_name_mapping.get(data_args.dataset_name, None)
if data_args.text_column is None:
text_column = dataset_columns[0] if dataset_columns is not None else column_names[0]
else:
text_column = data_args.text_column
if text_column not in column_names:
raise ValueError(
f"--text_column' value '{data_args.text_column}' needs to be one of: {', '.join(column_names)}"
)
if data_args.summary_column is None:
summary_column = dataset_columns[1] if dataset_columns is not None else column_names[1]
else:
summary_column = data_args.summary_column
if summary_column not in column_names:
raise ValueError(
f"--summary_column' value '{data_args.summary_column}' needs to be one of: {', '.join(column_names)}"
)
# Temporarily set max_target_length for training.
max_target_length = data_args.max_target_length
# In Flax, for seq2seq models we need to pass `decoder_input_ids`
# as the Flax models don't accept `labels`, we need to prepare the decoder_input_ids here
# for that dynamically import the `shift_tokens_right` function from the model file
model_module = __import__(model.__module__, fromlist=["shift_tokens_tight"])
shift_tokens_right_fn = getattr(model_module, "shift_tokens_right")
# Setting padding="max_length" as we need fixed length inputs for jitted functions
def preprocess_function(examples):
inputs = examples[text_column]
targets = examples[summary_column]
inputs = [prefix + inp for inp in inputs]
model_inputs = tokenizer(
inputs, max_length=data_args.max_source_length, padding="max_length", truncation=True, return_tensors="np"
)
# Setup the tokenizer for targets
labels = tokenizer(
text_target=targets,
max_length=max_target_length,
padding="max_length",
truncation=True,
return_tensors="np",
)
model_inputs["labels"] = labels["input_ids"]
decoder_input_ids = shift_tokens_right_fn(
labels["input_ids"], config.pad_token_id, config.decoder_start_token_id
)
model_inputs["decoder_input_ids"] = np.asarray(decoder_input_ids)
# We need decoder_attention_mask so we can ignore pad tokens from loss
model_inputs["decoder_attention_mask"] = labels["attention_mask"]
return model_inputs
if training_args.do_train:
train_dataset = dataset["train"]
if data_args.max_train_samples is not None:
max_train_samples = min(len(train_dataset), data_args.max_train_samples)
train_dataset = train_dataset.select(range(max_train_samples))
train_dataset = train_dataset.map(
preprocess_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
desc="Running tokenizer on train dataset",
)
if training_args.do_eval:
max_target_length = data_args.val_max_target_length
eval_dataset = dataset["validation"]
if data_args.max_eval_samples is not None:
max_eval_samples = min(len(eval_dataset), data_args.max_eval_samples)
eval_dataset = eval_dataset.select(range(max_eval_samples))
eval_dataset = eval_dataset.map(
preprocess_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
desc="Running tokenizer on validation dataset",
)
if training_args.do_predict:
max_target_length = data_args.val_max_target_length
predict_dataset = dataset["test"]
if data_args.max_predict_samples is not None:
max_predict_samples = min(len(predict_dataset), data_args.max_predict_samples)
predict_dataset = predict_dataset.select(range(max_predict_samples))
predict_dataset = predict_dataset.map(
preprocess_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
desc="Running tokenizer on prediction dataset",
)
# Metric
metric = evaluate.load("rouge")
def postprocess_text(preds, labels):
preds = [pred.strip() for pred in preds]
labels = [label.strip() for label in labels]
# rougeLSum expects newline after each sentence
preds = ["\n".join(nltk.sent_tokenize(pred)) for pred in preds]
labels = ["\n".join(nltk.sent_tokenize(label)) for label in labels]
return preds, labels
def compute_metrics(preds, labels):
decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True)
decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)
# Some simple post-processing
decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels)
result = metric.compute(predictions=decoded_preds, references=decoded_labels, use_stemmer=True)
result = {k: round(v * 100, 4) for k, v in result.items()}
prediction_lens = [np.count_nonzero(pred != tokenizer.pad_token_id) for pred in preds]
result["gen_len"] = np.mean(prediction_lens)
return result
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
rng, dropout_rng = jax.random.split(rng)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
eval_batch_size = per_device_eval_batch_size * jax.device_count()
steps_per_epoch = len(train_dataset) // train_batch_size
total_train_steps = steps_per_epoch * num_epochs
# Create learning rate schedule
linear_decay_lr_schedule_fn = create_learning_rate_fn(
len(train_dataset),
train_batch_size,
training_args.num_train_epochs,
training_args.warmup_steps,
training_args.learning_rate,
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
# find out all LayerNorm parameters
layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
layer_norm_named_params = {
layer[-2:]
for layer_norm_name in layer_norm_candidates
for layer in flat_params.keys()
if layer_norm_name in "".join(layer).lower()
}
flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
adamw = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state
state = TrainState.create(apply_fn=model.__call__, params=model.params, tx=adamw, dropout_rng=dropout_rng)
# label smoothed cross entropy
def loss_fn(logits, labels, padding_mask, label_smoothing_factor=0.0):
"""
The label smoothing implementation is adapted from Flax's official example:
https://github.com/google/flax/blob/87a211135c6a377c8f29048a1cac3840e38b9da4/examples/wmt/train.py#L104
"""
vocab_size = logits.shape[-1]
confidence = 1.0 - label_smoothing_factor
low_confidence = (1.0 - confidence) / (vocab_size - 1)
normalizing_constant = -(
confidence * jnp.log(confidence) + (vocab_size - 1) * low_confidence * jnp.log(low_confidence + 1e-20)
)
soft_labels = onehot(labels, vocab_size, on_value=confidence, off_value=low_confidence)
loss = optax.softmax_cross_entropy(logits, soft_labels)
loss = loss - normalizing_constant
# ignore padded tokens from loss
loss = loss * padding_mask
loss = loss.sum()
num_labels = padding_mask.sum()
return loss, num_labels
# Define gradient update step fn
def train_step(state, batch, label_smoothing_factor=0.0):
dropout_rng, new_dropout_rng = jax.random.split(state.dropout_rng)
def compute_loss(params):
labels = batch.pop("labels")
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
loss, num_labels = loss_fn(logits, labels, batch["decoder_attention_mask"], label_smoothing_factor)
return loss, num_labels
grad_fn = jax.value_and_grad(compute_loss, has_aux=True)
(loss, num_labels), grad = grad_fn(state.params)
num_labels = jax.lax.psum(num_labels, "batch")
# true loss = total loss / total samples
loss = jax.lax.psum(loss, "batch")
loss = jax.tree_util.tree_map(lambda x: x / num_labels, loss)
# true grad = total grad / total samples
grad = jax.lax.psum(grad, "batch")
grad = jax.tree_util.tree_map(lambda x: x / num_labels, grad)
new_state = state.apply_gradients(grads=grad, dropout_rng=new_dropout_rng)
metrics = {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}
return new_state, metrics
# Define eval fn
def eval_step(params, batch, label_smoothing_factor=0.0):
labels = batch.pop("labels")
logits = model(**batch, params=params, train=False)[0]
loss, num_labels = loss_fn(logits, labels, batch["decoder_attention_mask"], label_smoothing_factor)
num_labels = jax.lax.psum(num_labels, "batch")
# true loss = total loss / total samples
loss = jax.lax.psum(loss, "batch")
loss = jax.tree_util.tree_map(lambda x: x / num_labels, loss)
metrics = {"loss": loss}
return metrics
# Define generation function
max_length = (
data_args.val_max_target_length if data_args.val_max_target_length is not None else model.config.max_length
)
num_beams = data_args.num_beams if data_args.num_beams is not None else model.config.num_beams
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def generate_step(params, batch):
model.params = params
output_ids = model.generate(batch["input_ids"], attention_mask=batch["attention_mask"], **gen_kwargs)
return output_ids.sequences
# Create parallel version of the train and eval step
p_train_step = jax.pmap(
partial(train_step, label_smoothing_factor=training_args.label_smoothing_factor), "batch", donate_argnums=(0,)
)
p_eval_step = jax.pmap(partial(eval_step, label_smoothing_factor=training_args.label_smoothing_factor), "batch")
p_generate_step = jax.pmap(generate_step, "batch")
# Replicate the train state on each device
state = state.replicate()
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {num_epochs}")
logger.info(f" Instantaneous batch size per device = {training_args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel & distributed) = {train_batch_size}")
logger.info(f" Total optimization steps = {total_train_steps}")
train_time = 0
epochs = tqdm(range(num_epochs), desc=f"Epoch ... (1/{num_epochs})", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
# Create sampling rng
rng, input_rng = jax.random.split(rng)
train_metrics = []
# Generate an epoch by shuffling sampling indices from the train dataset
train_loader = data_loader(input_rng, train_dataset, train_batch_size, shuffle=True)
steps_per_epoch = len(train_dataset) // train_batch_size
# train
for _ in tqdm(range(steps_per_epoch), desc="Training...", position=1, leave=False):
batch = next(train_loader)
batch = shard(batch)
state, train_metric = p_train_step(state, batch)
train_metrics.append(train_metric)
train_time += time.time() - train_start
train_metric = unreplicate(train_metric)
epochs.write(
f"Epoch... ({epoch + 1}/{num_epochs} | Loss: {train_metric['loss']}, Learning Rate:"
f" {train_metric['learning_rate']})"
)
# ======================== Evaluating ==============================
eval_metrics = []
eval_preds = []
eval_labels = []
eval_loader = data_loader(input_rng, eval_dataset, eval_batch_size, drop_last=False)
eval_steps = math.ceil(len(eval_dataset) / eval_batch_size)
for _ in tqdm(range(eval_steps), desc="Evaluating...", position=2, leave=False):
# Model forward
batch = next(eval_loader)
labels = batch["labels"]
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, batch, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# generation
if data_args.predict_with_generate:
generated_ids = pad_shard_unpad(p_generate_step)(state.params, batch)
eval_preds.extend(jax.device_get(generated_ids.reshape(-1, gen_kwargs["max_length"])))
eval_labels.extend(labels)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.mean, eval_metrics)
# compute ROUGE metrics
rouge_desc = ""
if data_args.predict_with_generate:
rouge_metrics = compute_metrics(eval_preds, eval_labels)
eval_metrics.update(rouge_metrics)
rouge_desc = " ".join([f"Eval {key}: {value} |" for key, value in rouge_metrics.items()])
# Print metrics and update progress bar
desc = f"Epoch... ({epoch + 1}/{num_epochs} | Eval Loss: {eval_metrics['loss']} | {rouge_desc})"
epochs.write(desc)
epochs.desc = desc
# Save metrics
if has_tensorboard and jax.process_index() == 0:
cur_step = epoch * (len(train_dataset) // train_batch_size)
write_metric(summary_writer, train_metrics, eval_metrics, train_time, cur_step)
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(jax.tree_util.tree_map(lambda x: x[0], state.params))
model.save_pretrained(training_args.output_dir, params=params)
tokenizer.save_pretrained(training_args.output_dir)
if training_args.push_to_hub:
repo.push_to_hub(commit_message=f"Saving weights and logs of epoch {epoch}", blocking=False)
# ======================== Prediction loop ==============================
if training_args.do_predict:
logger.info("*** Predict ***")
pred_metrics = []
pred_generations = []
pred_labels = []
pred_loader = data_loader(input_rng, predict_dataset, eval_batch_size, drop_last=False)
pred_steps = math.ceil(len(predict_dataset) / eval_batch_size)
for _ in tqdm(range(pred_steps), desc="Predicting...", position=2, leave=False):
# Model forward
batch = next(pred_loader)
labels = batch["labels"]
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, batch, min_device_batch=per_device_eval_batch_size
)
pred_metrics.append(metrics)
# generation
if data_args.predict_with_generate:
generated_ids = pad_shard_unpad(p_generate_step)(state.params, batch)
pred_generations.extend(jax.device_get(generated_ids.reshape(-1, gen_kwargs["max_length"])))
pred_labels.extend(labels)
# normalize prediction metrics
pred_metrics = get_metrics(pred_metrics)
pred_metrics = jax.tree_util.tree_map(jnp.mean, pred_metrics)
# compute ROUGE metrics
rouge_desc = ""
if data_args.predict_with_generate:
rouge_metrics = compute_metrics(pred_generations, pred_labels)
pred_metrics.update(rouge_metrics)
rouge_desc = " ".join([f"Predict {key}: {value} |" for key, value in rouge_metrics.items()])
# Print metrics
desc = f"Predict Loss: {pred_metrics['loss']} | {rouge_desc})"
logger.info(desc)
# save final metrics in json
if jax.process_index() == 0:
rouge_metrics = {f"test_{metric_name}": value for metric_name, value in rouge_metrics.items()}
path = os.path.join(training_args.output_dir, "test_results.json")
with open(path, "w") as f:
json.dump(rouge_metrics, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/summarization/requirements.txt
|
datasets >= 1.1.3
jax>=0.2.8
jaxlib>=0.1.59
flax>=0.3.5
optax>=0.0.8
evaluate>=0.2.0
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/text-classification/run_flax_glue.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Inc. team. 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 a 🤗 Flax Transformers model for sequence classification on GLUE."""
import json
import logging
import math
import os
import random
import sys
import time
import warnings
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any, Callable, Dict, Optional, Tuple
import datasets
import evaluate
import jax
import jax.numpy as jnp
import numpy as np
import optax
from datasets import load_dataset
from flax import struct, traverse_util
from flax.jax_utils import pad_shard_unpad, replicate, unreplicate
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard
from huggingface_hub import Repository, create_repo
from tqdm import tqdm
import transformers
from transformers import (
AutoConfig,
AutoTokenizer,
FlaxAutoModelForSequenceClassification,
HfArgumentParser,
PretrainedConfig,
TrainingArguments,
is_tensorboard_available,
)
from transformers.utils import check_min_version, send_example_telemetry
logger = logging.getLogger(__name__)
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.36.0.dev0")
Array = Any
Dataset = datasets.arrow_dataset.Dataset
PRNGKey = Any
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"),
}
@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"}
)
use_slow_tokenizer: Optional[bool] = field(
default=False,
metadata={"help": "If passed, will use a slow tokenizer (not backed by the 🤗 Tokenizers library)."},
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
)
model_revision: str = field(
default="main",
metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
)
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
)
},
)
use_auth_token: bool = field(
default=None,
metadata={
"help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
},
)
trust_remote_code: bool = field(
default=False,
metadata={
"help": (
"Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
"should only be set to `True` for repositories you trust and in which you have read the code, as it will "
"execute code present on the Hub on your local machine."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
task_name: Optional[str] = field(
default=None, metadata={"help": f"The name of the glue task to train on. choices {list(task_to_keys.keys())}"}
)
dataset_config_name: Optional[str] = field(
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
train_file: Optional[str] = field(
default=None, metadata={"help": "The input training data file (a csv or JSON file)."}
)
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate on (a csv or JSON file)."},
)
test_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input test data file to predict on (a csv or JSON file)."},
)
text_column_name: Optional[str] = field(
default=None, metadata={"help": "The column name of text to input in the file (a csv or JSON file)."}
)
label_column_name: Optional[str] = field(
default=None, metadata={"help": "The column name of label to input in the file (a csv or JSON file)."}
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
max_seq_length: int = field(
default=None,
metadata={
"help": (
"The maximum total input sequence length after tokenization. If set, sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
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."
)
},
)
def __post_init__(self):
if self.task_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
self.task_name = self.task_name.lower() if isinstance(self.task_name, str) else self.task_name
def create_train_state(
model: FlaxAutoModelForSequenceClassification,
learning_rate_fn: Callable[[int], float],
is_regression: bool,
num_labels: int,
weight_decay: float,
) -> train_state.TrainState:
"""Create initial training state."""
class TrainState(train_state.TrainState):
"""Train state with an Optax optimizer.
The two functions below differ depending on whether the task is classification
or regression.
Args:
logits_fn: Applied to last layer to obtain the logits.
loss_fn: Function to compute the loss.
"""
logits_fn: Callable = struct.field(pytree_node=False)
loss_fn: Callable = struct.field(pytree_node=False)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
# find out all LayerNorm parameters
layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
layer_norm_named_params = {
layer[-2:]
for layer_norm_name in layer_norm_candidates
for layer in flat_params.keys()
if layer_norm_name in "".join(layer).lower()
}
flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
tx = optax.adamw(
learning_rate=learning_rate_fn, b1=0.9, b2=0.999, eps=1e-6, weight_decay=weight_decay, mask=decay_mask_fn
)
if is_regression:
def mse_loss(logits, labels):
return jnp.mean((logits[..., 0] - labels) ** 2)
return TrainState.create(
apply_fn=model.__call__,
params=model.params,
tx=tx,
logits_fn=lambda logits: logits[..., 0],
loss_fn=mse_loss,
)
else: # Classification.
def cross_entropy_loss(logits, labels):
xentropy = optax.softmax_cross_entropy(logits, onehot(labels, num_classes=num_labels))
return jnp.mean(xentropy)
return TrainState.create(
apply_fn=model.__call__,
params=model.params,
tx=tx,
logits_fn=lambda logits: logits.argmax(-1),
loss_fn=cross_entropy_loss,
)
def create_learning_rate_fn(
train_ds_size: int, train_batch_size: int, num_train_epochs: int, num_warmup_steps: int, learning_rate: float
) -> Callable[[int], jnp.ndarray]:
"""Returns a linear warmup, linear_decay learning rate function."""
steps_per_epoch = train_ds_size // train_batch_size
num_train_steps = steps_per_epoch * num_train_epochs
warmup_fn = optax.linear_schedule(init_value=0.0, end_value=learning_rate, transition_steps=num_warmup_steps)
decay_fn = optax.linear_schedule(
init_value=learning_rate, end_value=0, transition_steps=num_train_steps - num_warmup_steps
)
schedule_fn = optax.join_schedules(schedules=[warmup_fn, decay_fn], boundaries=[num_warmup_steps])
return schedule_fn
def glue_train_data_collator(rng: PRNGKey, dataset: Dataset, batch_size: int):
"""Returns shuffled batches of size `batch_size` from truncated `train dataset`, sharded over all local devices."""
steps_per_epoch = len(dataset) // batch_size
perms = jax.random.permutation(rng, len(dataset))
perms = perms[: steps_per_epoch * batch_size] # Skip incomplete batch.
perms = perms.reshape((steps_per_epoch, batch_size))
for perm in perms:
batch = dataset[perm]
batch = {k: np.array(v) for k, v in batch.items()}
batch = shard(batch)
yield batch
def glue_eval_data_collator(dataset: Dataset, batch_size: int):
"""Returns batches of size `batch_size` from `eval dataset`. Sharding handled by `pad_shard_unpad` in the eval loop."""
batch_idx = np.arange(len(dataset))
steps_per_epoch = math.ceil(len(dataset) / batch_size)
batch_idx = np.array_split(batch_idx, steps_per_epoch)
for idx in batch_idx:
batch = dataset[idx]
batch = {k: np.array(v) for k, v in batch.items()}
yield batch
def main():
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 model_args.use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
FutureWarning,
)
if model_args.token is not None:
raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
model_args.token = model_args.use_auth_token
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_glue", model_args, data_args, framework="flax")
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
# Setup logging, we only want one process per machine to log things on the screen.
logger.setLevel(logging.INFO if jax.process_index() == 0 else logging.ERROR)
if jax.process_index() == 0:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# Handle the repository creation
if training_args.push_to_hub:
# Retrieve of infer repo_name
repo_name = training_args.hub_model_id
if repo_name is None:
repo_name = Path(training_args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=training_args.hub_token).repo_id
# Clone repo locally
repo = Repository(training_args.output_dir, clone_from=repo_id, token=training_args.hub_token)
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
# For CSV/JSON files, this script will use as labels the column called 'label' and as pair of sentences the
# sentences in columns called 'sentence1' and 'sentence2' if such column exists or the first two columns not named
# label if at least two columns are provided.
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single 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.task_name is not None:
# Downloading and loading a dataset from the hub.
raw_datasets = load_dataset(
"glue",
data_args.task_name,
token=model_args.token,
)
else:
# Loading the dataset from local csv or json file.
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 if data_args.train_file is not None else data_args.valid_file).split(".")[-1]
raw_datasets = load_dataset(
extension,
data_files=data_files,
token=model_args.token,
)
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.
# 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["label"].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["label"].dtype in ["float32", "float64"]
if is_regression:
num_labels = 1
else:
# A useful fast method:
# https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.unique
label_list = raw_datasets["train"].unique("label")
label_list.sort() # Let's sort it for determinism
num_labels = len(label_list)
# Load pretrained model and tokenizer
config = AutoConfig.from_pretrained(
model_args.model_name_or_path,
num_labels=num_labels,
finetuning_task=data_args.task_name,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
use_fast=not model_args.use_slow_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
model = FlaxAutoModelForSequenceClassification.from_pretrained(
model_args.model_name_or_path,
config=config,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
# Preprocessing the datasets
if data_args.task_name is not None:
sentence1_key, sentence2_key = task_to_keys[data_args.task_name]
else:
# Again, we try to have some nice defaults but don't hesitate to tweak to your use case.
non_label_column_names = [name for name in raw_datasets["train"].column_names if name != "label"]
if "sentence1" in non_label_column_names and "sentence2" in non_label_column_names:
sentence1_key, sentence2_key = "sentence1", "sentence2"
else:
if len(non_label_column_names) >= 2:
sentence1_key, sentence2_key = non_label_column_names[:2]
else:
sentence1_key, sentence2_key = non_label_column_names[0], None
# Some models have set the order of the labels to use, so let's make sure we do use it.
label_to_id = None
if (
model.config.label2id != PretrainedConfig(num_labels=num_labels).label2id
and data_args.task_name is not None
and not is_regression
):
# Some have all caps in their config, some don't.
label_name_to_id = {k.lower(): v for k, v in model.config.label2id.items()}
if sorted(label_name_to_id.keys()) == sorted(label_list):
logger.info(
f"The configuration of the model provided the following label correspondence: {label_name_to_id}. "
"Using it!"
)
label_to_id = {i: label_name_to_id[label_list[i]] for i in range(num_labels)}
else:
logger.warning(
"Your model seems to have been trained with labels, but they don't match the dataset: ",
f"model labels: {sorted(label_name_to_id.keys())}, dataset labels: {sorted(label_list)}."
"\nIgnoring the model labels as a result.",
)
elif data_args.task_name is None:
label_to_id = {v: i for i, v in enumerate(label_list)}
def preprocess_function(examples):
# Tokenize the texts
texts = (
(examples[sentence1_key],) if sentence2_key is None else (examples[sentence1_key], examples[sentence2_key])
)
result = tokenizer(*texts, padding="max_length", max_length=data_args.max_seq_length, truncation=True)
if "label" in examples:
if label_to_id is not None:
# Map labels to IDs (not necessary for GLUE tasks)
result["labels"] = [label_to_id[l] for l in examples["label"]]
else:
# In all cases, rename the column to labels because the model will expect that.
result["labels"] = examples["label"]
return result
processed_datasets = raw_datasets.map(
preprocess_function, batched=True, remove_columns=raw_datasets["train"].column_names
)
train_dataset = processed_datasets["train"]
eval_dataset = processed_datasets["validation_matched" if data_args.task_name == "mnli" else "validation"]
# Log a few random samples from the training set:
for index in random.sample(range(len(train_dataset)), 3):
logger.info(f"Sample {index} of the training set: {train_dataset[index]}.")
# Define a summary writer
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(training_args.output_dir)
summary_writer.hparams({**training_args.to_dict(), **vars(model_args), **vars(data_args)})
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
num_epochs = int(training_args.num_train_epochs)
rng = jax.random.PRNGKey(training_args.seed)
dropout_rngs = jax.random.split(rng, jax.local_device_count())
train_batch_size = int(training_args.per_device_train_batch_size) * jax.local_device_count()
per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
eval_batch_size = per_device_eval_batch_size * jax.device_count()
learning_rate_fn = create_learning_rate_fn(
len(train_dataset),
train_batch_size,
training_args.num_train_epochs,
training_args.warmup_steps,
training_args.learning_rate,
)
state = create_train_state(
model, learning_rate_fn, is_regression, num_labels=num_labels, weight_decay=training_args.weight_decay
)
# define step functions
def train_step(
state: train_state.TrainState, batch: Dict[str, Array], dropout_rng: PRNGKey
) -> Tuple[train_state.TrainState, float]:
"""Trains model with an optimizer (both in `state`) on `batch`, returning a pair `(new_state, loss)`."""
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
targets = batch.pop("labels")
def loss_fn(params):
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
loss = state.loss_fn(logits, targets)
return loss
grad_fn = jax.value_and_grad(loss_fn)
loss, grad = grad_fn(state.params)
grad = jax.lax.pmean(grad, "batch")
new_state = state.apply_gradients(grads=grad)
metrics = jax.lax.pmean({"loss": loss, "learning_rate": learning_rate_fn(state.step)}, axis_name="batch")
return new_state, metrics, new_dropout_rng
p_train_step = jax.pmap(train_step, axis_name="batch", donate_argnums=(0,))
def eval_step(state, batch):
logits = state.apply_fn(**batch, params=state.params, train=False)[0]
return state.logits_fn(logits)
p_eval_step = jax.pmap(eval_step, axis_name="batch")
if data_args.task_name is not None:
metric = evaluate.load("glue", data_args.task_name)
else:
metric = evaluate.load("accuracy")
logger.info(f"===== Starting training ({num_epochs} epochs) =====")
train_time = 0
# make sure weights are replicated on each device
state = replicate(state)
steps_per_epoch = len(train_dataset) // train_batch_size
total_steps = steps_per_epoch * num_epochs
epochs = tqdm(range(num_epochs), desc=f"Epoch ... (0/{num_epochs})", position=0)
for epoch in epochs:
train_start = time.time()
train_metrics = []
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# train
train_loader = glue_train_data_collator(input_rng, train_dataset, train_batch_size)
for step, batch in enumerate(
tqdm(
train_loader,
total=steps_per_epoch,
desc="Training...",
position=1,
),
):
state, train_metric, dropout_rngs = p_train_step(state, batch, dropout_rngs)
train_metrics.append(train_metric)
cur_step = (epoch * steps_per_epoch) + (step + 1)
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_metric = unreplicate(train_metric)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step}/{total_steps} | Training Loss: {train_metric['loss']}, Learning Rate:"
f" {train_metric['learning_rate']})"
)
train_metrics = []
if (cur_step % training_args.eval_steps == 0 or cur_step % steps_per_epoch == 0) and cur_step > 0:
# evaluate
eval_loader = glue_eval_data_collator(eval_dataset, eval_batch_size)
for batch in tqdm(
eval_loader,
total=math.ceil(len(eval_dataset) / eval_batch_size),
desc="Evaluating ...",
position=2,
):
labels = batch.pop("labels")
predictions = pad_shard_unpad(p_eval_step)(
state, batch, min_device_batch=per_device_eval_batch_size
)
metric.add_batch(predictions=np.array(predictions), references=labels)
eval_metric = metric.compute()
logger.info(f"Step... ({cur_step}/{total_steps} | Eval metrics: {eval_metric})")
if has_tensorboard and jax.process_index() == 0:
write_eval_metric(summary_writer, eval_metric, cur_step)
if (cur_step % training_args.save_steps == 0 and cur_step > 0) or (cur_step == total_steps):
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(unreplicate(state.params))
model.save_pretrained(training_args.output_dir, params=params)
tokenizer.save_pretrained(training_args.output_dir)
if training_args.push_to_hub:
repo.push_to_hub(commit_message=f"Saving weights and logs of step {cur_step}", blocking=False)
epochs.desc = f"Epoch ... {epoch + 1}/{num_epochs}"
# save the eval metrics in json
if jax.process_index() == 0:
eval_metric = {f"eval_{metric_name}": value for metric_name, value in eval_metric.items()}
path = os.path.join(training_args.output_dir, "eval_results.json")
with open(path, "w") as f:
json.dump(eval_metric, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/text-classification/README.md
|
<!---
Copyright 2021 The Google Flax Team Authors and HuggingFace Team. 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.
-->
# Text classification examples
## GLUE tasks
Based on the script [`run_flax_glue.py`](https://github.com/huggingface/transformers/blob/main/examples/flax/text-classification/run_flax_glue.py).
Fine-tuning the library models for sequence classification on the GLUE benchmark: [General Language Understanding
Evaluation](https://gluebenchmark.com/). This script can fine-tune any of the models on the [hub](https://huggingface.co/models) and can also be used for a
dataset hosted on our [hub](https://huggingface.co/datasets) or your own data in a csv or a JSON file (the script might need some tweaks in that case,
refer to the comments inside for help).
GLUE is made up of a total of 9 different tasks. Here is how to run the script on one of them:
```bash
export TASK_NAME=mrpc
python run_flax_glue.py \
--model_name_or_path bert-base-cased \
--task_name ${TASK_NAME} \
--max_seq_length 128 \
--learning_rate 2e-5 \
--num_train_epochs 3 \
--per_device_train_batch_size 4 \
--eval_steps 100 \
--output_dir ./$TASK_NAME/ \
--push_to_hub
```
where task name can be one of cola, mnli, mnli_mismatched, mnli_matched, mrpc, qnli, qqp, rte, sst2, stsb, wnli.
Using the command above, the script will train for 3 epochs and run eval after each epoch.
Metrics and hyperparameters are stored in Tensorflow event files in `--output_dir`.
You can see the results by running `tensorboard` in that directory:
```bash
$ tensorboard --logdir .
```
or directly on the hub under *Training metrics*.
### Accuracy Evaluation
We train five replicas and report mean accuracy and stdev on the dev set below.
We use the settings as in the command above (with an exception for MRPC and
WNLI which are tiny and where we used 5 epochs instead of 3), and we use a total
train batch size of 32 (we train on 8 Cloud v3 TPUs, so a per-device batch size of 4),
On the task other than MRPC and WNLI we train for 3 these epochs because this is the standard,
but looking at the training curves of some of them (e.g., SST-2, STS-b), it appears the models
are undertrained and we could get better results when training longer.
In the Tensorboard results linked below, the random seed of each model is equal to the ID of the run. So in order to reproduce run 1, run the command above with `--seed=1`. The best run used random seed 3, which is the default in the script. The results of all runs are in [this Google Sheet](https://docs.google.com/spreadsheets/d/1p3XzReMO75m_XdEJvPue-PIq_PN-96J2IJpJW1yS-10/edit?usp=sharing).
| Task | Metric | Acc (best run) | Acc (avg/5runs) | Stdev | Metrics |
|-------|------------------------------|----------------|-----------------|-----------|--------------------------------------------------------------------------|
| CoLA | Matthews corr | 60.57 | 59.04 | 1.06 | [tfhub.dev](https://tensorboard.dev/experiment/lfr2adVpRtmLDALKrElkzg/) |
| SST-2 | Accuracy | 92.66 | 92.23 | 0.57 | [tfhub.dev](https://tensorboard.dev/experiment/jYvfv2trRHKMjoWnXVwrZA/) |
| MRPC | F1/Accuracy | 89.90/85.78 | 88.97/84.36 | 0.72/1.09 | [tfhub.dev](https://tensorboard.dev/experiment/bo3W3DEoRw2Q7YXjWrJkfg/) |
| STS-B | Pearson/Spearman corr. | 89.04/88.70 | 88.94/88.63 | 0.07/0.07 | [tfhub.dev](https://tensorboard.dev/experiment/fxVwbLD7QpKhbot0r9rn2w/) |
| QQP | Accuracy/F1 | 90.81/87.58 | 90.76/87.51 | 0.05/0.06 | [tfhub.dev](https://tensorboard.dev/experiment/di089Rc9TZmsnKRMrYNLsA/) |
| MNLI | Matched acc. | 84.10 | 83.80 | 0.16 | [tfhub.dev](https://tensorboard.dev/experiment/JgNCGHDJSRaW6HBx6YQFYQ/) |
| QNLI | Accuracy | 91.01 | 90.82 | 0.17 | [tfhub.dev](https://tensorboard.dev/experiment/Bq7cMGJnQMSggYgL8qNGeQ/) |
| RTE | Accuracy | 66.06 | 64.76 | 1.04 | [tfhub.dev](https://tensorboard.dev/experiment/66Eq24bhRjqN6CEhgDSGqQ/) |
| WNLI | Accuracy | 46.48 | 37.01 | 6.83 | [tfhub.dev](https://tensorboard.dev/experiment/TAqcnddqTkWvVEeGaWwIdQ/) |
Some of these results are significantly different from the ones reported on the test set of GLUE benchmark on the
website. For QQP and WNLI, please refer to [FAQ #12](https://gluebenchmark.com/faq) on the website.
### Runtime evaluation
We also ran each task once on a single V100 GPU, 8 V100 GPUs, and 8 Cloud v3 TPUs and report the
overall training time below. For comparison we ran Pytorch's [run_glue.py](https://github.com/huggingface/transformers/blob/main/examples/pytorch/text-classification/run_glue.py) on a single GPU (last column).
| Task | TPU v3-8 | 8 GPU | [1 GPU](https://tensorboard.dev/experiment/mkPS4Zh8TnGe1HB6Yzwj4Q) | 1 GPU (Pytorch) |
|-------|-----------|------------|------------|-----------------|
| CoLA | 1m 42s | 1m 26s | 3m 9s | 4m 6s |
| SST-2 | 5m 12s | 6m 28s | 22m 33s | 34m 37s |
| MRPC | 1m 29s | 1m 14s | 2m 20s | 2m 56s |
| STS-B | 1m 30s | 1m 12s | 2m 16s | 2m 48s |
| QQP | 22m 50s | 31m 48s | 1h 59m 41s | 2h 54m |
| MNLI | 25m 03s | 33m 55s | 2h 9m 37s | 3h 7m 6s |
| QNLI | 7m30s | 9m 40s | 34m 40s | 49m 8s |
| RTE | 1m 20s | 55s | 1m 10s | 1m 16s |
| WNLI | 1m 11s | 48s | 39s | 36s |
|-------|
| **TOTAL** | 1h 03m | 1h 28m | 5h 16m | 6h 37m |
*All experiments are ran on Google Cloud Platform.
GPU experiments are ran without further optimizations besides JAX
transformations. GPU experiments are ran with full precision (fp32). "TPU v3-8"
are 8 TPU cores on 4 chips (each chips has 2 cores), while "8 GPU" are 8 GPU chips.
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/text-classification/requirements.txt
|
datasets >= 1.1.3
jax>=0.2.8
jaxlib>=0.1.59
flax>=0.3.5
optax>=0.0.8
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/README.md
|
<!---
Copyright 2021 The HuggingFace Team. 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.
-->
# Language model training examples
The following example showcases how to train a language model from scratch
using the JAX/Flax backend.
JAX/Flax allows you to trace pure functions and compile them into efficient, fused accelerator code on both GPU and TPU.
Models written in JAX/Flax are **immutable** and updated in a purely functional
way which enables simple and efficient model parallelism.
## Masked language modeling
In the following, we demonstrate how to train a bi-directional transformer model
using masked language modeling objective as introduced in [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://arxiv.org/abs/1810.04805).
More specifically, we demonstrate how JAX/Flax can be leveraged
to pre-train [**`roberta-base`**](https://huggingface.co/roberta-base)
in Norwegian on a single TPUv3-8 pod.
The example script uses the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.
To setup all relevant files for training, let's create a directory.
```bash
mkdir ./norwegian-roberta-base
```
### Train tokenizer
In the first step, we train a tokenizer to efficiently process the text input for the model. Similar to how it is shown in [How to train a new language model from scratch using Transformers and Tokenizers](https://huggingface.co/blog/how-to-train), we use a **`ByteLevelBPETokenizer`**.
The tokenizer is trained on the complete Norwegian dataset of OSCAR
and consequently saved in the cloned model directory.
This can take up to 10 minutes depending on your hardware ☕.
```python
from datasets import load_dataset
from tokenizers import trainers, Tokenizer, normalizers, ByteLevelBPETokenizer
# load dataset
dataset = load_dataset("oscar", "unshuffled_deduplicated_no", split="train")
# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()
def batch_iterator(batch_size=1000):
for i in range(0, len(dataset), batch_size):
yield dataset[i: i + batch_size]["text"]
# Customized training
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
"<s>",
"<pad>",
"</s>",
"<unk>",
"<mask>",
])
# Save files to disk
tokenizer.save("./norwegian-roberta-base/tokenizer.json")
```
### Create configuration
Next, we create the model's configuration file. This is as simple
as loading and storing [`**roberta-base**`](https://huggingface.co/roberta-base)
in the local model folder:
```python
from transformers import RobertaConfig
config = RobertaConfig.from_pretrained("roberta-base", vocab_size=50265)
config.save_pretrained("./norwegian-roberta-base")
```
Great, we have set up our model repository. During training, we will automatically
push the training logs and model weights to the repo.
### Train model
Next we can run the example script to pretrain the model:
```bash
python run_mlm_flax.py \
--output_dir="./norwegian-roberta-base" \
--model_type="roberta" \
--config_name="./norwegian-roberta-base" \
--tokenizer_name="./norwegian-roberta-base" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_no" \
--max_seq_length="128" \
--weight_decay="0.01" \
--per_device_train_batch_size="128" \
--per_device_eval_batch_size="128" \
--learning_rate="3e-4" \
--warmup_steps="1000" \
--overwrite_output_dir \
--num_train_epochs="18" \
--adam_beta1="0.9" \
--adam_beta2="0.98" \
--logging_steps="500" \
--save_steps="2500" \
--eval_steps="2500" \
--push_to_hub
```
Training should converge at a loss and accuracy
of 1.78 and 0.64 respectively after 18 epochs on a single TPUv3-8.
This should take less than 18 hours.
Training statistics can be accessed on [tfhub.dev](https://tensorboard.dev/experiment/GdYmdak2TWeVz0DDRYOrrg).
For a step-by-step walkthrough of how to do masked language modeling in Flax, please have a
look at [this](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/masked_language_modeling_flax.ipynb) google colab.
## Causal language modeling
In the following, we demonstrate how to train an auto-regressive causal transformer model
in JAX/Flax.
More specifically, we pretrain a randomly initialized [**`gpt2`**](https://huggingface.co/gpt2) model in Norwegian on a single TPUv3-8.
to pre-train 124M [**`gpt2`**](https://huggingface.co/gpt2)
in Norwegian on a single TPUv3-8 pod.
The example script uses the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.
To setup all relevant files for training, let's create a directory.
```bash
mkdir ./norwegian-gpt2
```
### Train tokenizer
In the first step, we train a tokenizer to efficiently process the text input for the model. Similar to how it is shown in [How to train a new language model from scratch using Transformers and Tokenizers](https://huggingface.co/blog/how-to-train), we use a **`ByteLevelBPETokenizer`**.
The tokenizer is trained on the complete Norwegian dataset of OSCAR
and consequently saved in the cloned model directory.
This can take up to 10 minutes depending on your hardware ☕.
```python
from datasets import load_dataset
from tokenizers import trainers, Tokenizer, normalizers, ByteLevelBPETokenizer
# load dataset
dataset = load_dataset("oscar", "unshuffled_deduplicated_no", split="train")
# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()
def batch_iterator(batch_size=1000):
for i in range(0, len(dataset), batch_size):
yield dataset[i: i + batch_size]["text"]
# Customized training
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50257, min_frequency=2, special_tokens=[
"<s>",
"<pad>",
"</s>",
"<unk>",
"<mask>",
])
# Save files to disk
tokenizer.save("./norwegian-gpt2/tokenizer.json")
```
### Create configuration
Next, we create the model's configuration file. This is as simple
as loading and storing [`**gpt2**`](https://huggingface.co/gpt2)
in the local model folder:
```python
from transformers import GPT2Config
config = GPT2Config.from_pretrained("gpt2", resid_pdrop=0.0, embd_pdrop=0.0, attn_pdrop=0.0, vocab_size=50257)
config.save_pretrained("./norwegian-gpt2")
```
Great, we have set up our model repository. During training, we will now automatically
push the training logs and model weights to the repo.
### Train model
Finally, we can run the example script to pretrain the model:
```bash
python run_clm_flax.py \
--output_dir="./norwegian-gpt2" \
--model_type="gpt2" \
--config_name="./norwegian-gpt2" \
--tokenizer_name="./norwegian-gpt2" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_no" \
--do_train --do_eval \
--block_size="512" \
--per_device_train_batch_size="64" \
--per_device_eval_batch_size="64" \
--learning_rate="5e-3" --warmup_steps="1000" \
--adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.01" \
--overwrite_output_dir \
--num_train_epochs="20" \
--logging_steps="500" \
--save_steps="2500" \
--eval_steps="2500" \
--push_to_hub
```
Training should converge at a loss and perplexity
of 3.24 and 25.72 respectively after 20 epochs on a single TPUv3-8.
This should take less than ~21 hours.
Training statistics can be accessed on [tfhub.de](https://tensorboard.dev/experiment/2zEhLwJ0Qp2FAkI3WVH9qA).
For a step-by-step walkthrough of how to do causal language modeling in Flax, please have a
look at [this](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/causal_language_modeling_flax.ipynb) google colab.
## T5-like span-masked language modeling
In the following, we demonstrate how to train a T5 model using the span-masked language model
objective as proposed in the [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683).
More specifically, we demonstrate how JAX/Flax can be leveraged
to pre-train [**`google/t5-v1_1-base`**](https://huggingface.co/google/t5-v1_1-base)
in Norwegian on a single TPUv3-8 pod.
The example script uses the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.
Let's start by creating a model repository to save the trained model and logs.
Here we call the model `"norwegian-t5-base"`, but you can change the model name as you like.
To setup all relevant files for training, let's create a directory.
```bash
cd ./norwegian-t5-base
```
### Train tokenizer
In the first step, we train a tokenizer to efficiently process the text input for the model.
We make use of the [tokenizers](https://github.com/huggingface/tokenizers) library to train
a sentencepiece unigram tokenizer as shown in [t5_tokenizer_model.py](https://github.com/huggingface/transformers/tree/main/examples/flax/language-modeling/t5_tokenizer_model.py)
which is heavily inspired from [yandex-research/DeDLOC's tokenizer model](https://github.com/yandex-research/DeDLOC/blob/5c994bc64e573702a9a79add3ecd68b38f14b548/sahajbert/tokenizer/tokenizer_model.py) .
The tokenizer is trained on the complete Norwegian dataset of OSCAR
and consequently saved in the cloned model directory.
This can take up to 120 minutes depending on your hardware ☕☕☕ .
```python
import datasets
from t5_tokenizer_model import SentencePieceUnigramTokenizer
vocab_size = 32_000
input_sentence_size = None
# Initialize a dataset
dataset = datasets.load_dataset("oscar", name="unshuffled_deduplicated_no", split="train")
tokenizer = SentencePieceUnigramTokenizer(unk_token="<unk>", eos_token="</s>", pad_token="<pad>")
# Build an iterator over this dataset
def batch_iterator(input_sentence_size=None):
if input_sentence_size is None:
input_sentence_size = len(dataset)
batch_length = 100
for i in range(0, input_sentence_size, batch_length):
yield dataset[i: i + batch_length]["text"]
# Train tokenizer
tokenizer.train_from_iterator(
iterator=batch_iterator(input_sentence_size=input_sentence_size),
vocab_size=vocab_size,
show_progress=True,
)
# Save files to disk
tokenizer.save("./norwegian-t5-base/tokenizer.json")
```
### Create configuration
Next, we create the model's configuration file. This is as simple
as loading and storing [`**google/t5-v1_1-base**`](https://huggingface.co/google/t5-v1_1-base)
in the local model folder:
```python
from transformers import T5Config
config = T5Config.from_pretrained("google/t5-v1_1-base", vocab_size=tokenizer.get_vocab_size())
config.save_pretrained("./norwegian-t5-base")
```
Great, we have set up our model repository. During training, we will automatically
push the training logs and model weights to the repo.
### Train model
Next we can run the example script to pretrain the model:
```bash
python run_t5_mlm_flax.py \
--output_dir="./norwegian-t5-base" \
--model_type="t5" \
--config_name="./norwegian-t5-base" \
--tokenizer_name="./norwegian-t5-base" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_no" \
--max_seq_length="512" \
--per_device_train_batch_size="32" \
--per_device_eval_batch_size="32" \
--adafactor \
--learning_rate="0.005" \
--weight_decay="0.001" \
--warmup_steps="2000" \
--overwrite_output_dir \
--logging_steps="500" \
--save_steps="10000" \
--eval_steps="2500" \
--push_to_hub
```
Training should converge at a loss and accuracy
of 2.36 and 57.0 respectively after 3 epochs on a single TPUv3-8.
This should take around 4.5 hours.
Training statistics can be accessed on directly on the 🤗 [hub](https://huggingface.co/patrickvonplaten/t5-base-norwegian/tensorboard)
## BART: Denoising language modeling
In the following, we demonstrate how to train a BART model
using denoising language modeling objective as introduced in [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461).
More specifically, we demonstrate how JAX/Flax can be leveraged
to pre-train [**`bart-base`**](https://huggingface.co/facebook/bart-base)
in Norwegian on a single TPUv3-8 pod.
The example script uses the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.
To setup all relevant files for training, let's create a directory.
```bash
mkdir ./norwegian-bart-base
```
### Train tokenizer
In the first step, we train a tokenizer to efficiently process the text input for the model. Similar to how it is shown in [How to train a new language model from scratch using Transformers and Tokenizers](https://huggingface.co/blog/how-to-train), we use a **`ByteLevelBPETokenizer`**.
The tokenizer is trained on the complete Norwegian dataset of OSCAR
and consequently saved in the cloned model directory.
This can take up to 10 minutes depending on your hardware ☕.
```python
from datasets import load_dataset
from tokenizers import trainers, Tokenizer, normalizers, ByteLevelBPETokenizer
# load dataset
dataset = load_dataset("oscar", "unshuffled_deduplicated_no", split="train")
# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()
def batch_iterator(batch_size=1000):
for i in range(0, len(dataset), batch_size):
yield dataset[i: i + batch_size]["text"]
# Customized training
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
"<s>",
"<pad>",
"</s>",
"<unk>",
"<mask>",
])
# Save files to disk
tokenizer.save("./norwegian-bart-base/tokenizer.json")
```
### Create configuration
Next, we create the model's configuration file. This is as simple
as loading and storing [`**facebook/bart-base**`](https://huggingface.co/facebook/bart-base)
in the local model folder:
```python
from transformers import BartConfig
config = BartConfig.from_pretrained("facebook/bart-base", vocab_size=50265)
config.save_pretrained("./norwegian-bart-base")
```
Great, we have set up our model repository. During training, we will automatically
push the training logs and model weights to the repo.
### Train model
Next we can run the example script to pretrain the model:
```bash
python run_bart_dlm_flax.py \
--output_dir="./norwegian-bart-base" \
--config_name="./norwegian-bart-base" \
--tokenizer_name="./norwegian-bart-base" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_no" \
--max_seq_length="1024" \
--per_device_train_batch_size="32" \
--per_device_eval_batch_size="32" \
--learning_rate="1e-4" \
--warmup_steps="2000" \
--overwrite_output_dir \
--logging_steps="500" \
--save_steps="2000" \
--eval_steps="2000" \
--push_to_hub
```
Training should converge at a loss and accuracy
of 1.36 and 0.77 respectively after 3 epochs on a single TPUv3-8.
This should take less than 6 hours.
Training statistics can be accessed on [tfhub.dev](https://tensorboard.dev/experiment/Maw62QlaSXWS0MOf2V2lbg/).
## Runtime evaluation
We also ran masked language modeling using PyTorch/XLA on a TPUv3-8, and PyTorch on 8 V100 GPUs. We report the
overall training time below.
For reproducibility, we state the training commands used for PyTorch/XLA and PyTorch further below.
| Task | [TPU v3-8 (Flax)](https://tensorboard.dev/experiment/GdYmdak2TWeVz0DDRYOrrg/) | [TPU v3-8 (Pytorch/XLA)](https://tensorboard.dev/experiment/7Jq1kcQQRAmy12KOdXek7A/)| [8 GPU (PyTorch)](https://tensorboard.dev/experiment/PJneV8FQRxa2unPw1QnVHA) |
|-------|-----------|------------|------------|
| MLM | 15h32m | 23h46m | 44h14m |
*All experiments are ran on Google Cloud Platform.
GPU experiments are ran without further optimizations besides JAX
transformations. GPU experiments are ran with full precision (fp32). "TPU v3-8"
are 8 TPU cores on 4 chips (each chips has 2 cores), while "8 GPU" are 8 GPU chips.
### Script to run MLM with PyTorch/XLA on TPUv3-8
For comparison one can run the same pre-training with PyTorch/XLA on TPU. To set up PyTorch/XLA on Cloud TPU VMs, please
refer to [this](https://cloud.google.com/tpu/docs/pytorch-xla-ug-tpu-vm) guide.
Having created the tokenzier and configuration in `norwegian-roberta-base`, we create the following symbolic links:
```bash
ln -s ~/transformers/examples/pytorch/language-modeling/run_mlm.py ./
ln -s ~/transformers/examples/pytorch/xla_spawn.py ./
```
, set the following environment variables:
```bash
export XRT_TPU_CONFIG="localservice;0;localhost:51011"
unset LD_PRELOAD
export NUM_TPUS=8
export TOKENIZERS_PARALLELISM=0
export MODEL_DIR="./norwegian-roberta-base"
mkdir -p ${MODEL_DIR}
```
, and start training as follows:
```bash
python3 xla_spawn.py --num_cores ${NUM_TPUS} run_mlm.py --output_dir="./runs" \
--model_type="roberta" \
--config_name="${MODEL_DIR}" \
--tokenizer_name="${MODEL_DIR}" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_no" \
--max_seq_length="128" \
--weight_decay="0.01" \
--per_device_train_batch_size="128" \
--per_device_eval_batch_size="128" \
--learning_rate="3e-4" \
--warmup_steps="1000" \
--overwrite_output_dir \
--num_train_epochs="18" \
--adam_beta1="0.9" \
--adam_beta2="0.98" \
--do_train \
--do_eval \
--logging_steps="500" \
--evaluation_strategy="epoch" \
--report_to="tensorboard" \
--save_strategy="no"
```
### Script to compare pre-training with PyTorch on 8 GPU V100's
For comparison you can run the same pre-training with PyTorch on GPU. Note that we have to make use of `gradient_accumulation`
because the maximum batch size that fits on a single V100 GPU is 32 instead of 128.
Having created the tokenzier and configuration in `norwegian-roberta-base`, we create the following symbolic links:
```bash
ln -s ~/transformers/examples/pytorch/language-modeling/run_mlm.py ./
```
, set some environment variables:
```bash
export NUM_GPUS=8
export TOKENIZERS_PARALLELISM=0
export MODEL_DIR="./norwegian-roberta-base"
mkdir -p ${MODEL_DIR}
```
, and can start training as follows:
```bash
python3 -m torch.distributed.launch --nproc_per_node ${NUM_GPUS} run_mlm.py \
--output_dir="${MODEL_DIR}" \
--model_type="roberta" \
--config_name="${MODEL_DIR}" \
--tokenizer_name="${MODEL_DIR}" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_no" \
--max_seq_length="128" \
--weight_decay="0.01" \
--per_device_train_batch_size="32" \
--per_device_eval_batch_size="32" \
--gradient_accumulation="4" \
--learning_rate="3e-4" \
--warmup_steps="1000" \
--overwrite_output_dir \
--num_train_epochs="18" \
--adam_beta1="0.9" \
--adam_beta2="0.98" \
--do_train \
--do_eval \
--logging_steps="500" \
--evaluation_strategy="steps" \
--report_to="tensorboard" \
--save_strategy="no"
```
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/run_t5_mlm_flax.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Pretraining the library models for T5-like span-masked language modeling on a text file or a dataset.
Here is the full list of checkpoints on the hub that can be pretrained by this script:
https://huggingface.co/models?filter=t5
"""
import json
import logging
import math
import os
import sys
import time
import warnings
from dataclasses import asdict, dataclass, field
# You can also adapt this script on your own masked language modeling task. Pointers for this are left as comments.
from enum import Enum
from itertools import chain
from pathlib import Path
from typing import Dict, List, Optional
import flax
import jax
import jax.numpy as jnp
import numpy as np
import optax
from datasets import load_dataset
from flax import jax_utils, traverse_util
from flax.jax_utils import pad_shard_unpad
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard
from huggingface_hub import Repository, create_repo
from tqdm import tqdm
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_MASKED_LM_MAPPING,
AutoTokenizer,
BatchEncoding,
FlaxT5ForConditionalGeneration,
HfArgumentParser,
PreTrainedTokenizerBase,
T5Config,
is_tensorboard_available,
set_seed,
)
from transformers.models.t5.modeling_flax_t5 import shift_tokens_right
from transformers.utils import send_example_telemetry
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class TrainingArguments:
output_dir: str = field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."},
)
overwrite_output_dir: bool = field(
default=False,
metadata={
"help": (
"Overwrite the content of the output directory. "
"Use this to continue training if output_dir points to a checkpoint directory."
)
},
)
do_train: bool = field(default=False, metadata={"help": "Whether to run training."})
do_eval: bool = field(default=False, metadata={"help": "Whether to run eval on the dev set."})
per_device_train_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for training."}
)
per_device_eval_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for evaluation."}
)
learning_rate: float = field(default=5e-5, metadata={"help": "The initial learning rate for AdamW."})
weight_decay: float = field(default=0.0, metadata={"help": "Weight decay for AdamW if we apply some."})
adam_beta1: float = field(default=0.9, metadata={"help": "Beta1 for AdamW optimizer"})
adam_beta2: float = field(default=0.999, metadata={"help": "Beta2 for AdamW optimizer"})
adam_epsilon: float = field(default=1e-8, metadata={"help": "Epsilon for AdamW optimizer."})
adafactor: bool = field(default=False, metadata={"help": "Whether or not to replace AdamW by Adafactor."})
num_train_epochs: float = field(default=3.0, metadata={"help": "Total number of training epochs to perform."})
warmup_steps: int = field(default=0, metadata={"help": "Linear warmup over warmup_steps."})
logging_steps: int = field(default=500, metadata={"help": "Log every X updates steps."})
save_steps: int = field(default=500, metadata={"help": "Save checkpoint every X updates steps."})
eval_steps: int = field(default=None, metadata={"help": "Run an evaluation every X steps."})
seed: int = field(default=42, metadata={"help": "Random seed that will be set at the beginning of training."})
push_to_hub: bool = field(
default=False, metadata={"help": "Whether or not to upload the trained model to the model hub after training."}
)
hub_model_id: str = field(
default=None, metadata={"help": "The name of the repository to keep in sync with the local `output_dir`."}
)
hub_token: str = field(default=None, metadata={"help": "The token to use to push to the Model Hub."})
def __post_init__(self):
if self.output_dir is not None:
self.output_dir = os.path.expanduser(self.output_dir)
def to_dict(self):
"""
Serializes this instance while replace `Enum` by their values (for JSON serialization support). It obfuscates
the token values by removing their value.
"""
d = asdict(self)
for k, v in d.items():
if isinstance(v, Enum):
d[k] = v.value
if isinstance(v, list) and len(v) > 0 and isinstance(v[0], Enum):
d[k] = [x.value for x in v]
if k.endswith("_token"):
d[k] = f"<{k.upper()}>"
return d
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
)
},
)
use_auth_token: bool = field(
default=None,
metadata={
"help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
train_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input train ref data file for whole word masking in Chinese."},
)
validation_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input validation ref data file for whole word masking in Chinese."},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
max_seq_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total input sequence length after tokenization and masking. Sequences longer than this"
" will be truncated. Default to the max input length of the model."
)
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
mlm_probability: float = field(
default=0.15, metadata={"help": "Ratio of tokens to mask for span masked language modeling loss"}
)
mean_noise_span_length: float = field(
default=3.0,
metadata={"help": "Mean span length of masked tokens"},
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file."
def compute_input_and_target_lengths(inputs_length, noise_density, mean_noise_span_length):
"""This function is copy of `random_spans_helper <https://github.com/google-research/text-to-text-transfer-transformer/blob/84f8bcc14b5f2c03de51bd3587609ba8f6bbd1cd/t5/data/preprocessors.py#L2466>`__ .
Training parameters to avoid padding with random_spans_noise_mask.
When training a model with random_spans_noise_mask, we would like to set the other
training hyperparmeters in a way that avoids padding.
This function helps us compute these hyperparameters.
We assume that each noise span in the input is replaced by extra_tokens_per_span_inputs sentinel tokens,
and each non-noise span in the targets is replaced by extra_tokens_per_span_targets sentinel tokens.
This function tells us the required number of tokens in the raw example (for split_tokens())
as well as the length of the encoded targets. Note that this function assumes
the inputs and targets will have EOS appended and includes that in the reported length.
Args:
inputs_length: an integer - desired length of the tokenized inputs sequence
noise_density: a float
mean_noise_span_length: a float
Returns:
tokens_length: length of original text in tokens
targets_length: an integer - length in tokens of encoded targets sequence
"""
def _tokens_length_to_inputs_length_targets_length(tokens_length):
num_noise_tokens = int(round(tokens_length * noise_density))
num_nonnoise_tokens = tokens_length - num_noise_tokens
num_noise_spans = int(round(num_noise_tokens / mean_noise_span_length))
# inputs contain all nonnoise tokens, sentinels for all noise spans
# and one EOS token.
_input_length = num_nonnoise_tokens + num_noise_spans + 1
_output_length = num_noise_tokens + num_noise_spans + 1
return _input_length, _output_length
tokens_length = inputs_length
while _tokens_length_to_inputs_length_targets_length(tokens_length + 1)[0] <= inputs_length:
tokens_length += 1
inputs_length, targets_length = _tokens_length_to_inputs_length_targets_length(tokens_length)
# minor hack to get the targets length to be equal to inputs length
# which is more likely to have been set to a nice round number.
if noise_density == 0.5 and targets_length > inputs_length:
tokens_length -= 1
targets_length -= 1
return tokens_length, targets_length
@flax.struct.dataclass
class FlaxDataCollatorForT5MLM:
"""
Data collator used for T5 span-masked language modeling.
It is made sure that after masking the inputs are of length `data_args.max_seq_length` and targets are also of fixed length.
For more information on how T5 span-masked language modeling works, one can take a look
at the `official paper <https://arxiv.org/pdf/1910.10683.pdf>`__
or the `official code for preprocessing <https://github.com/google-research/text-to-text-transfer-transformer/blob/master/t5/data/preprocessors.py>`__ .
Args:
tokenizer (:class:`~transformers.PreTrainedTokenizer` or :class:`~transformers.PreTrainedTokenizerFast`):
The tokenizer used for encoding the data.
noise_density (:obj:`float`):
The probability with which to (randomly) mask tokens in the input.
mean_noise_span_length (:obj:`float`):
The average span length of the masked tokens.
input_length (:obj:`int`):
The expected input length after masking.
target_length (:obj:`int`):
The expected target length after masking.
pad_token_id: (:obj:`int`):
The pad token id of the model
decoder_start_token_id: (:obj:`int):
The decoder start token id of the model
"""
tokenizer: PreTrainedTokenizerBase
noise_density: float
mean_noise_span_length: float
input_length: int
target_length: int
pad_token_id: int
decoder_start_token_id: int
def __call__(self, examples: List[Dict[str, np.ndarray]]) -> BatchEncoding:
# convert list to dict and tensorize input
batch = BatchEncoding(
{k: np.array([examples[i][k] for i in range(len(examples))]) for k, v in examples[0].items()}
)
input_ids = batch["input_ids"]
batch_size, expandend_input_length = input_ids.shape
mask_indices = np.asarray([self.random_spans_noise_mask(expandend_input_length) for i in range(batch_size)])
labels_mask = ~mask_indices
input_ids_sentinel = self.create_sentinel_ids(mask_indices.astype(np.int8))
labels_sentinel = self.create_sentinel_ids(labels_mask.astype(np.int8))
batch["input_ids"] = self.filter_input_ids(input_ids, input_ids_sentinel)
batch["labels"] = self.filter_input_ids(input_ids, labels_sentinel)
if batch["input_ids"].shape[-1] != self.input_length:
raise ValueError(
f"`input_ids` are incorrectly preprocessed. `input_ids` length is {batch['input_ids'].shape[-1]}, but"
f" should be {self.input_length}."
)
if batch["labels"].shape[-1] != self.target_length:
raise ValueError(
f"`labels` are incorrectly preprocessed. `labels` length is {batch['labels'].shape[-1]}, but should be"
f" {self.target_length}."
)
# to check that tokens are correctly preprocessed, one can run `self.tokenizer.batch_decode(input_ids)` and `self.tokenizer.batch_decode(labels)` here...
batch["decoder_input_ids"] = shift_tokens_right(
batch["labels"], self.pad_token_id, self.decoder_start_token_id
)
return batch
def create_sentinel_ids(self, mask_indices):
"""
Sentinel ids creation given the indices that should be masked.
The start indices of each mask are replaced by the sentinel ids in increasing
order. Consecutive mask indices to be deleted are replaced with `-1`.
"""
start_indices = mask_indices - np.roll(mask_indices, 1, axis=-1) * mask_indices
start_indices[:, 0] = mask_indices[:, 0]
sentinel_ids = np.where(start_indices != 0, np.cumsum(start_indices, axis=-1), start_indices)
sentinel_ids = np.where(sentinel_ids != 0, (len(self.tokenizer) - sentinel_ids), 0)
sentinel_ids -= mask_indices - start_indices
return sentinel_ids
def filter_input_ids(self, input_ids, sentinel_ids):
"""
Puts sentinel mask on `input_ids` and fuse consecutive mask tokens into a single mask token by deleting.
This will reduce the sequence length from `expanded_inputs_length` to `input_length`.
"""
batch_size = input_ids.shape[0]
input_ids_full = np.where(sentinel_ids != 0, sentinel_ids, input_ids)
# input_ids tokens and sentinel tokens are >= 0, tokens < 0 are
# masked tokens coming after sentinel tokens and should be removed
input_ids = input_ids_full[input_ids_full >= 0].reshape((batch_size, -1))
input_ids = np.concatenate(
[input_ids, np.full((batch_size, 1), self.tokenizer.eos_token_id, dtype=np.int32)], axis=-1
)
return input_ids
def random_spans_noise_mask(self, length):
"""This function is copy of `random_spans_helper <https://github.com/google-research/text-to-text-transfer-transformer/blob/84f8bcc14b5f2c03de51bd3587609ba8f6bbd1cd/t5/data/preprocessors.py#L2682>`__ .
Noise mask consisting of random spans of noise tokens.
The number of noise tokens and the number of noise spans and non-noise spans
are determined deterministically as follows:
num_noise_tokens = round(length * noise_density)
num_nonnoise_spans = num_noise_spans = round(num_noise_tokens / mean_noise_span_length)
Spans alternate between non-noise and noise, beginning with non-noise.
Subject to the above restrictions, all masks are equally likely.
Args:
length: an int32 scalar (length of the incoming token sequence)
noise_density: a float - approximate density of output mask
mean_noise_span_length: a number
Returns:
a boolean tensor with shape [length]
"""
orig_length = length
num_noise_tokens = int(np.round(length * self.noise_density))
num_nonnoise_tokens = length - num_noise_tokens
# avoid degeneracy by ensuring positive numbers of noise and nonnoise tokens.
num_noise_tokens = min(max(num_noise_tokens, 1), length - 1)
# num_noise_tokens should be less than num_noise_tokens and num_nonnoise_tokens
num_noise_spans = int(np.round(min(num_noise_tokens, num_nonnoise_tokens) / self.mean_noise_span_length))
# avoid degeneracy by ensuring positive number of noise spans
num_noise_spans = max(num_noise_spans, 1)
# pick the lengths of the noise spans and the non-noise spans
def _random_segmentation(num_items, num_segments):
"""Partition a sequence of items randomly into non-empty segments.
Args:
num_items: an integer scalar > 0
num_segments: an integer scalar in [1, num_items]
Returns:
a Tensor with shape [num_segments] containing positive integers that add
up to num_items
"""
mask_indices = np.arange(num_items - 1) < (num_segments - 1)
np.random.shuffle(mask_indices)
first_in_segment = np.pad(mask_indices, [[1, 0]])
segment_id = np.cumsum(first_in_segment)
# count length of sub segments assuming that list is sorted
_, segment_length = np.unique(segment_id, return_counts=True)
return segment_length
noise_span_lengths = _random_segmentation(num_noise_tokens, num_noise_spans)
nonnoise_span_lengths = _random_segmentation(num_nonnoise_tokens, num_noise_spans)
interleaved_span_lengths = np.reshape(
np.stack([nonnoise_span_lengths, noise_span_lengths], axis=1), [num_noise_spans * 2]
)
span_starts = np.cumsum(interleaved_span_lengths)[:-1]
span_start_indicator = np.zeros((length,), dtype=np.int8)
span_start_indicator[span_starts] = True
span_num = np.cumsum(span_start_indicator)
is_noise = np.equal(span_num % 2, 1)
return is_noise[:orig_length]
def generate_batch_splits(samples_idx: np.ndarray, batch_size: int, drop_last=True) -> np.ndarray:
"""Generate batches of data for a specified batch size from sample indices. If the dataset size is not divisible by
the batch size and `drop_last` is `True`, the last incomplete batch is dropped. Else, it is returned."""
num_samples = len(samples_idx)
if drop_last:
samples_to_remove = num_samples % batch_size
if samples_to_remove != 0:
samples_idx = samples_idx[:-samples_to_remove]
sections_split = num_samples // batch_size
samples_idx = samples_idx.reshape((sections_split, batch_size))
else:
sections_split = math.ceil(num_samples / batch_size)
samples_idx = np.array_split(samples_idx, sections_split)
return samples_idx
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
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 model_args.use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
FutureWarning,
)
if model_args.token is not None:
raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
model_args.token = model_args.use_auth_token
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_t5_mlm", model_args, data_args, framework="flax")
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",
level=logging.INFO,
datefmt="[%X]",
)
# Log on each process the small summary:
logger = logging.getLogger(__name__)
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
# Set seed before initializing model.
set_seed(training_args.seed)
# Handle the repository creation
if training_args.push_to_hub:
# Retrieve of infer repo_name
repo_name = training_args.hub_model_id
if repo_name is None:
repo_name = Path(training_args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=training_args.hub_token).repo_id
# Clone repo locally
repo = Repository(training_args.output_dir, clone_from=repo_id, token=training_args.hub_token)
# 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 if no column called
# 'text' is found. You can easily tweak this behavior (see below).
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,
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
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}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
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,
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
if "validation" not in datasets.keys():
datasets["validation"] = load_dataset(
extension,
data_files=data_files,
split=f"train[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
datasets["train"] = load_dataset(
extension,
data_files=data_files,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
# 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.
# Load pretrained model and tokenizer
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
if model_args.config_name:
config = T5Config.from_pretrained(
model_args.config_name,
cache_dir=model_args.cache_dir,
vocab_size=len(tokenizer),
token=model_args.token,
)
elif model_args.model_name_or_path:
config = T5Config.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
token=model_args.token,
)
else:
config = CONFIG_MAPPING[model_args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
# 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]
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
# Otherwise, we tokenize every text, then concatenate them together before splitting them in smaller parts.
# Since we make sure that all sequences are of the same length, no attention_mask is needed.
def tokenize_function(examples):
return tokenizer(examples[text_column_name], return_attention_mask=False)
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,
)
# T5-like span masked language modeling will fuse consecutively masked tokens to a single sentinel token.
# To ensure that the input length is `max_seq_length`, we need to increase the maximum length
# according to `mlm_probability` and `mean_noise_span_length`. We can also define the label length accordingly.
expanded_inputs_length, targets_length = compute_input_and_target_lengths(
inputs_length=max_seq_length,
noise_density=data_args.mlm_probability,
mean_noise_span_length=data_args.mean_noise_span_length,
)
# Main data processing function that will concatenate all texts from our dataset and generate chunks of expanded_inputs_length.
def group_texts(examples):
# Concatenate all texts.
concatenated_examples = {k: list(chain(*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.
if total_length >= expanded_inputs_length:
total_length = (total_length // expanded_inputs_length) * expanded_inputs_length
# Split by chunks of max_len.
result = {
k: [t[i : i + expanded_inputs_length] for i in range(0, total_length, expanded_inputs_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/process#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,
)
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
dropout_rngs = jax.random.split(rng, jax.local_device_count())
if model_args.model_name_or_path:
model = FlaxT5ForConditionalGeneration.from_pretrained(
model_args.model_name_or_path,
config=config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
token=model_args.token,
)
else:
config.vocab_size = len(tokenizer)
model = FlaxT5ForConditionalGeneration(
config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
)
# Data collator
# This one will take care of randomly masking the tokens.
data_collator = FlaxDataCollatorForT5MLM(
tokenizer=tokenizer,
noise_density=data_args.mlm_probability,
mean_noise_span_length=data_args.mean_noise_span_length,
input_length=max_seq_length,
target_length=targets_length,
pad_token_id=model.config.pad_token_id,
decoder_start_token_id=model.config.decoder_start_token_id,
)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
eval_batch_size = per_device_eval_batch_size * jax.device_count()
num_train_steps = len(tokenized_datasets["train"]) // train_batch_size * num_epochs
num_of_hosts = jax.process_count()
current_host_idx = jax.process_index()
# Create learning rate schedule
warmup_fn = optax.linear_schedule(
init_value=0.0, end_value=training_args.learning_rate, transition_steps=training_args.warmup_steps
)
decay_fn = optax.linear_schedule(
init_value=training_args.learning_rate,
end_value=0,
transition_steps=num_train_steps - training_args.warmup_steps,
)
linear_decay_lr_schedule_fn = optax.join_schedules(
schedules=[warmup_fn, decay_fn], boundaries=[training_args.warmup_steps]
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
# find out all LayerNorm parameters
layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
layer_norm_named_params = {
layer[-2:]
for layer_norm_name in layer_norm_candidates
for layer in flat_params.keys()
if layer_norm_name in "".join(layer).lower()
}
flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
if training_args.adafactor:
# We use the default parameters here to initialize adafactor,
# For more details about the parameters please check https://github.com/deepmind/optax/blob/ed02befef9bf81cbbf236be3d2b0e032e9ed4a40/optax/_src/alias.py#L74
optimizer = optax.adafactor(
learning_rate=linear_decay_lr_schedule_fn,
)
else:
optimizer = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state
state = train_state.TrainState.create(apply_fn=model.__call__, params=model.params, tx=optimizer)
# Define gradient update step fn
def train_step(state, batch, dropout_rng):
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
def loss_fn(params):
labels = batch.pop("labels")
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
# compute loss
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])).mean()
return loss
grad_fn = jax.value_and_grad(loss_fn)
loss, grad = grad_fn(state.params)
grad = jax.lax.pmean(grad, "batch")
new_state = state.apply_gradients(grads=grad)
metrics = jax.lax.pmean(
{"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}, axis_name="batch"
)
return new_state, metrics, new_dropout_rng
# Create parallel version of the train step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
# Define eval fn
def eval_step(params, batch):
labels = batch.pop("labels")
logits = model(**batch, params=params, train=False)[0]
# compute loss
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1]))
# compute accuracy
accuracy = jnp.equal(jnp.argmax(logits, axis=-1), labels)
# summarize metrics
metrics = {"loss": loss.mean(), "accuracy": accuracy.mean()}
metrics = jax.lax.pmean(metrics, axis_name="batch")
return metrics
p_eval_step = jax.pmap(eval_step, "batch", donate_argnums=(0,))
# Replicate the train state on each device
state = jax_utils.replicate(state)
train_time = 0
epochs = tqdm(range(num_epochs), desc="Epoch ... ", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
train_metrics = []
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# Generate an epoch by shuffling sampling indices from the train dataset
num_train_samples = len(tokenized_datasets["train"])
# Avoid using jax.numpy here in case of TPU training
train_samples_idx = np.random.permutation(np.arange(num_train_samples))
train_batch_idx = generate_batch_splits(train_samples_idx, train_batch_size)
# Gather the indexes for creating the batch and do a training step
for step, batch_idx in enumerate(tqdm(train_batch_idx, desc="Training...", position=1)):
samples = [tokenized_datasets["train"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
local_host_model_inputs = {
key: np.split(model_inputs.data[key], num_of_hosts, axis=0)[current_host_idx]
for key, value in model_inputs.data.items()
}
# Model forward
model_inputs = shard(local_host_model_inputs)
state, train_metric, dropout_rngs = p_train_step(state, model_inputs, dropout_rngs)
train_metrics.append(train_metric)
cur_step = epoch * (num_train_samples // train_batch_size) + step
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_metric = jax_utils.unreplicate(train_metric)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step} | Loss: {train_metric['loss'].mean()}, Learning Rate:"
f" {train_metric['learning_rate'].mean()})"
)
train_metrics = []
if cur_step % training_args.eval_steps == 0 and cur_step > 0:
# ======================== Evaluating ==============================
num_eval_samples = len(tokenized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size, drop_last=False)
eval_metrics = []
for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
# Model forward
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# get eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.mean, eval_metrics)
# Update progress bar
epochs.write(f"Step... ({cur_step} | Loss: {eval_metrics['loss']}, Acc: {eval_metrics['accuracy']})")
# Save metrics
if has_tensorboard and jax.process_index() == 0:
write_eval_metric(summary_writer, eval_metrics, cur_step)
if cur_step % training_args.save_steps == 0 and cur_step > 0:
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(jax.tree_util.tree_map(lambda x: x[0], state.params))
model.save_pretrained(training_args.output_dir, params=params)
tokenizer.save_pretrained(training_args.output_dir)
if training_args.push_to_hub:
repo.push_to_hub(commit_message=f"Saving weights and logs of step {cur_step}", blocking=False)
# Eval after training
if training_args.do_eval:
num_eval_samples = len(tokenized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size, drop_last=False)
eval_metrics = []
for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
# Model forward
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# get eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(lambda metric: jnp.mean(metric).item(), eval_metrics)
if jax.process_index() == 0:
eval_metrics = {f"eval_{metric_name}": value for metric_name, value in eval_metrics.items()}
path = os.path.join(training_args.output_dir, "eval_results.json")
with open(path, "w") as f:
json.dump(eval_metrics, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/run_clm_flax.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Pre-training/Fine-tuning the library models for causal language modeling (GPT, GPT-2, CTRL, ...) on a text file or a dataset.
Here is the full list of checkpoints on the hub that can be fine-tuned by this script:
https://huggingface.co/models?filter=text-generation
"""
# You can also adapt this script on your own causal language modeling task. Pointers for this are left as comments.
import json
import logging
import math
import os
import sys
import time
import warnings
from dataclasses import asdict, dataclass, field
from enum import Enum
from itertools import chain
from pathlib import Path
from typing import Callable, Optional
import datasets
import jax
import jax.numpy as jnp
import numpy as np
import optax
from datasets import Dataset, load_dataset
from flax import jax_utils, traverse_util
from flax.jax_utils import pad_shard_unpad, unreplicate
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard, shard_prng_key
from huggingface_hub import Repository, create_repo
from tqdm import tqdm
import transformers
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_CAUSAL_LM_MAPPING,
AutoConfig,
AutoTokenizer,
FlaxAutoModelForCausalLM,
HfArgumentParser,
is_tensorboard_available,
set_seed,
)
from transformers.testing_utils import CaptureLogger
from transformers.utils import send_example_telemetry
logger = logging.getLogger(__name__)
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_CAUSAL_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class TrainingArguments:
output_dir: str = field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."},
)
overwrite_output_dir: bool = field(
default=False,
metadata={
"help": (
"Overwrite the content of the output directory. "
"Use this to continue training if output_dir points to a checkpoint directory."
)
},
)
do_train: bool = field(default=False, metadata={"help": "Whether to run training."})
do_eval: bool = field(default=False, metadata={"help": "Whether to run eval on the dev set."})
per_device_train_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for training."}
)
per_device_eval_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for evaluation."}
)
learning_rate: float = field(default=5e-5, metadata={"help": "The initial learning rate for AdamW."})
weight_decay: float = field(default=0.0, metadata={"help": "Weight decay for AdamW if we apply some."})
adam_beta1: float = field(default=0.9, metadata={"help": "Beta1 for AdamW optimizer"})
adam_beta2: float = field(default=0.999, metadata={"help": "Beta2 for AdamW optimizer"})
adam_epsilon: float = field(default=1e-8, metadata={"help": "Epsilon for AdamW optimizer."})
adafactor: bool = field(default=False, metadata={"help": "Whether or not to replace AdamW by Adafactor."})
num_train_epochs: float = field(default=3.0, metadata={"help": "Total number of training epochs to perform."})
warmup_steps: int = field(default=0, metadata={"help": "Linear warmup over warmup_steps."})
logging_steps: int = field(default=500, metadata={"help": "Log every X updates steps."})
save_steps: int = field(default=500, metadata={"help": "Save checkpoint every X updates steps."})
eval_steps: int = field(default=None, metadata={"help": "Run an evaluation every X steps."})
seed: int = field(default=42, metadata={"help": "Random seed that will be set at the beginning of training."})
push_to_hub: bool = field(
default=False, metadata={"help": "Whether or not to upload the trained model to the model hub after training."}
)
hub_model_id: str = field(
default=None, metadata={"help": "The name of the repository to keep in sync with the local `output_dir`."}
)
hub_token: str = field(default=None, metadata={"help": "The token to use to push to the Model Hub."})
def __post_init__(self):
if self.output_dir is not None:
self.output_dir = os.path.expanduser(self.output_dir)
def to_dict(self):
"""
Serializes this instance while replace `Enum` by their values (for JSON serialization support). It obfuscates
the token values by removing their value.
"""
d = asdict(self)
for k, v in d.items():
if isinstance(v, Enum):
d[k] = v.value
if isinstance(v, list) and len(v) > 0 and isinstance(v[0], Enum):
d[k] = [x.value for x in v]
if k.endswith("_token"):
d[k] = f"<{k.upper()}>"
return d
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
)
},
)
use_auth_token: bool = field(
default=None,
metadata={
"help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
},
)
trust_remote_code: bool = field(
default=False,
metadata={
"help": (
"Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
"should only be set to `True` for repositories you trust and in which you have read the code, as it will "
"execute code present on the Hub on your local machine."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
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."
)
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
block_size: Optional[int] = field(
default=None,
metadata={
"help": (
"Optional input sequence length after tokenization. "
"The training dataset will be truncated in block of this size for training. "
"Default to the model max input length for single sentence inputs (take into account special tokens)."
)
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
keep_linebreaks: bool = field(
default=True, metadata={"help": "Whether to keep line breaks when using TXT files or not."}
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
if extension not in ["csv", "json", "txt"]:
raise ValueError("train_file` should be a csv, json or text file.")
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
if extension not in ["csv", "json", "txt"]:
raise ValueError("`validation_file` should be a csv, json or text file.")
class TrainState(train_state.TrainState):
dropout_rng: jnp.ndarray
def replicate(self):
return jax_utils.replicate(self).replace(dropout_rng=shard_prng_key(self.dropout_rng))
def data_loader(rng: jax.random.PRNGKey, dataset: Dataset, batch_size: int, shuffle: bool = False, drop_last=True):
"""
Returns batches of size `batch_size` from `dataset`. If `drop_last` is set to `False`, the final batch may be incomplete,
and range in size from 1 to `batch_size`. Shuffle batches if `shuffle` is `True`.
"""
if shuffle:
batch_idx = jax.random.permutation(rng, len(dataset))
batch_idx = np.asarray(batch_idx)
else:
batch_idx = np.arange(len(dataset))
if drop_last:
steps_per_epoch = len(dataset) // batch_size
batch_idx = batch_idx[: steps_per_epoch * batch_size] # Skip incomplete batch.
batch_idx = batch_idx.reshape((steps_per_epoch, batch_size))
else:
steps_per_epoch = math.ceil(len(dataset) / batch_size)
batch_idx = np.array_split(batch_idx, steps_per_epoch)
for idx in batch_idx:
batch = dataset[idx]
batch = {k: np.array(v) for k, v in batch.items()}
yield batch
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
def create_learning_rate_fn(
train_ds_size: int, train_batch_size: int, num_train_epochs: int, num_warmup_steps: int, learning_rate: float
) -> Callable[[int], jnp.ndarray]:
"""Returns a linear warmup, linear_decay learning rate function."""
steps_per_epoch = train_ds_size // train_batch_size
num_train_steps = steps_per_epoch * num_train_epochs
warmup_fn = optax.linear_schedule(init_value=0.0, end_value=learning_rate, transition_steps=num_warmup_steps)
decay_fn = optax.linear_schedule(
init_value=learning_rate, end_value=0, transition_steps=num_train_steps - num_warmup_steps
)
schedule_fn = optax.join_schedules(schedules=[warmup_fn, decay_fn], boundaries=[num_warmup_steps])
return schedule_fn
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 model_args.use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
FutureWarning,
)
if model_args.token is not None:
raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
model_args.token = model_args.use_auth_token
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_clm", model_args, data_args, framework="flax")
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."
)
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
# Setup logging, we only want one process per machine to log things on the screen.
logger.setLevel(logging.INFO if jax.process_index() == 0 else logging.ERROR)
if jax.process_index() == 0:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
# Set seed before initializing model.
set_seed(training_args.seed)
# Handle the repository creation
if training_args.push_to_hub:
# Retrieve of infer repo_name
repo_name = training_args.hub_model_id
if repo_name is None:
repo_name = Path(training_args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=training_args.hub_token).repo_id
# Clone repo locally
repo = Repository(training_args.output_dir, clone_from=repo_id, token=training_args.hub_token)
# 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 if no column called
# 'text' is found. You can easily tweak this behavior (see below).
#
# In distributed training, the load_dataset function guarantees 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.
dataset = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
cache_dir=model_args.cache_dir,
keep_in_memory=False,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
if "validation" not in dataset.keys():
dataset["validation"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
dataset["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
else:
data_files = {}
dataset_args = {}
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"
dataset_args["keep_linebreaks"] = data_args.keep_linebreaks
dataset = load_dataset(
extension,
data_files=data_files,
cache_dir=model_args.cache_dir,
**dataset_args,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
if "validation" not in dataset.keys():
dataset["validation"] = load_dataset(
extension,
data_files=data_files,
split=f"train[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
**dataset_args,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
dataset["train"] = load_dataset(
extension,
data_files=data_files,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
**dataset_args,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
# 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.
# 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,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
elif model_args.model_name_or_path:
config = AutoConfig.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
config = CONFIG_MAPPING[model_args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
if model_args.model_name_or_path:
model = FlaxAutoModelForCausalLM.from_pretrained(
model_args.model_name_or_path,
config=config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
model = FlaxAutoModelForCausalLM.from_config(
config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
trust_remote_code=model_args.trust_remote_code,
)
# Preprocessing the datasets.
# First we tokenize all the texts.
if training_args.do_train:
column_names = dataset["train"].column_names
else:
column_names = dataset["validation"].column_names
text_column_name = "text" if "text" in column_names else column_names[0]
# since this will be pickled to avoid _LazyModule error in Hasher force logger loading before tokenize_function
tok_logger = transformers.utils.logging.get_logger("transformers.tokenization_utils_base")
def tokenize_function(examples):
with CaptureLogger(tok_logger) as cl:
output = tokenizer(examples[text_column_name])
# clm input could be much much longer than block_size
if "Token indices sequence length is longer than the" in cl.out:
tok_logger.warning(
"^^^^^^^^^^^^^^^^ Please ignore the warning above - this long input will be chunked into smaller bits"
" before being passed to the model."
)
return output
tokenized_datasets = dataset.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.block_size is None:
block_size = tokenizer.model_max_length
if block_size > config.max_position_embeddings:
logger.warning(
f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
f"Using block_size={min(1024, config.max_position_embeddings)} instead. You can change that default value by passing --block_size xxx."
)
block_size = min(1024, config.max_position_embeddings)
else:
if data_args.block_size > tokenizer.model_max_length:
logger.warning(
f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model "
f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
)
block_size = min(data_args.block_size, tokenizer.model_max_length)
# Main data processing function that will concatenate all texts from our dataset and generate chunks of block_size.
def group_texts(examples):
# Concatenate all texts.
concatenated_examples = {k: list(chain(*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.
if total_length >= block_size:
total_length = (total_length // block_size) * block_size
# Split by chunks of max_len.
result = {
k: [t[i : i + block_size] for i in range(0, total_length, block_size)]
for k, t in concatenated_examples.items()
}
result["labels"] = result["input_ids"].copy()
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/process#map
lm_datasets = tokenized_datasets.map(
group_texts,
batched=True,
num_proc=data_args.preprocessing_num_workers,
load_from_cache_file=not data_args.overwrite_cache,
)
if training_args.do_train:
if "train" not in tokenized_datasets:
raise ValueError("--do_train requires a train dataset")
train_dataset = lm_datasets["train"]
if data_args.max_train_samples is not None:
max_train_samples = min(len(train_dataset), data_args.max_train_samples)
train_dataset = train_dataset.select(range(max_train_samples))
if training_args.do_eval:
if "validation" not in tokenized_datasets:
raise ValueError("--do_eval requires a validation dataset")
eval_dataset = lm_datasets["validation"]
if data_args.max_eval_samples is not None:
max_eval_samples = min(len(eval_dataset), data_args.max_eval_samples)
eval_dataset = eval_dataset.select(range(max_eval_samples))
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
rng, dropout_rng = jax.random.split(rng)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
eval_batch_size = per_device_eval_batch_size * jax.device_count()
steps_per_epoch = len(train_dataset) // train_batch_size
total_train_steps = steps_per_epoch * num_epochs
# Create learning rate schedule
linear_decay_lr_schedule_fn = create_learning_rate_fn(
len(train_dataset),
train_batch_size,
training_args.num_train_epochs,
training_args.warmup_steps,
training_args.learning_rate,
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
# find out all LayerNorm parameters
layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
layer_norm_named_params = {
layer[-2:]
for layer_norm_name in layer_norm_candidates
for layer in flat_params.keys()
if layer_norm_name in "".join(layer).lower()
}
flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
if training_args.adafactor:
# We use the default parameters here to initialize adafactor,
# For more details about the parameters please check https://github.com/deepmind/optax/blob/ed02befef9bf81cbbf236be3d2b0e032e9ed4a40/optax/_src/alias.py#L74
optimizer = optax.adafactor(
learning_rate=linear_decay_lr_schedule_fn,
)
else:
optimizer = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state
state = TrainState.create(apply_fn=model.__call__, params=model.params, tx=optimizer, dropout_rng=dropout_rng)
def loss_fn(logits, labels):
shift_logits = logits[..., :-1, :]
shift_labels = labels[..., 1:]
loss = optax.softmax_cross_entropy(shift_logits, onehot(shift_labels, shift_logits.shape[-1]))
return loss.mean()
# Define gradient update step fn
def train_step(state, batch):
dropout_rng, new_dropout_rng = jax.random.split(state.dropout_rng)
def compute_loss(params):
labels = batch.pop("labels")
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
loss = loss_fn(logits, labels)
return loss
grad_fn = jax.value_and_grad(compute_loss)
loss, grad = grad_fn(state.params)
grad = jax.lax.pmean(grad, "batch")
new_state = state.apply_gradients(grads=grad, dropout_rng=new_dropout_rng)
metrics = {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}
metrics = jax.lax.pmean(metrics, axis_name="batch")
return new_state, metrics
# Define eval fn
def eval_step(params, batch):
labels = batch.pop("labels")
logits = model(**batch, params=params, train=False)[0]
loss = loss_fn(logits, labels)
# summarize metrics
metrics = {"loss": loss}
metrics = jax.lax.pmean(metrics, axis_name="batch")
return metrics
# Create parallel version of the train and eval step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
p_eval_step = jax.pmap(eval_step, "batch")
# Replicate the train state on each device
state = state.replicate()
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {num_epochs}")
logger.info(f" Instantaneous batch size per device = {training_args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel & distributed) = {train_batch_size}")
logger.info(f" Total optimization steps = {total_train_steps}")
train_time = 0
train_metrics = []
epochs = tqdm(range(num_epochs), desc="Epoch ... ", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# Generate an epoch by shuffling sampling indices from the train dataset
train_loader = data_loader(input_rng, train_dataset, train_batch_size, shuffle=True)
steps_per_epoch = len(train_dataset) // train_batch_size
# train
for step in tqdm(range(steps_per_epoch), desc="Training...", position=1, leave=False):
batch = next(train_loader)
batch = shard(batch)
state, train_metric = p_train_step(state, batch)
train_metrics.append(train_metric)
cur_step = epoch * (len(train_dataset) // train_batch_size) + step
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_metric = unreplicate(train_metric)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step} | Loss: {train_metric['loss'].mean()}, Learning Rate:"
f" {train_metric['learning_rate'].mean()})"
)
train_metrics = []
if cur_step % training_args.eval_steps == 0 and cur_step > 0:
# ======================== Evaluating ==============================
eval_metrics = []
eval_loader = data_loader(input_rng, eval_dataset, eval_batch_size, drop_last=False)
eval_steps = math.ceil(len(eval_dataset) / eval_batch_size)
for _ in tqdm(range(eval_steps), desc="Evaluating...", position=2, leave=False):
# Model forward
batch = next(eval_loader)
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, batch, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.mean, eval_metrics)
try:
eval_metrics["perplexity"] = math.exp(eval_metrics["loss"])
except OverflowError:
eval_metrics["perplexity"] = float("inf")
# Print metrics and update progress bar
desc = (
f"Step... ({cur_step} | Eval Loss: {eval_metrics['loss']} | Eval Perplexity:"
f" {eval_metrics['perplexity']})"
)
epochs.write(desc)
epochs.desc = desc
# Save metrics
if has_tensorboard and jax.process_index() == 0:
write_eval_metric(summary_writer, eval_metrics, cur_step)
if cur_step % training_args.save_steps == 0 and cur_step > 0:
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(unreplicate(state.params))
model.save_pretrained(training_args.output_dir, params=params)
tokenizer.save_pretrained(training_args.output_dir)
if training_args.push_to_hub:
repo.push_to_hub(commit_message=f"Saving weights and logs of step {cur_step}", blocking=False)
# Eval after training
if training_args.do_eval:
eval_metrics = []
eval_loader = data_loader(input_rng, eval_dataset, eval_batch_size, drop_last=False)
eval_steps = math.ceil(len(eval_dataset) / eval_batch_size)
for _ in tqdm(range(eval_steps), desc="Evaluating...", position=2, leave=False):
# Model forward
batch = next(eval_loader)
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, batch, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(lambda x: jnp.mean(x).item(), eval_metrics)
try:
eval_metrics["perplexity"] = math.exp(eval_metrics["loss"])
except OverflowError:
eval_metrics["perplexity"] = float("inf")
if jax.process_index() == 0:
eval_metrics = {f"eval_{metric_name}": value for metric_name, value in eval_metrics.items()}
path = os.path.join(training_args.output_dir, "eval_results.json")
with open(path, "w") as f:
json.dump(eval_metrics, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/run_bart_dlm_flax.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Pretraining the library models for denoising language modeling on a text file or a dataset.
Here is the full list of checkpoints on the hub that can be pretrained by this script:
https://huggingface.co/models?filter=bart
"""
# You can also adapt this script on your own denoising language modeling task. Pointers for this are left as comments.
import json
import logging
import math
import os
import sys
import time
import warnings
from dataclasses import asdict, dataclass, field
from enum import Enum
from itertools import chain
from pathlib import Path
from typing import Dict, List, Optional
import flax
import jax
import jax.numpy as jnp
import nltk
import numpy as np
import optax
from datasets import load_dataset
from flax import jax_utils, traverse_util
from flax.jax_utils import pad_shard_unpad
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard
from huggingface_hub import Repository, create_repo
from tqdm import tqdm
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_MASKED_LM_MAPPING,
AutoTokenizer,
BartConfig,
BatchEncoding,
FlaxBartForConditionalGeneration,
HfArgumentParser,
PreTrainedTokenizerBase,
is_tensorboard_available,
set_seed,
)
from transformers.models.bart.modeling_flax_bart import shift_tokens_right
from transformers.utils import send_example_telemetry
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class TrainingArguments:
output_dir: str = field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."},
)
overwrite_output_dir: bool = field(
default=False,
metadata={
"help": (
"Overwrite the content of the output directory. "
"Use this to continue training if output_dir points to a checkpoint directory."
)
},
)
do_train: bool = field(default=False, metadata={"help": "Whether to run training."})
do_eval: bool = field(default=False, metadata={"help": "Whether to run eval on the dev set."})
per_device_train_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for training."}
)
per_device_eval_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for evaluation."}
)
learning_rate: float = field(default=5e-5, metadata={"help": "The initial learning rate for AdamW."})
weight_decay: float = field(default=0.0, metadata={"help": "Weight decay for AdamW if we apply some."})
adam_beta1: float = field(default=0.9, metadata={"help": "Beta1 for AdamW optimizer"})
adam_beta2: float = field(default=0.999, metadata={"help": "Beta2 for AdamW optimizer"})
adam_epsilon: float = field(default=1e-8, metadata={"help": "Epsilon for AdamW optimizer."})
adafactor: bool = field(default=False, metadata={"help": "Whether or not to replace AdamW by Adafactor."})
num_train_epochs: float = field(default=3.0, metadata={"help": "Total number of training epochs to perform."})
warmup_steps: int = field(default=0, metadata={"help": "Linear warmup over warmup_steps."})
logging_steps: int = field(default=500, metadata={"help": "Log every X updates steps."})
save_steps: int = field(default=500, metadata={"help": "Save checkpoint every X updates steps."})
eval_steps: int = field(default=None, metadata={"help": "Run an evaluation every X steps."})
seed: int = field(default=42, metadata={"help": "Random seed that will be set at the beginning of training."})
push_to_hub: bool = field(
default=False, metadata={"help": "Whether or not to upload the trained model to the model hub after training."}
)
hub_model_id: str = field(
default=None, metadata={"help": "The name of the repository to keep in sync with the local `output_dir`."}
)
hub_token: str = field(default=None, metadata={"help": "The token to use to push to the Model Hub."})
def __post_init__(self):
if self.output_dir is not None:
self.output_dir = os.path.expanduser(self.output_dir)
def to_dict(self):
"""
Serializes this instance while replace `Enum` by their values (for JSON serialization support). It obfuscates
the token values by removing their value.
"""
d = asdict(self)
for k, v in d.items():
if isinstance(v, Enum):
d[k] = v.value
if isinstance(v, list) and len(v) > 0 and isinstance(v[0], Enum):
d[k] = [x.value for x in v]
if k.endswith("_token"):
d[k] = f"<{k.upper()}>"
return d
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
)
},
)
use_auth_token: bool = field(
default=None,
metadata={
"help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
train_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input train ref data file for whole word masking in Chinese."},
)
validation_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input validation ref data file for whole word masking in Chinese."},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
max_seq_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total input sequence length after tokenization and masking. Sequences longer than this"
" will be truncated. Default to the max input length of the model."
)
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
mlm_probability: float = field(
default=0.3, metadata={"help": "Ratio of tokens to mask for span masked language modeling loss"}
)
permute_sentence_ratio: float = field(
default=1.0, metadata={"help": "Ratio of sentences to be permuted in each document"}
)
poisson_lambda: float = field(
default=3.0, metadata={"help": "Mean of Poisson distribution used to generate span-lengths to be masked"}
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
if extension not in ["csv", "json", "txt"]:
raise ValueError("train_file` should be a csv, json or text file.")
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
if extension not in ["csv", "json", "txt"]:
raise ValueError("`validation_file` should be a csv, json or text file.")
@flax.struct.dataclass
class FlaxDataCollatorForBartDenoisingLM:
"""
Data collator used for BART denoising language modeling. The code is largely copied from
`<https://github.com/morganmcg1/rotobart/blob/main/data_collator.py#L223>`__.
For more information on how BART denoising language modeling works, one can take a look
at the `official paper <https://arxiv.org/pdf/1910.13461.pdf>`__
or the `official code for preprocessing <https://github.com/facebookresearch/fairseq/blob/main/fairseq/data/denoising_dataset.py>`__ .
Args:
tokenizer (:class:`~transformers.PreTrainedTokenizer` or :class:`~transformers.PreTrainedTokenizerFast`):
The tokenizer used for encoding the data
mask_ratio (:obj:`float`):
The probability with which to (randomly) mask tokens in the input
poisson_lambda (:obj:`float`):
Mean parameter of Poisson distribution used to generate span-lengths to be masked
permute_sentence_ratio (:obj:`float`):
Ratio of sentences to be permuted in each document
decoder_start_token_id: (:obj:`int):
The decoder start token id of the model
"""
tokenizer: PreTrainedTokenizerBase
decoder_start_token_id: int
mask_ratio: float = 0.3
poisson_lambda: float = 3.0
permute_sentence_ratio: float = 1.0
def __post_init__(self):
if self.tokenizer.mask_token is None or self.tokenizer.eos_token is None:
raise ValueError(
"This tokenizer does not have a mask token or eos token token which is necessary for denoising"
" language modeling. "
)
def __call__(self, examples: List[Dict[str, List[int]]]) -> BatchEncoding:
# convert list to dict and tensorize input
batch = BatchEncoding(
{k: np.array([examples[i][k] for i in range(len(examples))]) for k, v in examples[0].items()}
)
batch["labels"] = batch["input_ids"].copy()
batch["decoder_input_ids"] = shift_tokens_right(
batch["labels"], self.tokenizer.pad_token_id, self.decoder_start_token_id
)
# permuting sentences
do_permute = False
if self.permute_sentence_ratio > 0.0:
batch["input_ids"] = self.permute_sentences(batch["input_ids"])
do_permute = True
# masking span of tokens (text infilling in the paper)
if self.mask_ratio:
batch["input_ids"], batch["labels"] = self.span_mask_tokens(
batch["input_ids"], batch["labels"], do_permute
)
# ignore pad tokens
batch["attention_mask"] = (batch["input_ids"] != self.tokenizer.pad_token_id).astype(int)
batch["decoder_attention_mask"] = (batch["decoder_input_ids"] != self.tokenizer.pad_token_id).astype(int)
return batch
def permute_sentences(self, input_ids):
"""
Shuffle sentences in each document.
"""
results = input_ids.copy()
# find end locations of sentences
end_sentence_mask = input_ids == self.tokenizer.pad_token_id
sentence_ends = np.argwhere(end_sentence_mask)
sentence_ends[:, 1] += 1
example_has_multiple_sentences, num_sentences = np.unique(sentence_ends[:, 0], return_counts=True)
num_sentences_map = dict(zip(example_has_multiple_sentences, num_sentences))
num_to_permute = np.ceil(num_sentences * self.permute_sentence_ratio).astype(int)
num_to_permute_map = dict(zip(example_has_multiple_sentences, num_to_permute))
sentence_ends = np.split(sentence_ends[:, 1], np.unique(sentence_ends[:, 0], return_index=True)[1][1:])
sentence_ends_map = dict(zip(example_has_multiple_sentences, sentence_ends))
for i in range(input_ids.shape[0]):
if i not in example_has_multiple_sentences:
continue
substitutions = np.random.permutation(num_sentences_map[i])[: num_to_permute_map[i]]
ordering = np.arange(0, num_sentences_map[i])
ordering[substitutions] = substitutions[np.random.permutation(num_to_permute_map[i])]
# write shuffled sentences into results
index = 0
for j in ordering:
sentence = input_ids[i, (sentence_ends_map[i][j - 1] if j > 0 else 0) : sentence_ends_map[i][j]]
results[i, index : index + sentence.shape[0]] = sentence
index += sentence.shape[0]
return results
def span_mask_tokens(self, input_ids, labels, do_permute):
"""
Sampling text spans with span lengths drawn from a Poisson distribution and masking them.
"""
special_tokens_mask_labels = [
self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
]
special_tokens_mask_inputs = [
self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in input_ids.tolist()
]
special_tokens_mask_labels = np.array(special_tokens_mask_labels, dtype=bool)
special_tokens_mask_inputs = np.array(special_tokens_mask_inputs, dtype=bool)
# determine how many tokens we need to mask in total
is_token_mask = ~(input_ids == self.tokenizer.pad_token_id) & ~special_tokens_mask_inputs
num_tokens_to_mask = int(math.ceil(is_token_mask.astype(float).sum() * self.mask_ratio))
if num_tokens_to_mask == 0:
return input_ids, labels
# generate a sufficient number of span lengths
span_lengths = np.random.poisson(lam=self.poisson_lambda, size=(num_tokens_to_mask,))
while np.cumsum(span_lengths, 0)[-1] < num_tokens_to_mask:
span_lengths = np.concatenate(
[span_lengths, np.random.poisson(lam=self.poisson_lambda, size=(num_tokens_to_mask,))]
)
# remove all spans of length 0
# note that BART inserts additional mask tokens where length == 0,
# which we do not implement for now as it adds additional complexity
span_lengths = span_lengths[span_lengths > 0]
# trim to about num_tokens_to_mask tokens
cutoff_idx = np.argmin(np.abs(np.cumsum(span_lengths, 0) - num_tokens_to_mask)) + 1
span_lengths = span_lengths[:cutoff_idx]
# randomly choose starting positions for masking
token_indices = np.argwhere(is_token_mask == 1)
span_starts = np.random.permutation(token_indices.shape[0])[: span_lengths.shape[0]]
# prepare mask
masked_indices = np.array(token_indices[span_starts])
mask = np.full_like(input_ids, fill_value=False)
# mask starting positions
for mi in masked_indices:
mask[tuple(mi)] = True
span_lengths -= 1
# fill up spans
max_index = input_ids.shape[1] - 1
remaining = (span_lengths > 0) & (masked_indices[:, 1] < max_index)
while np.any(remaining):
masked_indices[remaining, 1] += 1
for mi in masked_indices:
mask[tuple(mi)] = True
span_lengths -= 1
remaining = (span_lengths > 0) & (masked_indices[:, 1] < max_index)
# place the mask tokens
mask[np.where(special_tokens_mask_inputs)] = False
input_ids[np.where(mask)] = self.tokenizer.mask_token_id
if not do_permute:
labels[np.where(mask == 0)] = -100
else:
labels[np.where(special_tokens_mask_labels)] = -100
# remove mask tokens that are not starts of spans
to_remove = (mask == 1) & np.roll((mask == 1), 1, 1)
new_input_ids = np.full_like(input_ids, fill_value=self.tokenizer.pad_token_id)
for i, example in enumerate(input_ids):
new_example = example[~to_remove[i]]
new_input_ids[i, : new_example.shape[0]] = new_example
return new_input_ids, labels
def generate_batch_splits(samples_idx: np.ndarray, batch_size: int, drop_last=True) -> np.ndarray:
"""Generate batches of data for a specified batch size from sample indices. If the dataset size is not divisible by
the batch size and `drop_last` is `True`, the last incomplete batch is dropped. Else, it is returned."""
num_samples = len(samples_idx)
if drop_last:
samples_to_remove = num_samples % batch_size
if samples_to_remove != 0:
samples_idx = samples_idx[:-samples_to_remove]
sections_split = num_samples // batch_size
samples_idx = samples_idx.reshape((sections_split, batch_size))
else:
sections_split = math.ceil(num_samples / batch_size)
samples_idx = np.array_split(samples_idx, sections_split)
return samples_idx
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
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 model_args.use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
FutureWarning,
)
if model_args.token is not None:
raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
model_args.token = model_args.use_auth_token
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_bart_dlm", model_args, data_args, framework="flax")
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",
level=logging.INFO,
datefmt="[%X]",
)
# Log on each process the small summary:
logger = logging.getLogger(__name__)
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
# Set seed before initializing model.
set_seed(training_args.seed)
# Handle the repository creation
if training_args.push_to_hub:
# Retrieve of infer repo_name
repo_name = training_args.hub_model_id
if repo_name is None:
repo_name = Path(training_args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=training_args.hub_token).repo_id
# Clone repo locally
repo = Repository(training_args.output_dir, clone_from=repo_id, token=training_args.hub_token)
# 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 if no column called
# 'text' is found. You can easily tweak this behavior (see below).
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,
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
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}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
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,
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
if "validation" not in datasets.keys():
datasets["validation"] = load_dataset(
extension,
data_files=data_files,
split=f"train[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
datasets["train"] = load_dataset(
extension,
data_files=data_files,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
# 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.
# Load pretrained model and tokenizer
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
if model_args.config_name:
config = BartConfig.from_pretrained(
model_args.config_name,
cache_dir=model_args.cache_dir,
vocab_size=len(tokenizer),
token=model_args.token,
)
elif model_args.model_name_or_path:
config = BartConfig.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
token=model_args.token,
)
else:
config = CONFIG_MAPPING[model_args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
# 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]
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
# Use Punkt Sentence Tokenizer to divide a document into a list of sentences
nltk.download("punkt")
sentence_tokenizer = nltk.data.load("tokenizers/punkt/english.pickle")
def sentence_split_function(example):
sents = sentence_tokenizer.tokenize(example["text"])
# use pad token as end of sentence indicator
new_text = tokenizer.bos_token + f"{tokenizer.pad_token}".join(sents) + tokenizer.eos_token
return {"text": new_text}
split_datasets = datasets.map(
sentence_split_function,
batched=False,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
# Tokenize every text, then concatenate them together before splitting them in smaller parts.
# Since we make sure that all sequences are of the same length, no attention_mask is needed.
def tokenize_function(examples):
return tokenizer(examples[text_column_name], add_special_tokens=False, return_attention_mask=False)
tokenized_datasets = split_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,
)
# 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: list(chain(*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.
if total_length >= max_seq_length:
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/process#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,
)
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
dropout_rngs = jax.random.split(rng, jax.local_device_count())
if model_args.model_name_or_path:
model = FlaxBartForConditionalGeneration.from_pretrained(
model_args.model_name_or_path,
config=config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
token=model_args.token,
)
else:
config.vocab_size = len(tokenizer)
model = FlaxBartForConditionalGeneration(
config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
)
# Data collator
# This one will take care of randomly masking the tokens and permuting the sentences.
data_collator = FlaxDataCollatorForBartDenoisingLM(
tokenizer=tokenizer,
decoder_start_token_id=model.config.decoder_start_token_id,
mask_ratio=data_args.mlm_probability,
poisson_lambda=data_args.poisson_lambda,
permute_sentence_ratio=data_args.permute_sentence_ratio,
)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
eval_batch_size = per_device_eval_batch_size * jax.device_count()
num_train_steps = len(tokenized_datasets["train"]) // train_batch_size * num_epochs
# Create learning rate schedule
warmup_fn = optax.linear_schedule(
init_value=0.0, end_value=training_args.learning_rate, transition_steps=training_args.warmup_steps
)
decay_fn = optax.linear_schedule(
init_value=training_args.learning_rate,
end_value=0,
transition_steps=num_train_steps - training_args.warmup_steps,
)
linear_decay_lr_schedule_fn = optax.join_schedules(
schedules=[warmup_fn, decay_fn], boundaries=[training_args.warmup_steps]
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
# find out all LayerNorm parameters
layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
layer_norm_named_params = {
layer[-2:]
for layer_norm_name in layer_norm_candidates
for layer in flat_params.keys()
if layer_norm_name in "".join(layer).lower()
}
flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
if training_args.adafactor:
# We use the default parameters here to initialize adafactor,
# For more details about the parameters please check https://github.com/deepmind/optax/blob/ed02befef9bf81cbbf236be3d2b0e032e9ed4a40/optax/_src/alias.py#L74
optimizer = optax.adafactor(
learning_rate=linear_decay_lr_schedule_fn,
)
else:
optimizer = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state
state = train_state.TrainState.create(apply_fn=model.__call__, params=model.params, tx=optimizer)
# Define gradient update step fn
def train_step(state, batch, dropout_rng):
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
def loss_fn(params):
labels = batch.pop("labels")
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
# compute loss, ignore padded input tokens and special tokens
label_mask = jnp.where(labels > 0, 1.0, 0.0)
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
# take average
loss = loss.sum()
num_labels = label_mask.sum()
return loss, num_labels
grad_fn = jax.value_and_grad(loss_fn, has_aux=True)
(loss, num_labels), grad = grad_fn(state.params)
num_labels = jax.lax.psum(num_labels, "batch")
# true loss = total loss / total samples
loss = jax.lax.psum(loss, "batch")
loss = jax.tree_util.tree_map(lambda x: x / num_labels, loss)
# true grad = total grad / total samples
grad = jax.lax.psum(grad, "batch")
grad = jax.tree_util.tree_map(lambda x: x / num_labels, grad)
new_state = state.apply_gradients(grads=grad)
metrics = {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}
return new_state, metrics, new_dropout_rng
# Create parallel version of the train step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
# Define eval fn
def eval_step(params, batch):
labels = batch.pop("labels")
logits = model(**batch, params=params, train=False)[0]
# compute loss, ignore padded input tokens and special tokens
label_mask = jnp.where(labels > 0, 1.0, 0.0)
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
# compute accuracy
accuracy = jnp.equal(jnp.argmax(logits, axis=-1), labels) * label_mask
# summarize metrics
metrics = {"loss": loss.sum(), "accuracy": accuracy.sum(), "normalizer": label_mask.sum()}
metrics = jax.lax.psum(metrics, axis_name="batch")
return metrics
p_eval_step = jax.pmap(eval_step, "batch", donate_argnums=(0,))
# Replicate the train state on each device
state = jax_utils.replicate(state)
train_time = 0
epochs = tqdm(range(num_epochs), desc="Epoch ... ", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
train_metrics = []
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# Generate an epoch by shuffling sampling indices from the train dataset
num_train_samples = len(tokenized_datasets["train"])
# Avoid using jax.numpy here in case of TPU training
train_samples_idx = np.random.permutation(np.arange(num_train_samples))
train_batch_idx = generate_batch_splits(train_samples_idx, train_batch_size)
# Gather the indexes for creating the batch and do a training step
for step, batch_idx in enumerate(tqdm(train_batch_idx, desc="Training...", position=1)):
samples = [tokenized_datasets["train"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
# Model forward
model_inputs = shard(model_inputs.data)
state, train_metric, dropout_rngs = p_train_step(state, model_inputs, dropout_rngs)
train_metrics.append(train_metric)
cur_step = epoch * (num_train_samples // train_batch_size) + step
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_metric = jax_utils.unreplicate(train_metric)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step} | Loss: {train_metric['loss']}, Learning Rate:"
f" {train_metric['learning_rate']})"
)
train_metrics = []
if cur_step % training_args.eval_steps == 0 and cur_step > 0:
# ======================== Evaluating ==============================
num_eval_samples = len(tokenized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size)
eval_metrics = []
for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
# Model forward
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.sum, eval_metrics)
eval_normalizer = eval_metrics.pop("normalizer")
eval_metrics = jax.tree_util.tree_map(lambda x: x / eval_normalizer, eval_metrics)
# Update progress bar
epochs.desc = f"Step... ({cur_step} | Loss: {eval_metrics['loss']}, Acc: {eval_metrics['accuracy']})"
# Save metrics
if has_tensorboard and jax.process_index() == 0:
write_eval_metric(summary_writer, eval_metrics, cur_step)
if cur_step % training_args.save_steps == 0 and cur_step > 0:
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(jax.tree_util.tree_map(lambda x: x[0], state.params))
model.save_pretrained(training_args.output_dir, params=params)
tokenizer.save_pretrained(training_args.output_dir)
if training_args.push_to_hub:
repo.push_to_hub(commit_message=f"Saving weights and logs of step {cur_step}", blocking=False)
# Eval after training
if training_args.do_eval:
num_eval_samples = len(tokenized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size)
eval_metrics = []
for _, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
# Model forward
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(lambda metric: jnp.sum(metric).item(), eval_metrics)
eval_normalizer = eval_metrics.pop("normalizer")
eval_metrics = jax.tree_util.tree_map(lambda x: x / eval_normalizer, eval_metrics)
try:
perplexity = math.exp(eval_metrics["loss"])
except OverflowError:
perplexity = float("inf")
eval_metrics["perplexity"] = perplexity
if jax.process_index() == 0:
eval_metrics = {f"eval_{metric_name}": value for metric_name, value in eval_metrics.items()}
path = os.path.join(training_args.output_dir, "eval_results.json")
with open(path, "w") as f:
json.dump(eval_metrics, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/run_mlm_flax.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Fine-tuning the library models for masked language modeling (BERT, ALBERT, RoBERTa...) with whole word masking on a
text file or a dataset.
Here is the full list of checkpoints on the hub that can be fine-tuned by this script:
https://huggingface.co/models?filter=fill-mask
"""
import json
import logging
import math
import os
import sys
import time
import warnings
from dataclasses import asdict, dataclass, field
from enum import Enum
from itertools import chain
# You can also adapt this script on your own masked language modeling task. Pointers for this are left as comments.
from pathlib import Path
from typing import Dict, List, Optional, Tuple
import flax
import jax
import jax.numpy as jnp
import numpy as np
import optax
from datasets import load_dataset
from flax import jax_utils, traverse_util
from flax.jax_utils import pad_shard_unpad
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard
from huggingface_hub import Repository, create_repo
from tqdm import tqdm
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_MASKED_LM_MAPPING,
AutoConfig,
AutoTokenizer,
FlaxAutoModelForMaskedLM,
HfArgumentParser,
PreTrainedTokenizerBase,
TensorType,
is_tensorboard_available,
set_seed,
)
from transformers.utils import send_example_telemetry
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class TrainingArguments:
output_dir: str = field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."},
)
overwrite_output_dir: bool = field(
default=False,
metadata={
"help": (
"Overwrite the content of the output directory. "
"Use this to continue training if output_dir points to a checkpoint directory."
)
},
)
do_train: bool = field(default=False, metadata={"help": "Whether to run training."})
do_eval: bool = field(default=False, metadata={"help": "Whether to run eval on the dev set."})
per_device_train_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for training."}
)
per_device_eval_batch_size: int = field(
default=8, metadata={"help": "Batch size per GPU/TPU core/CPU for evaluation."}
)
learning_rate: float = field(default=5e-5, metadata={"help": "The initial learning rate for AdamW."})
weight_decay: float = field(default=0.0, metadata={"help": "Weight decay for AdamW if we apply some."})
adam_beta1: float = field(default=0.9, metadata={"help": "Beta1 for AdamW optimizer"})
adam_beta2: float = field(default=0.999, metadata={"help": "Beta2 for AdamW optimizer"})
adam_epsilon: float = field(default=1e-8, metadata={"help": "Epsilon for AdamW optimizer."})
adafactor: bool = field(default=False, metadata={"help": "Whether or not to replace AdamW by Adafactor."})
num_train_epochs: float = field(default=3.0, metadata={"help": "Total number of training epochs to perform."})
warmup_steps: int = field(default=0, metadata={"help": "Linear warmup over warmup_steps."})
logging_steps: int = field(default=500, metadata={"help": "Log every X updates steps."})
save_steps: int = field(default=500, metadata={"help": "Save checkpoint every X updates steps."})
eval_steps: int = field(default=None, metadata={"help": "Run an evaluation every X steps."})
seed: int = field(default=42, metadata={"help": "Random seed that will be set at the beginning of training."})
push_to_hub: bool = field(
default=False, metadata={"help": "Whether or not to upload the trained model to the model hub after training."}
)
hub_model_id: str = field(
default=None, metadata={"help": "The name of the repository to keep in sync with the local `output_dir`."}
)
hub_token: str = field(default=None, metadata={"help": "The token to use to push to the Model Hub."})
gradient_checkpointing: bool = field(
default=False,
metadata={
"help": "If True, use gradient checkpointing to save memory at the expense of slower backward pass."
},
)
def __post_init__(self):
if self.output_dir is not None:
self.output_dir = os.path.expanduser(self.output_dir)
def to_dict(self):
"""
Serializes this instance while replace `Enum` by their values (for JSON serialization support). It obfuscates
the token values by removing their value.
"""
d = asdict(self)
for k, v in d.items():
if isinstance(v, Enum):
d[k] = v.value
if isinstance(v, list) and len(v) > 0 and isinstance(v[0], Enum):
d[k] = [x.value for x in v]
if k.endswith("_token"):
d[k] = f"<{k.upper()}>"
return d
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
token: str = field(
default=None,
metadata={
"help": (
"The token to use as HTTP bearer authorization for remote files. If not specified, will use the token "
"generated when running `huggingface-cli login` (stored in `~/.huggingface`)."
)
},
)
use_auth_token: bool = field(
default=None,
metadata={
"help": "The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead."
},
)
trust_remote_code: bool = field(
default=False,
metadata={
"help": (
"Whether or not to allow for custom models defined on the Hub in their own modeling files. This option"
"should only be set to `True` for repositories you trust and in which you have read the code, as it will "
"execute code present on the Hub on your local machine."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
train_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input train ref data file for whole word masking in Chinese."},
)
validation_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input validation ref data file for whole word masking in Chinese."},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
max_seq_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated. Default to the max input length of the model."
)
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
mlm_probability: float = field(
default=0.15, metadata={"help": "Ratio of tokens to mask for masked language modeling loss"}
)
pad_to_max_length: bool = field(
default=False,
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."
)
},
)
line_by_line: bool = field(
default=False,
metadata={"help": "Whether distinct lines of text in the dataset are to be handled as distinct sequences."},
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file."
@flax.struct.dataclass
class FlaxDataCollatorForLanguageModeling:
"""
Data collator used for language modeling. Inputs are dynamically padded to the maximum length of a batch if they
are not all of the same length.
Args:
tokenizer (:class:`~transformers.PreTrainedTokenizer` or :class:`~transformers.PreTrainedTokenizerFast`):
The tokenizer used for encoding the data.
mlm_probability (:obj:`float`, `optional`, defaults to 0.15):
The probability with which to (randomly) mask tokens in the input.
.. note::
For best performance, this data collator should be used with a dataset having items that are dictionaries or
BatchEncoding, with the :obj:`"special_tokens_mask"` key, as returned by a
:class:`~transformers.PreTrainedTokenizer` or a :class:`~transformers.PreTrainedTokenizerFast` with the
argument :obj:`return_special_tokens_mask=True`.
"""
tokenizer: PreTrainedTokenizerBase
mlm_probability: float = 0.15
def __post_init__(self):
if self.tokenizer.mask_token is None:
raise ValueError(
"This tokenizer does not have a mask token which is necessary for masked language modeling. "
"You should pass `mlm=False` to train on causal language modeling instead."
)
def __call__(self, examples: List[Dict[str, np.ndarray]], pad_to_multiple_of: int) -> Dict[str, np.ndarray]:
# Handle dict or lists with proper padding and conversion to tensor.
batch = self.tokenizer.pad(examples, pad_to_multiple_of=pad_to_multiple_of, return_tensors=TensorType.NUMPY)
# If special token mask has been preprocessed, pop it from the dict.
special_tokens_mask = batch.pop("special_tokens_mask", None)
batch["input_ids"], batch["labels"] = self.mask_tokens(
batch["input_ids"], special_tokens_mask=special_tokens_mask
)
return batch
def mask_tokens(
self, inputs: np.ndarray, special_tokens_mask: Optional[np.ndarray]
) -> Tuple[np.ndarray, np.ndarray]:
"""
Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.
"""
labels = inputs.copy()
# We sample a few tokens in each sequence for MLM training (with probability `self.mlm_probability`)
probability_matrix = np.full(labels.shape, self.mlm_probability)
special_tokens_mask = special_tokens_mask.astype("bool")
probability_matrix[special_tokens_mask] = 0.0
masked_indices = np.random.binomial(1, probability_matrix).astype("bool")
labels[~masked_indices] = -100 # We only compute loss on masked tokens
# 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
indices_replaced = np.random.binomial(1, np.full(labels.shape, 0.8)).astype("bool") & masked_indices
inputs[indices_replaced] = self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token)
# 10% of the time, we replace masked input tokens with random word
indices_random = np.random.binomial(1, np.full(labels.shape, 0.5)).astype("bool")
indices_random &= masked_indices & ~indices_replaced
random_words = np.random.randint(self.tokenizer.vocab_size, size=labels.shape, dtype="i4")
inputs[indices_random] = random_words[indices_random]
# The rest of the time (10% of the time) we keep the masked input tokens unchanged
return inputs, labels
def generate_batch_splits(samples_idx: np.ndarray, batch_size: int, drop_last=True) -> np.ndarray:
"""Generate batches of data for a specified batch size from sample indices. If the dataset size is not divisible by
the batch size and `drop_last` is `True`, the last incomplete batch is dropped. Else, it is returned."""
num_samples = len(samples_idx)
if drop_last:
samples_to_remove = num_samples % batch_size
if samples_to_remove != 0:
samples_idx = samples_idx[:-samples_to_remove]
sections_split = num_samples // batch_size
samples_idx = samples_idx.reshape((sections_split, batch_size))
else:
sections_split = math.ceil(num_samples / batch_size)
samples_idx = np.array_split(samples_idx, sections_split)
return samples_idx
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
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 model_args.use_auth_token is not None:
warnings.warn(
"The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token` instead.",
FutureWarning,
)
if model_args.token is not None:
raise ValueError("`token` and `use_auth_token` are both specified. Please set only the argument `token`.")
model_args.token = model_args.use_auth_token
# Sending telemetry. Tracking the example usage helps us better allocate resources to maintain them. The
# information sent is the one passed as arguments along with your Python/PyTorch versions.
send_example_telemetry("run_mlm", model_args, data_args, framework="flax")
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",
level=logging.INFO,
datefmt="[%X]",
)
# Log on each process the small summary:
logger = logging.getLogger(__name__)
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
# Set seed before initializing model.
set_seed(training_args.seed)
# Handle the repository creation
if training_args.push_to_hub:
# Retrieve of infer repo_name
repo_name = training_args.hub_model_id
if repo_name is None:
repo_name = Path(training_args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=training_args.hub_token).repo_id
# Clone repo locally
repo = Repository(training_args.output_dir, clone_from=repo_id, token=training_args.hub_token)
# 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 if no column called
# 'text' is found. You can easily tweak this behavior (see below).
#
# In distributed training, the load_dataset function guarantees 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,
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
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}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
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,
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
if "validation" not in datasets.keys():
datasets["validation"] = load_dataset(
extension,
data_files=data_files,
split=f"train[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
datasets["train"] = load_dataset(
extension,
data_files=data_files,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
token=model_args.token,
num_proc=data_args.preprocessing_num_workers,
)
# 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.
# 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,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
elif model_args.model_name_or_path:
config = AutoConfig.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
config = CONFIG_MAPPING[model_args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=model_args.use_fast_tokenizer,
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
# 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]
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
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 = [line for line in examples if len(line) > 0 and not line.isspace()]
return tokenizer(
examples,
return_special_tokens_mask=True,
padding=padding,
truncation=True,
max_length=max_seq_length,
)
tokenized_datasets = datasets.map(
tokenize_function,
input_columns=[text_column_name],
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
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,
)
# 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: list(chain(*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.
if total_length >= max_seq_length:
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/process#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,
)
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Data collator
# This one will take care of randomly masking the tokens.
data_collator = FlaxDataCollatorForLanguageModeling(tokenizer=tokenizer, mlm_probability=data_args.mlm_probability)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
dropout_rngs = jax.random.split(rng, jax.local_device_count())
if model_args.model_name_or_path:
model = FlaxAutoModelForMaskedLM.from_pretrained(
model_args.model_name_or_path,
config=config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
token=model_args.token,
trust_remote_code=model_args.trust_remote_code,
)
else:
model = FlaxAutoModelForMaskedLM.from_config(
config,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
trust_remote_code=model_args.trust_remote_code,
)
if training_args.gradient_checkpointing:
model.enable_gradient_checkpointing()
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
per_device_eval_batch_size = int(training_args.per_device_eval_batch_size)
eval_batch_size = per_device_eval_batch_size * jax.device_count()
num_train_steps = len(tokenized_datasets["train"]) // train_batch_size * num_epochs
# Create learning rate schedule
warmup_fn = optax.linear_schedule(
init_value=0.0, end_value=training_args.learning_rate, transition_steps=training_args.warmup_steps
)
decay_fn = optax.linear_schedule(
init_value=training_args.learning_rate,
end_value=0,
transition_steps=num_train_steps - training_args.warmup_steps,
)
linear_decay_lr_schedule_fn = optax.join_schedules(
schedules=[warmup_fn, decay_fn], boundaries=[training_args.warmup_steps]
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
# find out all LayerNorm parameters
layer_norm_candidates = ["layernorm", "layer_norm", "ln"]
layer_norm_named_params = {
layer[-2:]
for layer_norm_name in layer_norm_candidates
for layer in flat_params.keys()
if layer_norm_name in "".join(layer).lower()
}
flat_mask = {path: (path[-1] != "bias" and path[-2:] not in layer_norm_named_params) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
if training_args.adafactor:
# We use the default parameters here to initialize adafactor,
# For more details about the parameters please check https://github.com/deepmind/optax/blob/ed02befef9bf81cbbf236be3d2b0e032e9ed4a40/optax/_src/alias.py#L74
optimizer = optax.adafactor(
learning_rate=linear_decay_lr_schedule_fn,
)
else:
optimizer = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state
state = train_state.TrainState.create(apply_fn=model.__call__, params=model.params, tx=optimizer)
# Define gradient update step fn
def train_step(state, batch, dropout_rng):
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
def loss_fn(params):
labels = batch.pop("labels")
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
# compute loss, ignore padded input tokens
label_mask = jnp.where(labels > 0, 1.0, 0.0)
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
# take average
loss = loss.sum()
num_labels = label_mask.sum()
return loss, num_labels
grad_fn = jax.value_and_grad(loss_fn, has_aux=True)
(loss, num_labels), grad = grad_fn(state.params)
num_labels = jax.lax.psum(num_labels, "batch")
# true loss = total loss / total samples
loss = jax.lax.psum(loss, "batch")
loss = jax.tree_util.tree_map(lambda x: x / num_labels, loss)
# true grad = total grad / total samples
grad = jax.lax.psum(grad, "batch")
grad = jax.tree_util.tree_map(lambda x: x / num_labels, grad)
new_state = state.apply_gradients(grads=grad)
metrics = {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}
return new_state, metrics, new_dropout_rng
# Create parallel version of the train step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
# Define eval fn
def eval_step(params, batch):
labels = batch.pop("labels")
logits = model(**batch, params=params, train=False)[0]
# compute loss, ignore padded input tokens
label_mask = jnp.where(labels > 0, 1.0, 0.0)
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
# compute accuracy
accuracy = jnp.equal(jnp.argmax(logits, axis=-1), labels) * label_mask
# summarize metrics
metrics = {"loss": loss.sum(), "accuracy": accuracy.sum(), "normalizer": label_mask.sum()}
metrics = jax.lax.psum(metrics, axis_name="batch")
return metrics
p_eval_step = jax.pmap(eval_step, "batch", donate_argnums=(0,))
# Replicate the train state on each device
state = jax_utils.replicate(state)
train_time = 0
epochs = tqdm(range(num_epochs), desc=f"Epoch ... (1/{num_epochs})", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
train_metrics = []
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# Generate an epoch by shuffling sampling indices from the train dataset
num_train_samples = len(tokenized_datasets["train"])
# Avoid using jax.numpy here in case of TPU training
train_samples_idx = np.random.permutation(np.arange(num_train_samples))
train_batch_idx = generate_batch_splits(train_samples_idx, train_batch_size)
# Gather the indexes for creating the batch and do a training step
for step, batch_idx in enumerate(tqdm(train_batch_idx, desc="Training...", position=1)):
samples = [tokenized_datasets["train"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples, pad_to_multiple_of=16)
# Model forward
model_inputs = shard(model_inputs.data)
state, train_metric, dropout_rngs = p_train_step(state, model_inputs, dropout_rngs)
train_metrics.append(train_metric)
cur_step = epoch * (num_train_samples // train_batch_size) + step
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_metric = jax_utils.unreplicate(train_metric)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step} | Loss: {train_metric['loss']}, Learning Rate:"
f" {train_metric['learning_rate']})"
)
train_metrics = []
if cur_step % training_args.eval_steps == 0 and cur_step > 0:
# ======================== Evaluating ==============================
num_eval_samples = len(tokenized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size, drop_last=False)
eval_metrics = []
for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples, pad_to_multiple_of=16)
# Model forward
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.sum, eval_metrics)
eval_normalizer = eval_metrics.pop("normalizer")
eval_metrics = jax.tree_util.tree_map(lambda x: x / eval_normalizer, eval_metrics)
# Update progress bar
epochs.desc = f"Step... ({cur_step} | Loss: {eval_metrics['loss']}, Acc: {eval_metrics['accuracy']})"
# Save metrics
if has_tensorboard and jax.process_index() == 0:
write_eval_metric(summary_writer, eval_metrics, cur_step)
if cur_step % training_args.save_steps == 0 and cur_step > 0:
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(jax.tree_util.tree_map(lambda x: x[0], state.params))
model.save_pretrained(training_args.output_dir, params=params)
tokenizer.save_pretrained(training_args.output_dir)
if training_args.push_to_hub:
repo.push_to_hub(commit_message=f"Saving weights and logs of step {cur_step}", blocking=False)
# Eval after training
if training_args.do_eval:
num_eval_samples = len(tokenized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size, drop_last=False)
eval_metrics = []
for _, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [tokenized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples, pad_to_multiple_of=16)
# Model forward
metrics = pad_shard_unpad(p_eval_step, static_return=True)(
state.params, model_inputs.data, min_device_batch=per_device_eval_batch_size
)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(lambda metric: jnp.sum(metric).item(), eval_metrics)
eval_normalizer = eval_metrics.pop("normalizer")
eval_metrics = jax.tree_util.tree_map(lambda x: x / eval_normalizer, eval_metrics)
try:
perplexity = math.exp(eval_metrics["loss"])
except OverflowError:
perplexity = float("inf")
eval_metrics["perplexity"] = perplexity
if jax.process_index() == 0:
eval_metrics = {f"eval_{metric_name}": value for metric_name, value in eval_metrics.items()}
path = os.path.join(training_args.output_dir, "eval_results.json")
with open(path, "w") as f:
json.dump(eval_metrics, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/t5_tokenizer_model.py
|
#!/usr/bin/env python3
import json
from typing import Iterator, List, Union
from tokenizers import AddedToken, Regex, Tokenizer, decoders, normalizers, pre_tokenizers, trainers
from tokenizers.implementations.base_tokenizer import BaseTokenizer
from tokenizers.models import Unigram
from tokenizers.processors import TemplateProcessing
class SentencePieceUnigramTokenizer(BaseTokenizer):
"""
This class is a copy of `DeDLOC's tokenizer implementation <https://github.com/yandex-research/DeDLOC/blob/main/sahajbert/tokenizer/tokenizer_model.py>`__ .
Custom SentencePiece Unigram Tokenizer with NMT, NKFC, spaces and lower-casing characters normalization
Represents the Unigram algorithm, with the pretokenization used by SentencePiece
"""
def __init__(
self,
replacement: str = "▁",
add_prefix_space: bool = True,
unk_token: Union[str, AddedToken] = "<unk>",
eos_token: Union[str, AddedToken] = "</s>",
pad_token: Union[str, AddedToken] = "<pad>",
):
self.special_tokens = {
"pad": {"id": 0, "token": pad_token},
"eos": {"id": 1, "token": eos_token},
"unk": {"id": 2, "token": unk_token},
}
self.special_tokens_list = [None] * len(self.special_tokens)
for token_dict in self.special_tokens.values():
self.special_tokens_list[token_dict["id"]] = token_dict["token"]
tokenizer = Tokenizer(Unigram())
tokenizer.normalizer = normalizers.Sequence(
[
normalizers.Nmt(),
normalizers.NFKC(),
normalizers.Replace(Regex(" {2,}"), " "),
normalizers.Lowercase(),
]
)
tokenizer.pre_tokenizer = pre_tokenizers.Sequence(
[
pre_tokenizers.Metaspace(replacement=replacement, add_prefix_space=add_prefix_space),
pre_tokenizers.Digits(individual_digits=True),
pre_tokenizers.Punctuation(),
]
)
tokenizer.decoder = decoders.Metaspace(replacement=replacement, add_prefix_space=add_prefix_space)
tokenizer.post_processor = TemplateProcessing(
single=f"$A {self.special_tokens['eos']['token']}",
special_tokens=[(self.special_tokens["eos"]["token"], self.special_tokens["eos"]["id"])],
)
parameters = {
"model": "SentencePieceUnigram",
"replacement": replacement,
"add_prefix_space": add_prefix_space,
}
super().__init__(tokenizer, parameters)
def train(
self,
files: Union[str, List[str]],
vocab_size: int = 8000,
show_progress: bool = True,
):
"""Train the model using the given files"""
trainer = trainers.UnigramTrainer(
vocab_size=vocab_size,
special_tokens=self.special_tokens_list,
show_progress=show_progress,
)
if isinstance(files, str):
files = [files]
self._tokenizer.train(files, trainer=trainer)
self.add_unk_id()
def train_from_iterator(
self,
iterator: Union[Iterator[str], Iterator[Iterator[str]]],
vocab_size: int = 8000,
show_progress: bool = True,
):
"""Train the model using the given iterator"""
trainer = trainers.UnigramTrainer(
vocab_size=vocab_size,
special_tokens=self.special_tokens_list,
show_progress=show_progress,
)
self._tokenizer.train_from_iterator(iterator, trainer=trainer)
self.add_unk_id()
def add_unk_id(self):
tokenizer_json = json.loads(self._tokenizer.to_str())
tokenizer_json["model"]["unk_id"] = self.special_tokens["unk"]["id"]
self._tokenizer = Tokenizer.from_str(json.dumps(tokenizer_json))
| 0
|
hf_public_repos/transformers/examples/flax
|
hf_public_repos/transformers/examples/flax/language-modeling/requirements.txt
|
datasets >= 1.1.3
jax>=0.2.8
jaxlib>=0.1.59
flax>=0.3.5
optax>=0.0.9
| 0
|
hf_public_repos/transformers/examples
|
hf_public_repos/transformers/examples/research_projects/README.md
|
<!---
Copyright 2020 The HuggingFace Team. 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.
-->
# Research projects
This folder contains various research projects using 🤗 Transformers. They are not maintained and require a specific
version of 🤗 Transformers that is indicated in the requirements file of each folder. Updating them to the most recent version of the library will require some work.
To use any of them, just run the command
```
pip install -r requirements.txt
```
inside the folder of your choice.
If you need help with any of those, contact the author(s), indicated at the top of the `README` of each folder.
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/vqgan-clip/loaders.py
|
import importlib
import torch
import yaml
from omegaconf import OmegaConf
from taming.models.vqgan import VQModel
def load_config(config_path, display=False):
config = OmegaConf.load(config_path)
if display:
print(yaml.dump(OmegaConf.to_container(config)))
return config
def load_vqgan(device, conf_path=None, ckpt_path=None):
if conf_path is None:
conf_path = "./model_checkpoints/vqgan_only.yaml"
config = load_config(conf_path, display=False)
model = VQModel(**config.model.params)
if ckpt_path is None:
ckpt_path = "./model_checkpoints/vqgan_only.pt"
sd = torch.load(ckpt_path, map_location=device)
if ".ckpt" in ckpt_path:
sd = sd["state_dict"]
model.load_state_dict(sd, strict=True)
model.to(device)
del sd
return model
def reconstruct_with_vqgan(x, model):
z, _, [_, _, indices] = model.encode(x)
print(f"VQGAN --- {model.__class__.__name__}: latent shape: {z.shape[2:]}")
xrec = model.decode(z)
return xrec
def get_obj_from_str(string, reload=False):
module, cls = string.rsplit(".", 1)
if reload:
module_imp = importlib.import_module(module)
importlib.reload(module_imp)
return getattr(importlib.import_module(module, package=None), cls)
def instantiate_from_config(config):
if "target" not in config:
raise KeyError("Expected key `target` to instantiate.")
return get_obj_from_str(config["target"])(**config.get("params", {}))
def load_model_from_config(config, sd, gpu=True, eval_mode=True):
model = instantiate_from_config(config)
if sd is not None:
model.load_state_dict(sd)
if gpu:
model.cuda()
if eval_mode:
model.eval()
return {"model": model}
def load_model(config, ckpt, gpu, eval_mode):
# load the specified checkpoint
if ckpt:
pl_sd = torch.load(ckpt, map_location="cpu")
global_step = pl_sd["global_step"]
print(f"loaded model from global step {global_step}.")
else:
pl_sd = {"state_dict": None}
global_step = None
model = load_model_from_config(config.model, pl_sd["state_dict"], gpu=gpu, eval_mode=eval_mode)["model"]
return model, global_step
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/vqgan-clip/README.md
|
# Simple VQGAN CLIP
Author: @ErwannMillon
This is a very simple VQGAN-CLIP implementation that was built as a part of the <a href= "https://github.com/ErwannMillon/face-editor"> Face Editor project </a> . This simplified version allows you to generate or edit images using text with just three lines of code. For a more full featured implementation with masking, more advanced losses, and a full GUI, check out the Face Editor project.
By default this uses a CelebA checkpoint (for generating/editing faces), but also has an imagenet checkpoint that can be loaded by specifying vqgan_config and vqgan_checkpoint when instantiating VQGAN_CLIP.
Learning rate and iterations can be set by modifying vqgan_clip.lr and vqgan_clip.iterations .
You can edit images by passing `image_path` to the generate function.
See the generate function's docstring to learn more about how to format prompts.
## Usage
The easiest way to test this out is by <a href="https://colab.research.google.com/drive/1Ez4D1J6-hVkmlXeR5jBPWYyu6CLA9Yor?usp=sharing
">using the Colab demo</a>
To install locally:
- Clone this repo
- Install git-lfs (ubuntu: sudo apt-get install git-lfs , MacOS: brew install git-lfs)
In the root of the repo run:
```
conda create -n vqganclip python=3.8
conda activate vqganclip
git-lfs install
git clone https://huggingface.co/datasets/erwann/face_editor_model_ckpt model_checkpoints
pip install -r requirements.txt
```
### Generate new images
```
from VQGAN_CLIP import VQGAN_CLIP
vqgan_clip = VQGAN_CLIP()
vqgan_clip.generate("a picture of a smiling woman")
```
### Edit an image
To get a test image, run
`git clone https://huggingface.co/datasets/erwann/vqgan-clip-pic test_images`
To edit:
```
from VQGAN_CLIP import VQGAN_CLIP
vqgan_clip = VQGAN_CLIP()
vqgan_clip.lr = .07
vqgan_clip.iterations = 15
vqgan_clip.generate(
pos_prompts= ["a picture of a beautiful asian woman", "a picture of a woman from Japan"],
neg_prompts=["a picture of an Indian person", "a picture of a white person"],
image_path="./test_images/face.jpeg",
show_intermediate=True,
save_intermediate=True,
)
```
### Make an animation from the most recent generation
`vqgan_clip.make_animation()`
## Features:
- Positive and negative prompts
- Multiple prompts
- Prompt Weights
- Creating GIF animations of the transformations
- Wandb logging
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/vqgan-clip/utils.py
|
from datetime import datetime
import matplotlib.pyplot as plt
import torch
def freeze_module(module):
for param in module.parameters():
param.requires_grad = False
def get_device():
device = "cuda" if torch.cuda.is_available() else "cpu"
if torch.backends.mps.is_available() and torch.backends.mps.is_built():
device = "mps"
if device == "mps":
print(
"WARNING: MPS currently doesn't seem to work, and messes up backpropagation without any visible torch"
" errors. I recommend using CUDA on a colab notebook or CPU instead if you're facing inexplicable issues"
" with generations."
)
return device
def show_pil(img):
fig = plt.imshow(img)
fig.axes.get_xaxis().set_visible(False)
fig.axes.get_yaxis().set_visible(False)
plt.show()
def get_timestamp():
current_time = datetime.now()
timestamp = current_time.strftime("%H:%M:%S")
return timestamp
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/vqgan-clip/img_processing.py
|
import numpy as np
import PIL
import torch
import torchvision.transforms as T
import torchvision.transforms.functional as TF
from PIL import Image
def preprocess(img, target_image_size=256):
s = min(img.size)
if s < target_image_size:
raise ValueError(f"min dim for image {s} < {target_image_size}")
r = target_image_size / s
s = (round(r * img.size[1]), round(r * img.size[0]))
img = TF.resize(img, s, interpolation=PIL.Image.LANCZOS)
img = TF.center_crop(img, output_size=2 * [target_image_size])
img = torch.unsqueeze(T.ToTensor()(img), 0)
return img
def preprocess_vqgan(x):
x = 2.0 * x - 1.0
return x
def custom_to_pil(x, process=True, mode="RGB"):
x = x.detach().cpu()
if process:
x = post_process_tensor(x)
x = x.numpy()
if process:
x = (255 * x).astype(np.uint8)
x = Image.fromarray(x)
if not x.mode == mode:
x = x.convert(mode)
return x
def post_process_tensor(x):
x = torch.clamp(x, -1.0, 1.0)
x = (x + 1.0) / 2.0
x = x.permute(1, 2, 0)
return x
def loop_post_process(x):
x = post_process_tensor(x.squeeze())
return x.permute(2, 0, 1).unsqueeze(0)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/vqgan-clip/requirements.txt
|
einops
gradio
icecream
imageio
lpips
matplotlib
more_itertools
numpy
omegaconf
opencv_python_headless
Pillow
pudb
pytorch_lightning
PyYAML
requests
scikit_image
scipy
setuptools
streamlit
taming-transformers
torch
torchvision
tqdm
transformers==4.26.0
tokenizers==0.13.2
typing_extensions
wandb
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/vqgan-clip/VQGAN_CLIP.py
|
import os
from glob import glob
import imageio
import torch
import torchvision
import wandb
from img_processing import custom_to_pil, loop_post_process, preprocess, preprocess_vqgan
from loaders import load_vqgan
from PIL import Image
from torch import nn
from transformers import CLIPModel, CLIPTokenizerFast
from utils import get_device, get_timestamp, show_pil
class ProcessorGradientFlow:
"""
This wraps the huggingface CLIP processor to allow backprop through the image processing step.
The original processor forces conversion to PIL images, which is faster for image processing but breaks gradient flow.
We call the original processor to get the text embeddings, but use our own image processing to keep images as torch tensors.
"""
def __init__(self, device: str = "cpu", clip_model: str = "openai/clip-vit-large-patch14") -> None:
self.device = device
self.tokenizer = CLIPTokenizerFast.from_pretrained(clip_model)
self.image_mean = [0.48145466, 0.4578275, 0.40821073]
self.image_std = [0.26862954, 0.26130258, 0.27577711]
self.normalize = torchvision.transforms.Normalize(self.image_mean, self.image_std)
self.resize = torchvision.transforms.Resize(224)
self.center_crop = torchvision.transforms.CenterCrop(224)
def preprocess_img(self, images):
images = self.resize(images)
images = self.center_crop(images)
images = self.normalize(images)
return images
def __call__(self, text=None, images=None, **kwargs):
encoding = self.tokenizer(text=text, **kwargs)
encoding["pixel_values"] = self.preprocess_img(images)
encoding = {key: value.to(self.device) for (key, value) in encoding.items()}
return encoding
class VQGAN_CLIP(nn.Module):
def __init__(
self,
iterations=10,
lr=0.01,
vqgan=None,
vqgan_config=None,
vqgan_checkpoint=None,
clip=None,
clip_preprocessor=None,
device=None,
log=False,
save_vector=True,
return_val="image",
quantize=True,
save_intermediate=False,
show_intermediate=False,
make_grid=False,
) -> None:
"""
Instantiate a VQGAN_CLIP model. If you want to use a custom VQGAN model, pass it as vqgan.
"""
super().__init__()
self.latent = None
self.device = device if device else get_device()
if vqgan:
self.vqgan = vqgan
else:
self.vqgan = load_vqgan(self.device, conf_path=vqgan_config, ckpt_path=vqgan_checkpoint)
self.vqgan.eval()
if clip:
self.clip = clip
else:
self.clip = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
self.clip.to(self.device)
self.clip_preprocessor = ProcessorGradientFlow(device=self.device)
self.iterations = iterations
self.lr = lr
self.log = log
self.make_grid = make_grid
self.return_val = return_val
self.quantize = quantize
self.latent_dim = self.vqgan.decoder.z_shape
def make_animation(self, input_path=None, output_path=None, total_duration=5, extend_frames=True):
"""
Make an animation from the intermediate images saved during generation.
By default, uses the images from the most recent generation created by the generate function.
If you want to use images from a different generation, pass the path to the folder containing the images as input_path.
"""
images = []
if output_path is None:
output_path = "./animation.gif"
if input_path is None:
input_path = self.save_path
paths = sorted(glob(input_path + "/*"))
if not len(paths):
raise ValueError(
"No images found in save path, aborting (did you pass save_intermediate=True to the generate"
" function?)"
)
if len(paths) == 1:
print("Only one image found in save path, (did you pass save_intermediate=True to the generate function?)")
frame_duration = total_duration / len(paths)
durations = [frame_duration] * len(paths)
if extend_frames:
durations[0] = 1.5
durations[-1] = 3
for file_name in paths:
if file_name.endswith(".png"):
images.append(imageio.imread(file_name))
imageio.mimsave(output_path, images, duration=durations)
print(f"gif saved to {output_path}")
def _get_latent(self, path=None, img=None):
if not (path or img):
raise ValueError("Input either path or tensor")
if img is not None:
raise NotImplementedError
x = preprocess(Image.open(path), target_image_size=256).to(self.device)
x_processed = preprocess_vqgan(x)
z, *_ = self.vqgan.encode(x_processed)
return z
def _add_vector(self, transform_vector):
"""Add a vector transform to the base latent and returns the resulting image."""
base_latent = self.latent.detach().requires_grad_()
trans_latent = base_latent + transform_vector
if self.quantize:
z_q, *_ = self.vqgan.quantize(trans_latent)
else:
z_q = trans_latent
return self.vqgan.decode(z_q)
def _get_clip_similarity(self, prompts, image, weights=None):
clip_inputs = self.clip_preprocessor(text=prompts, images=image, return_tensors="pt", padding=True)
clip_outputs = self.clip(**clip_inputs)
similarity_logits = clip_outputs.logits_per_image
if weights is not None:
similarity_logits = similarity_logits * weights
return similarity_logits.sum()
def _get_clip_loss(self, pos_prompts, neg_prompts, image):
pos_logits = self._get_clip_similarity(pos_prompts["prompts"], image, weights=(1 / pos_prompts["weights"]))
if neg_prompts:
neg_logits = self._get_clip_similarity(neg_prompts["prompts"], image, weights=neg_prompts["weights"])
else:
neg_logits = torch.tensor([1], device=self.device)
loss = -torch.log(pos_logits) + torch.log(neg_logits)
return loss
def _optimize_CLIP(self, original_img, pos_prompts, neg_prompts):
vector = torch.randn_like(self.latent, requires_grad=True, device=self.device)
optim = torch.optim.Adam([vector], lr=self.lr)
for i in range(self.iterations):
optim.zero_grad()
transformed_img = self._add_vector(vector)
processed_img = loop_post_process(transformed_img)
clip_loss = self._get_CLIP_loss(pos_prompts, neg_prompts, processed_img)
print("CLIP loss", clip_loss)
if self.log:
wandb.log({"CLIP Loss": clip_loss})
clip_loss.backward(retain_graph=True)
optim.step()
if self.return_val == "image":
yield custom_to_pil(transformed_img[0])
else:
yield vector
def _init_logging(self, positive_prompts, negative_prompts, image_path):
wandb.init(reinit=True, project="face-editor")
wandb.config.update({"Positive Prompts": positive_prompts})
wandb.config.update({"Negative Prompts": negative_prompts})
wandb.config.update({"lr": self.lr, "iterations": self.iterations})
if image_path:
image = Image.open(image_path)
image = image.resize((256, 256))
wandb.log("Original Image", wandb.Image(image))
def process_prompts(self, prompts):
if not prompts:
return []
processed_prompts = []
weights = []
if isinstance(prompts, str):
prompts = [prompt.strip() for prompt in prompts.split("|")]
for prompt in prompts:
if isinstance(prompt, (tuple, list)):
processed_prompt = prompt[0]
weight = float(prompt[1])
elif ":" in prompt:
processed_prompt, weight = prompt.split(":")
weight = float(weight)
else:
processed_prompt = prompt
weight = 1.0
processed_prompts.append(processed_prompt)
weights.append(weight)
return {
"prompts": processed_prompts,
"weights": torch.tensor(weights, device=self.device),
}
def generate(
self,
pos_prompts,
neg_prompts=None,
image_path=None,
show_intermediate=True,
save_intermediate=False,
show_final=True,
save_final=True,
save_path=None,
):
"""Generate an image from the given prompts.
If image_path is provided, the image is used as a starting point for the optimization.
If image_path is not provided, a random latent vector is used as a starting point.
You must provide at least one positive prompt, and optionally provide negative prompts.
Prompts must be formatted in one of the following ways:
- A single prompt as a string, e.g "A smiling woman"
- A set of prompts separated by pipes: "A smiling woman | a woman with brown hair"
- A set of prompts and their weights separated by colons: "A smiling woman:1 | a woman with brown hair: 3" (default weight is 1)
- A list of prompts, e.g ["A smiling woman", "a woman with brown hair"]
- A list of prompts and weights, e.g [("A smiling woman", 1), ("a woman with brown hair", 3)]
"""
if image_path:
self.latent = self._get_latent(image_path)
else:
self.latent = torch.randn(self.latent_dim, device=self.device)
if self.log:
self._init_logging(pos_prompts, neg_prompts, image_path)
assert pos_prompts, "You must provide at least one positive prompt."
pos_prompts = self.process_prompts(pos_prompts)
neg_prompts = self.process_prompts(neg_prompts)
if save_final and save_path is None:
save_path = os.path.join("./outputs/", "_".join(pos_prompts["prompts"]))
if not os.path.exists(save_path):
os.makedirs(save_path)
else:
save_path = save_path + "_" + get_timestamp()
os.makedirs(save_path)
self.save_path = save_path
original_img = self.vqgan.decode(self.latent)[0]
if show_intermediate:
print("Original Image")
show_pil(custom_to_pil(original_img))
original_img = loop_post_process(original_img)
for iter, transformed_img in enumerate(self._optimize_CLIP(original_img, pos_prompts, neg_prompts)):
if show_intermediate:
show_pil(transformed_img)
if save_intermediate:
transformed_img.save(os.path.join(self.save_path, f"iter_{iter:03d}.png"))
if self.log:
wandb.log({"Image": wandb.Image(transformed_img)})
if show_final:
show_pil(transformed_img)
if save_final:
transformed_img.save(os.path.join(self.save_path, f"iter_{iter:03d}_final.png"))
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/decision_transformer/run_decision_transformer.py
|
import gym
import numpy as np
import torch
from mujoco_py import GlfwContext
from transformers import DecisionTransformerModel
GlfwContext(offscreen=True) # Create a window to init GLFW.
def get_action(model, states, actions, rewards, returns_to_go, timesteps):
# we don't care about the past rewards in this model
states = states.reshape(1, -1, model.config.state_dim)
actions = actions.reshape(1, -1, model.config.act_dim)
returns_to_go = returns_to_go.reshape(1, -1, 1)
timesteps = timesteps.reshape(1, -1)
if model.config.max_length is not None:
states = states[:, -model.config.max_length :]
actions = actions[:, -model.config.max_length :]
returns_to_go = returns_to_go[:, -model.config.max_length :]
timesteps = timesteps[:, -model.config.max_length :]
# pad all tokens to sequence length
attention_mask = torch.cat(
[torch.zeros(model.config.max_length - states.shape[1]), torch.ones(states.shape[1])]
)
attention_mask = attention_mask.to(dtype=torch.long, device=states.device).reshape(1, -1)
states = torch.cat(
[
torch.zeros(
(states.shape[0], model.config.max_length - states.shape[1], model.config.state_dim),
device=states.device,
),
states,
],
dim=1,
).to(dtype=torch.float32)
actions = torch.cat(
[
torch.zeros(
(actions.shape[0], model.config.max_length - actions.shape[1], model.config.act_dim),
device=actions.device,
),
actions,
],
dim=1,
).to(dtype=torch.float32)
returns_to_go = torch.cat(
[
torch.zeros(
(returns_to_go.shape[0], model.config.max_length - returns_to_go.shape[1], 1),
device=returns_to_go.device,
),
returns_to_go,
],
dim=1,
).to(dtype=torch.float32)
timesteps = torch.cat(
[
torch.zeros(
(timesteps.shape[0], model.config.max_length - timesteps.shape[1]), device=timesteps.device
),
timesteps,
],
dim=1,
).to(dtype=torch.long)
else:
attention_mask = None
_, action_preds, _ = model(
states=states,
actions=actions,
rewards=rewards,
returns_to_go=returns_to_go,
timesteps=timesteps,
attention_mask=attention_mask,
return_dict=False,
)
return action_preds[0, -1]
# build the environment
env = gym.make("Hopper-v3")
state_dim = env.observation_space.shape[0]
act_dim = env.action_space.shape[0]
max_ep_len = 1000
device = "cuda"
scale = 1000.0 # normalization for rewards/returns
TARGET_RETURN = 3600 / scale # evaluation conditioning targets, 3600 is reasonable from the paper LINK
state_mean = np.array(
[
1.311279,
-0.08469521,
-0.5382719,
-0.07201576,
0.04932366,
2.1066856,
-0.15017354,
0.00878345,
-0.2848186,
-0.18540096,
-0.28461286,
]
)
state_std = np.array(
[
0.17790751,
0.05444621,
0.21297139,
0.14530419,
0.6124444,
0.85174465,
1.4515252,
0.6751696,
1.536239,
1.6160746,
5.6072536,
]
)
state_mean = torch.from_numpy(state_mean).to(device=device)
state_std = torch.from_numpy(state_std).to(device=device)
# Create the decision transformer model
model = DecisionTransformerModel.from_pretrained("edbeeching/decision-transformer-gym-hopper-medium")
model = model.to(device)
model.eval()
for ep in range(10):
episode_return, episode_length = 0, 0
state = env.reset()
target_return = torch.tensor(TARGET_RETURN, device=device, dtype=torch.float32).reshape(1, 1)
states = torch.from_numpy(state).reshape(1, state_dim).to(device=device, dtype=torch.float32)
actions = torch.zeros((0, act_dim), device=device, dtype=torch.float32)
rewards = torch.zeros(0, device=device, dtype=torch.float32)
timesteps = torch.tensor(0, device=device, dtype=torch.long).reshape(1, 1)
for t in range(max_ep_len):
env.render()
# add padding
actions = torch.cat([actions, torch.zeros((1, act_dim), device=device)], dim=0)
rewards = torch.cat([rewards, torch.zeros(1, device=device)])
action = get_action(
model,
(states.to(dtype=torch.float32) - state_mean) / state_std,
actions.to(dtype=torch.float32),
rewards.to(dtype=torch.float32),
target_return.to(dtype=torch.float32),
timesteps.to(dtype=torch.long),
)
actions[-1] = action
action = action.detach().cpu().numpy()
state, reward, done, _ = env.step(action)
cur_state = torch.from_numpy(state).to(device=device).reshape(1, state_dim)
states = torch.cat([states, cur_state], dim=0)
rewards[-1] = reward
pred_return = target_return[0, -1] - (reward / scale)
target_return = torch.cat([target_return, pred_return.reshape(1, 1)], dim=1)
timesteps = torch.cat([timesteps, torch.ones((1, 1), device=device, dtype=torch.long) * (t + 1)], dim=1)
episode_return += reward
episode_length += 1
if done:
break
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/decision_transformer/requirements.txt
|
absl-py==1.0.0
aiohttp==3.8.5
aiosignal==1.2.0
alembic==1.7.7
appdirs==1.4.4
APScheduler==3.9.1
arrow==1.2.2
asttokens==2.0.5
astunparse==1.6.3
async-timeout==4.0.2
attrs==21.4.0
audioread==2.1.9
autopage==0.5.0
backcall==0.2.0
backoff==1.11.1
backports.zoneinfo==0.2.1
binaryornot==0.4.4
black==22.1.0
boto3==1.16.34
botocore==1.19.63
Brotli==1.0.9
cachetools==5.0.0
certifi==2023.7.22
cffi==1.15.0
chardet==4.0.0
charset-normalizer==2.0.12
chex==0.1.1
click==8.0.4
cliff==3.10.1
clldutils==3.11.1
cloudpickle==2.0.0
cmaes==0.8.2
cmd2==2.4.0
codecarbon==1.2.0
colorlog==6.6.0
cookiecutter==2.1.1
cryptography==41.0.2
csvw==2.0.0
cycler==0.11.0
Cython==0.29.28
dash==2.3.0
dash-bootstrap-components==1.0.3
dash-core-components==2.0.0
dash-html-components==2.0.0
dash-table==5.0.0
datasets==2.0.0
decorator==5.1.1
Deprecated==1.2.13
dill==0.3.4
dlinfo==1.2.1
dm-tree==0.1.6
docker==4.4.4
execnet==1.9.0
executing==0.8.3
faiss-cpu==1.7.2
fasteners==0.17.3
filelock==3.6.0
fire==0.4.0
flake8==4.0.1
Flask==2.3.2
Flask-Compress==1.11
flatbuffers==2.0
flax==0.4.0
fonttools==4.31.1
frozenlist==1.3.0
fsspec==2022.2.0
fugashi==1.1.2
gast==0.5.3
gitdb==4.0.9
GitPython==3.1.32
glfw==2.5.1
google-auth==2.6.2
google-auth-oauthlib==0.4.6
google-pasta==0.2.0
greenlet==1.1.2
grpcio==1.44.0
gym==0.23.1
gym-notices==0.0.6
h5py==3.6.0
huggingface-hub==0.4.0
hypothesis==6.39.4
idna==3.3
imageio==2.16.1
importlib-metadata==4.11.3
importlib-resources==5.4.0
iniconfig==1.1.1
ipadic==1.0.0
ipython==8.10.0
isodate==0.6.1
isort==5.10.1
itsdangerous==2.1.1
jax==0.3.4
jaxlib==0.3.2
jedi==0.18.1
Jinja2==2.11.3
jinja2-time==0.2.0
jmespath==0.10.0
joblib==1.2.0
jsonschema==4.4.0
keras==2.8.0
Keras-Preprocessing==1.1.2
kiwisolver==1.4.0
kubernetes==12.0.1
libclang==13.0.0
librosa==0.9.1
llvmlite==0.38.0
Mako==1.2.2
Markdown==3.3.6
MarkupSafe==1.1.1
matplotlib==3.5.1
matplotlib-inline==0.1.3
mccabe==0.6.1
msgpack==1.0.3
mujoco-py==2.1.2.14
multidict==6.0.2
multiprocess==0.70.12.2
mypy-extensions==0.4.3
nltk==3.7
numba==0.55.1
numpy==1.22.3
oauthlib==3.2.2
onnx==1.13.0
onnxconverter-common==1.9.0
opt-einsum==3.3.0
optax==0.1.1
optuna==2.10.0
packaging==21.3
pandas==1.4.1
parameterized==0.8.1
parso==0.8.3
pathspec==0.9.0
pbr==5.8.1
pexpect==4.8.0
phonemizer==3.0.1
pickleshare==0.7.5
Pillow==10.0.1
Pint==0.16.1
plac==1.3.4
platformdirs==2.5.1
plotly==5.6.0
pluggy==1.0.0
pooch==1.6.0
portalocker==2.0.0
poyo==0.5.0
prettytable==3.2.0
prompt-toolkit==3.0.28
protobuf==3.19.5
psutil==5.9.0
ptyprocess==0.7.0
pure-eval==0.2.2
py==1.11.0
py-cpuinfo==8.0.0
pyarrow==7.0.0
pyasn1==0.4.8
pyasn1-modules==0.2.8
pycodestyle==2.8.0
pycparser==2.21
pyctcdecode==0.3.0
pyflakes==2.4.0
Pygments==2.15.0
pygtrie==2.4.2
pynvml==11.4.1
pyOpenSSL==22.0.0
pyparsing==3.0.7
pyperclip==1.8.2
pypng==0.0.21
pyrsistent==0.18.1
pytest==7.1.1
pytest-forked==1.4.0
pytest-timeout==2.1.0
pytest-xdist==2.5.0
python-dateutil==2.8.2
python-slugify==6.1.1
pytz==2022.1
pytz-deprecation-shim==0.1.0.post0
PyYAML==6.0
ray==1.11.0
redis==4.5.4
regex==2022.3.15
requests==2.31.0
requests-oauthlib==1.3.1
resampy==0.2.2
responses==0.18.0
rfc3986==1.5.0
rouge-score==0.0.4
rsa==4.8
s3transfer==0.3.7
sacrebleu==1.5.1
sacremoses==0.0.49
scikit-learn==1.0.2
scipy==1.8.0
segments==2.2.0
sentencepiece==0.1.96
sigopt==8.2.0
six==1.16.0
smmap==5.0.0
sortedcontainers==2.4.0
SoundFile==0.10.3.post1
SQLAlchemy==1.4.32
stack-data==0.2.0
stevedore==3.5.0
tabulate==0.8.9
tenacity==8.0.1
tensorboard==2.8.0
tensorboard-data-server==0.6.1
tensorboard-plugin-wit==1.8.1
tensorboardX==2.5
tensorflow==2.8.1
tensorflow-io-gcs-filesystem==0.24.0
termcolor==1.1.0
text-unidecode==1.3
tf-estimator-nightly==2.8.0.dev2021122109
tf2onnx==1.9.3
threadpoolctl==3.1.0
timeout-decorator==0.5.0
timm==0.5.4
tokenizers==0.11.6
tomli==2.0.1
toolz==0.11.2
torch==1.11.0
torchaudio==0.11.0
torchvision==0.12.0
tqdm==4.63.0
traitlets==5.1.1
-e git+git@github.com:edbeeching/transformers.git@77b90113ca0a0e4058b046796c874bdc98f1da61#egg=transformers
typing-extensions==4.1.1
tzdata==2022.1
tzlocal==4.1
unidic==1.1.0
unidic-lite==1.0.8
uritemplate==4.1.1
urllib3==1.26.18
wasabi==0.9.0
wcwidth==0.2.5
websocket-client==1.3.1
Werkzeug==3.0.1
wrapt==1.14.0
xxhash==3.0.0
yarl==1.7.2
zipp==3.7.0
| 0
|
hf_public_repos/transformers/examples/research_projects/onnx
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization/run_onnx_exporter.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright The HuggingFace Team and The HuggingFace Inc. team. 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.
"""
"""
import argparse
import logging
import os
import sys
import numpy as np
import onnxruntime
import torch
from bart_onnx.generation_onnx import BARTBeamSearchGenerator
from bart_onnx.reduce_onnx_size import remove_dup_initializers
import transformers
from transformers import BartForConditionalGeneration, BartTokenizer
logging.basicConfig(
format="%(asctime)s | %(levelname)s | %(name)s | [%(filename)s:%(lineno)d] %(message)s",
datefmt="%Y-%m-%d %H:%M:%S",
level=os.environ.get("LOGLEVEL", "INFO").upper(),
stream=sys.stdout,
)
logger = logging.getLogger(__name__)
model_dict = {"facebook/bart-base": BartForConditionalGeneration}
tokenizer_dict = {"facebook/bart-base": BartTokenizer}
def parse_args():
parser = argparse.ArgumentParser(description="Export Bart model + Beam Search to ONNX graph.")
parser.add_argument(
"--validation_file", type=str, default=None, help="A csv or a json file containing the validation data."
)
parser.add_argument(
"--max_length",
type=int,
default=5,
help="The maximum total input sequence length after tokenization.",
)
parser.add_argument(
"--num_beams",
type=int,
default=None,
help=(
"Number of beams to use for evaluation. This argument will be "
"passed to ``model.generate``, which is used during ``evaluate`` and ``predict``."
),
)
parser.add_argument(
"--model_name_or_path",
type=str,
help="Path to pretrained model or model identifier from huggingface.co/models.",
required=True,
)
parser.add_argument(
"--config_name",
type=str,
default=None,
help="Pretrained config name or path if not the same as model_name",
)
parser.add_argument(
"--device",
type=str,
default="cpu",
help="Device where the model will be run",
)
parser.add_argument("--output_file_path", type=str, default=None, help="Where to store the final ONNX file.")
args = parser.parse_args()
return args
def load_model_tokenizer(model_name, device="cpu"):
huggingface_model = model_dict[model_name].from_pretrained(model_name).to(device)
tokenizer = tokenizer_dict[model_name].from_pretrained(model_name)
if model_name in ["facebook/bart-base"]:
huggingface_model.config.no_repeat_ngram_size = 0
huggingface_model.config.forced_bos_token_id = None
huggingface_model.config.min_length = 0
return huggingface_model, tokenizer
def export_and_validate_model(model, tokenizer, onnx_file_path, num_beams, max_length):
model.eval()
ort_sess = None
bart_script_model = torch.jit.script(BARTBeamSearchGenerator(model))
with torch.no_grad():
ARTICLE_TO_SUMMARIZE = "My friends are cool but they eat too many carbs."
inputs = tokenizer([ARTICLE_TO_SUMMARIZE], max_length=1024, return_tensors="pt").to(model.device)
summary_ids = model.generate(
inputs["input_ids"],
attention_mask=inputs["attention_mask"],
num_beams=num_beams,
max_length=max_length,
early_stopping=True,
decoder_start_token_id=model.config.decoder_start_token_id,
)
torch.onnx.export(
bart_script_model,
(
inputs["input_ids"],
inputs["attention_mask"],
num_beams,
max_length,
model.config.decoder_start_token_id,
),
onnx_file_path,
opset_version=14,
input_names=["input_ids", "attention_mask", "num_beams", "max_length", "decoder_start_token_id"],
output_names=["output_ids"],
dynamic_axes={
"input_ids": {0: "batch", 1: "seq"},
"output_ids": {0: "batch", 1: "seq_out"},
},
example_outputs=summary_ids,
)
logger.info("Model exported to {}".format(onnx_file_path))
new_onnx_file_path = remove_dup_initializers(os.path.abspath(onnx_file_path))
logger.info("Deduplicated and optimized model written to {}".format(new_onnx_file_path))
ort_sess = onnxruntime.InferenceSession(new_onnx_file_path)
ort_out = ort_sess.run(
None,
{
"input_ids": inputs["input_ids"].cpu().numpy(),
"attention_mask": inputs["attention_mask"].cpu().numpy(),
"num_beams": np.array(num_beams),
"max_length": np.array(max_length),
"decoder_start_token_id": np.array(model.config.decoder_start_token_id),
},
)
np.testing.assert_allclose(summary_ids.cpu().numpy(), ort_out[0], rtol=1e-3, atol=1e-3)
logger.info("Model outputs from torch and ONNX Runtime are similar.")
logger.info("Success.")
def main():
args = parse_args()
max_length = 5
num_beams = 4
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
logger.setLevel(logging.INFO)
transformers.utils.logging.set_verbosity_error()
device = torch.device(args.device)
model, tokenizer = load_model_tokenizer(args.model_name_or_path, device)
if model.config.decoder_start_token_id is None:
raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")
model.to(device)
if args.max_length:
max_length = args.max_length
if args.num_beams:
num_beams = args.num_beams
if args.output_file_path:
output_name = args.output_file_path
else:
output_name = "BART.onnx"
logger.info("Exporting model to ONNX")
export_and_validate_model(model, tokenizer, output_name, num_beams, max_length)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/onnx
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization/README.md
|
<!---
Copyright 2021 The HuggingFace Team. 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.
-->
# Bart + Beam Search to ONNX
Author: [@fatcat-z](https://github.com/fatcat-z)
This folder contains an example of exporting Bart + Beam Search generation (`BartForConditionalGeneration`) to ONNX.
Beam Search contains a for-loop workflow, so we need to make them TorchScript-compatible for exporting to ONNX. This example shows how to make a Bart model be TorchScript-compatible by wrapping up it into a new model. In addition, some changes were made to the `beam_search()` function to make it TorchScript-compatible.
## How to run the example
To make sure you can successfully run the latest versions of the example scripts, you have to **install the library from source** and install some example-specific requirements. To do this, execute the following steps in a new virtual environment:
```bash
git clone https://github.com/huggingface/transformers
cd transformers
pip install '.[onnxruntime]'
```
Then cd in this example folder and run
```bash
pip install -r requirements.txt
```
Now you can run the example command below to get the example ONNX file:
```bash
python run_onnx_exporter.py --model_name_or_path facebook/bart-base
```
| 0
|
hf_public_repos/transformers/examples/research_projects/onnx
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization/requirements.txt
|
torch >= 1.10
| 0
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization/bart_onnx/generation_onnx.py
|
import copy
import itertools
from typing import List, Optional, Tuple
import torch
import torch.nn.functional as F
from transformers import BartConfig
from transformers.generation import GenerationMixin
def _convert_past_list_to_tuple(past_key_values):
"""
In Bart model, the type of past_key_values is tuple(tuple(torch.FloatTensor)) which is not
TorchScript-compatible. To support this, we have to convert it during the export process.
This function will convert past values from a list to tuple(tuple(torch.FloatTensor)) for
the inner decoder.
According to the definition of past_key_values, each inner tuple(torch.FloatTensor) has 4 tensors,
so we convert every 4 elements in the list as a tuple(torch.FloatTensor).
"""
count_of_each_inner_tuple = 4
results = ()
temp_result = ()
count_n = len(past_key_values) // count_of_each_inner_tuple
for idx in range(count_n):
real_idx = idx * count_of_each_inner_tuple
temp_result = tuple(past_key_values[real_idx : real_idx + count_of_each_inner_tuple])
results += ((temp_result),)
return results
class EncoderForONNX(torch.nn.Module):
def __init__(self, encoder):
super().__init__()
self.encoder = encoder
def forward(self, input_ids, attention_mask):
return self.encoder(
input_ids=input_ids,
attention_mask=attention_mask,
return_dict=False,
)
class DecoderForONNX(torch.nn.Module):
def __init__(self, decoder):
super().__init__()
self.decoder = decoder
def forward(self, input_ids, encoder_state, attention_mask, past=None):
all_results = None
if past is not None:
all_results = _convert_past_list_to_tuple(past)
input_ids = input_ids[:, -1:]
last_hidden_state, past_key_values = self.decoder(
input_ids=input_ids,
encoder_hidden_states=encoder_state,
encoder_attention_mask=attention_mask,
past_key_values=all_results,
return_dict=False,
)
past_values = []
for past in past_key_values:
past_values = past_values + list(past)
return last_hidden_state, past_values
def _create_traced_encoder(encoder, input_ids, attention_mask):
encoder_c = copy.deepcopy(encoder)
encoder_for_onnx = EncoderForONNX(encoder_c)
return torch.jit.trace(encoder_for_onnx, (input_ids, attention_mask))
def _create_traced_decoder(decoder, input_ids, encoder_state, attention_mask, past=None):
decoder_c = copy.deepcopy(decoder)
decoder_for_onnx = DecoderForONNX(decoder_c)
past_values = list(itertools.chain.from_iterable(past or ()))
# Do this twice so we got 2 different decoders for further work.
if past_values:
return torch.jit.trace(decoder_for_onnx, (input_ids, encoder_state, attention_mask, past_values))
else:
return torch.jit.trace(decoder_for_onnx, (input_ids, encoder_state, attention_mask))
class BartConfigTS(BartConfig, torch.nn.Module):
"""
BartConfigTS is a TorchScript-compatible transformers.models.bart.configuration_bart.BartConfig.
TorchScript only supports sub-classes of torch.nn.Module.
"""
def __init__(self, config):
BartConfig.__init__(self, config)
torch.nn.Module.__init__(self)
class MinLengthLogitsProcessorTS(torch.nn.Module):
r"""
:class:`transformers.LogitsProcessor` enforcing a min-length by setting EOS probability to 0.
Args:
min_length (:obj:`int`):
The minimum length below which the score of :obj:`eos_token_id` is set to :obj:`-float("Inf")`.
eos_token_id (:obj:`int`):
The id of the `end-of-sequence` token.
"""
def __init__(self, min_length: int, eos_token_id: int):
super().__init__()
if not isinstance(min_length, int) or min_length < 0:
raise ValueError(f"`min_length` has to be a positive integer, but is {min_length}")
if not isinstance(eos_token_id, int) or eos_token_id < 0:
raise ValueError(f"`eos_token_id` has to be a positive integer, but is {eos_token_id}")
self.min_length = min_length
self.eos_token_id = eos_token_id
def forward(self, input_ids, scores) -> torch.Tensor:
cur_len = input_ids.shape[-1]
if cur_len < self.min_length:
scores[:, self.eos_token_id] = -float("inf")
return scores
class BARTGenerator(torch.nn.Module, GenerationMixin):
def __init__(self, model):
super().__init__()
self.config = BartConfigTS(model.config)
self.config.force_bos_token_to_be_generated = False
self._trace_modules(model)
self.logits_processor = MinLengthLogitsProcessorTS(self.config.min_length, self.config.eos_token_id)
self.final_logits_weight = model.model.shared.weight
self.final_logits_bias = model.final_logits_bias
self.decoder_layers = model.config.decoder_layers
def _trace_modules(self, model):
input_ids = torch.tensor(
[
[
19,
669,
18,
420,
8,
664,
57,
42,
8,
664,
21,
3028,
195,
4445,
331,
1293,
34,
21,
10,
6174,
1100,
6,
69,
104,
42,
32,
2621,
1638,
144,
4,
6174,
558,
108,
4419,
1091,
28,
4,
1668,
9,
1509,
1621,
279,
35,
867,
2734,
85,
11,
2216,
2734,
85,
203,
2244,
7,
6,
15,
8102,
7,
57,
8629,
5,
model.config.eos_token_id,
]
],
device=model.device,
dtype=torch.long,
)
attention_mask = torch.tensor(
[[True] * input_ids.shape[-1]],
device=model.device,
dtype=torch.bool,
)
self.encoder = _create_traced_encoder(model.get_encoder(), input_ids, attention_mask)
encoder_outputs = model.get_encoder()(input_ids, attention_mask=attention_mask, return_dict=True)
decoder = model.model.decoder
decoder_outputs = decoder(input_ids, attention_mask, encoder_outputs["last_hidden_state"], None, None, None)
self.decoder_no_past = _create_traced_decoder(
model.model.decoder, input_ids, encoder_outputs["last_hidden_state"], attention_mask
)
self.decoder_with_past = _create_traced_decoder(
model.model.decoder, input_ids, encoder_outputs["last_hidden_state"], attention_mask, decoder_outputs[1]
)
def _encoder_forward(self, input_ids, attention_mask):
return self.encoder(input_ids, attention_mask)[0]
@staticmethod
def _init_sequence_length_for_generation(
input_ids: torch.LongTensor, max_length: int
) -> Tuple[torch.Tensor, torch.Tensor, int]:
unfinished_sequences = torch.zeros(input_ids.shape[0], dtype=torch.long, device=input_ids.device) + 1
sequence_lengths = torch.zeros(input_ids.shape[0], dtype=torch.long, device=input_ids.device) + max_length
cur_len = input_ids.shape[-1]
return sequence_lengths, unfinished_sequences, cur_len
def _decoder_forward(self, input_ids, encoder_output, attention_mask, past: List[torch.Tensor]):
# Update here to use different decoder for different values of past.
if past is None or len(past) == 0:
decoder_output, past = self.decoder_no_past(
input_ids=input_ids, encoder_state=encoder_output, attention_mask=attention_mask
)
else:
decoder_output, past = self.decoder_with_past(
input_ids=input_ids, encoder_state=encoder_output, attention_mask=attention_mask, past=past
)
lm_logits = F.linear(decoder_output, self.final_logits_weight, bias=self.final_logits_bias)
return lm_logits, past
def greedy_search(
self, input_ids, encoder_output, attention_mask, max_length, pad_token_id: int, eos_token_id: int
):
# init sequence length tensors
sequence_lengths, unfinished_sequences, cur_len = self._init_sequence_length_for_generation(
input_ids, max_length
)
past: List[torch.Tensor] = []
while cur_len < max_length:
logits, past = self._decoder_forward(input_ids, encoder_output, attention_mask, past)
next_token_logits = logits[:, -1, :]
# pre-process distribution
scores = self.logits_processor(input_ids, next_token_logits)
# argmax
next_tokens = torch.argmax(scores, dim=-1)
# add code that transfomers next_tokens to tokens_to_add
if eos_token_id is not None:
assert pad_token_id is not None, "If eos_token_id is defined, make sure that pad_token_id is defined."
next_tokens = next_tokens * unfinished_sequences + (pad_token_id) * (1 - unfinished_sequences)
# add token and increase length by one
input_ids = torch.cat([input_ids, next_tokens[:, None]], dim=-1)
# update sequence length
if eos_token_id is not None:
sequence_lengths, unfinished_sequences = self._update_seq_length_for_generation(
sequence_lengths, unfinished_sequences, cur_len, next_tokens == eos_token_id
)
# stop when there is a </s> in each sentence, or if we exceed the maximul length
if unfinished_sequences.max() == 0:
break
# increase cur_len
cur_len = cur_len + 1
return input_ids
def _prepare_decoder_input_ids_for_generation(
self,
input_ids: torch.LongTensor,
decoder_start_token_id,
bos_token_id: Optional[int] = None,
) -> torch.LongTensor:
decoder_input_ids = (
torch.ones((input_ids.shape[0], 1), dtype=input_ids.dtype, device=input_ids.device)
* decoder_start_token_id
)
return decoder_input_ids
def forward(self, input_ids, attention_mask, max_length, decoder_start_token_id):
pad_token_id = self.config.pad_token_id
bos_token_id = self.config.bos_token_id
eos_token_id = self.config.eos_token_id
# special case if pad_token_id is not defined
if pad_token_id is None and eos_token_id is not None:
# Setting `pad_token_id` to `eos_token_id`:{eos_token_id} for open-end generation.
pad_token_id = eos_token_id
encoder_output = self._encoder_forward(input_ids, attention_mask)
input_ids = self._prepare_decoder_input_ids_for_generation(
input_ids,
decoder_start_token_id=decoder_start_token_id,
bos_token_id=bos_token_id,
)
return self.greedy_search(
input_ids,
encoder_output,
attention_mask,
max_length=max_length,
pad_token_id=pad_token_id,
eos_token_id=eos_token_id,
)
# TorchScript compatible BeamSearchScorer
class BeamSearchScorerTS(torch.nn.Module):
def __init__(self):
super().__init__()
self.max_length: int = 200
self.num_beams: int = 3
self.batch_size: int = 1
self.length_penalty: float = 1.0
self.do_early_stopping: bool = True
self.num_beam_hyps_to_keep: int = 1
self.num_beam_groups: int = 1
self.group_size: int = self.num_beams // self.num_beam_groups
self._done = torch.zeros(self.batch_size, dtype=torch.bool)
self._beam_hyps_count = torch.zeros(self.batch_size, dtype=torch.long)
self._beam_hyps_worst_scores = torch.zeros(self.batch_size) + 1e9
self._beam_hyps_max_length: int = self.max_length - 1
self._beam_hyps: List[torch.Tensor] = [torch.zeros(2)] # placeholder for TorchScript compatibility
self._beam_scores: List[torch.Tensor] = [torch.zeros(2)] # placeholder for TorchScript compatibility
def is_done(self) -> torch.Tensor:
return self._done.all()
def init(
self,
batch_size: int,
max_length: int,
num_beams: int,
device: torch.device,
length_penalty: float = 1.0,
do_early_stopping: bool = False,
num_beam_hyps_to_keep: int = 1,
num_beam_groups: int = 1,
):
self.max_length = max_length
self.num_beams = num_beams
self.batch_size = batch_size
self.length_penalty = length_penalty
self.do_early_stopping = do_early_stopping
self.num_beam_hyps_to_keep = num_beam_hyps_to_keep
self.num_beam_groups = num_beam_groups
self.group_size = self.num_beams // self.num_beam_groups
# NOTE: TorchScript does not support List of Modules
# Rewritten BeamHypotheses with tensors and list of tensors.
self._done = torch.zeros(batch_size, dtype=torch.bool, device=device)
self._beam_hyps_count = torch.zeros(batch_size, dtype=torch.long, device=device)
self._beam_hyps_worst_scores = torch.zeros(batch_size, device=device) + 1e9
self._beam_hyps = []
self._beam_scores = []
self._beam_hyps_max_length = max_length - 1 # ignoring bos_token
if not isinstance(num_beams, int) or num_beams <= 1:
raise ValueError(
f"`num_beams` has to be an integer strictly greater than 1, but is {num_beams}. For `num_beams` == 1,"
" one should make use of `greedy_search` instead."
)
if not isinstance(num_beam_groups, int) or (num_beam_groups > num_beams) or (num_beams % num_beam_groups != 0):
raise ValueError(
"`num_beam_groups` has to be an integer smaller or equal than `num_beams` and `num_beams` has to be"
f" divisible by `num_beam_groups`, but is {num_beam_groups} with `num_beams` being {num_beams}."
)
def hypo_len(self, hypo_idx: int):
"""
Number of hypotheses in the list.
"""
return self._beam_hyps_count[hypo_idx]
def hypo_add(self, hyp: torch.Tensor, sum_logprobs: float, hypo_idx: int):
"""
Add a new hypothesis to the list.
"""
score = sum_logprobs / (hyp.shape[-1] ** self.length_penalty)
hyps_count = self.hypo_len(hypo_idx)
if hyps_count < self.num_beams or score > self._beam_hyps_worst_scores[hypo_idx]:
# NOTE: work around difference of torch.sum(empty_tensor) == 0, while error in onnx.
# Bug: https://msdata.visualstudio.com/Vienna/_workitems/edit/1486599
beam_idx = (
torch.sum(self._beam_hyps_count[:hypo_idx]) if hypo_idx != 0 else torch.tensor(0, dtype=torch.long)
)
self._beam_scores.insert(beam_idx, torch.tensor([score]))
self._beam_hyps.insert(beam_idx, hyp)
if hyps_count + 1 > self.num_beams:
sorted_next_scores, sorted_indices = torch.topk(
torch.cat(self._beam_scores)[beam_idx : beam_idx + hyps_count + 1], hyps_count + 1, largest=False
)
del self._beam_hyps[int((sorted_indices[0] + beam_idx))]
del self._beam_scores[int((sorted_indices[0] + beam_idx))]
self._beam_hyps_worst_scores[hypo_idx] = sorted_next_scores[1]
else:
self._beam_hyps_worst_scores[hypo_idx] = min(score, self._beam_hyps_worst_scores[hypo_idx])
self._beam_hyps_count[hypo_idx] = hyps_count + 1
def hypo_is_done(self, hypo_idx: int, best_sum_logprobs: float, cur_len: int) -> bool:
"""
If there are enough hypotheses and that none of the hypotheses being generated can become better than the worst
one in the heap, then we are done with this sentence.
"""
if self.hypo_len(hypo_idx) < self.num_beams:
return False
elif self.do_early_stopping:
return True
else:
cur_score = best_sum_logprobs / cur_len**self.length_penalty
ret = self._beam_hyps_worst_scores[hypo_idx].item() >= cur_score
return ret
def process(
self,
input_ids: torch.Tensor,
next_scores: torch.Tensor,
next_tokens: torch.Tensor,
next_indices: torch.Tensor,
pad_token_id: Optional[int] = None,
eos_token_id: Optional[int] = None,
) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
cur_len = input_ids.shape[-1]
batch_size = len(self._beam_hyps_count)
assert batch_size == (input_ids.shape[0] // self.group_size)
device = input_ids.device
next_beam_scores = torch.zeros((batch_size, self.group_size), dtype=next_scores.dtype, device=device)
next_beam_tokens = torch.zeros((batch_size, self.group_size), dtype=next_tokens.dtype, device=device)
next_beam_indices = torch.zeros((batch_size, self.group_size), dtype=next_indices.dtype, device=device)
for batch_idx in range(batch_size):
if self._done[batch_idx]:
assert (
self.hypo_len(batch_idx) >= self.num_beams
), "Batch can only be done if at least {} beams have been generated".format(self.num_beams)
assert (
eos_token_id is not None and pad_token_id is not None
), "generated beams >= num_beams -> eos_token_id and pad_token have to be defined"
# pad the batch
next_beam_scores[batch_idx, :] = 0
next_beam_tokens[batch_idx, :] = pad_token_id
next_beam_indices[batch_idx, :] = 0
continue
# next tokens for this sentence
beam_idx = 0
for beam_token_rank, (next_token, next_score, next_index) in enumerate(
zip(next_tokens[batch_idx], next_scores[batch_idx], next_indices[batch_idx])
):
batch_beam_idx = batch_idx * self.group_size + next_index
# add to generated hypotheses if end of sentence
if (eos_token_id is not None) and (next_token == eos_token_id):
# if beam_token does not belong to top num_beams tokens, it should not be added
is_beam_token_worse_than_top_num_beams = beam_token_rank >= self.group_size
if is_beam_token_worse_than_top_num_beams:
continue
self.hypo_add(
input_ids[batch_beam_idx].clone(),
next_score.item(),
batch_idx,
)
else:
# add next predicted token since it is not eos_token
next_beam_scores[batch_idx, beam_idx] = next_score
next_beam_tokens[batch_idx, beam_idx] = next_token
next_beam_indices[batch_idx, beam_idx] = batch_beam_idx
beam_idx += 1
# once the beam for next step is full, don't add more tokens to it.
if beam_idx == self.group_size:
break
if beam_idx < self.group_size:
raise ValueError(
f"At most {self.group_size} tokens in {next_tokens[batch_idx]} can be equal to `eos_token_id:"
f" {eos_token_id}`. Make sure {next_tokens[batch_idx]} are corrected."
)
# Check if we are done so that we can save a pad step if all(done)
self._done[batch_idx] = self._done[batch_idx] or self.hypo_is_done(
batch_idx,
next_scores[batch_idx].max().item(),
cur_len,
)
return next_beam_scores.view(-1), next_beam_tokens.view(-1), next_beam_indices.view(-1)
def finalize(
self,
input_ids: torch.Tensor,
final_beam_scores: torch.Tensor,
final_beam_tokens: torch.Tensor,
final_beam_indices: torch.Tensor,
pad_token_id: int,
eos_token_id: int,
) -> Tuple[torch.Tensor, torch.Tensor]:
batch_size = len(self._beam_hyps_count)
# finalize all open beam hypotheses and add to generated hypotheses
for batch_idx in range(batch_size):
if self._done[batch_idx]:
continue
# all open beam hypotheses are added to the beam hypothesis
# beam hypothesis class automatically keeps the best beams
for beam_id in range(self.num_beams):
batch_beam_idx = batch_idx * self.num_beams + beam_id
final_score = final_beam_scores[batch_beam_idx].item()
final_tokens = input_ids[batch_beam_idx]
self.hypo_add(final_tokens, final_score, batch_idx)
# select the best hypotheses
# NOTE: torch.Tensor.new_zeros() is not scriptable
sent_lengths = torch.zeros(batch_size * self.num_beam_hyps_to_keep, dtype=torch.long)
best = []
best_scores = torch.zeros(
batch_size * self.num_beam_hyps_to_keep, device=input_ids.device, dtype=torch.float32
)
# retrieve best hypotheses
for i in range(batch_size):
# NOTE: lambda is not scriptable
batch_hypo_start = torch.sum(self._beam_hyps_count[:i]) if i > 0 else torch.tensor(0, dtype=torch.long)
batch_hypo_end = torch.sum(self._beam_hyps_count[: i + 1])
beam_scores = torch.cat(self._beam_scores)[batch_hypo_start:batch_hypo_end]
sorted_next_scores, sorted_indices = torch.topk(beam_scores, len(beam_scores), largest=True)
for j in range(self.num_beam_hyps_to_keep):
best_score = beam_scores[sorted_indices[j]]
best_hyp = self._beam_hyps[batch_hypo_start + sorted_indices[j]]
sent_lengths[self.num_beam_hyps_to_keep * i + j] = len(best_hyp)
# append to lists
best.append(best_hyp)
best_scores[i * self.num_beam_hyps_to_keep + j] = best_score
# prepare for adding eos
sent_max_len = min(sent_lengths.max() + 1, self.max_length)
decoded = torch.zeros(batch_size * self.num_beam_hyps_to_keep, sent_max_len, dtype=torch.long)
# shorter batches are padded if needed
if sent_lengths.min() != sent_lengths.max():
assert pad_token_id is not None, "`pad_token_id` has to be defined"
decoded.fill_(pad_token_id)
# fill with hypotheses and eos_token_id if the latter fits in
for i, hypo in enumerate(best):
decoded[i, : sent_lengths[i]] = hypo
if sent_lengths[i] < self.max_length:
decoded[i, sent_lengths[i]] = eos_token_id
return decoded, best_scores
class BARTBeamSearchGenerator(BARTGenerator):
def __init__(self, model):
super().__init__(model)
self.beam_scorer = BeamSearchScorerTS()
self.device = model.device
@staticmethod
def _expand_inputs_for_generation(
input_ids: torch.Tensor,
attention_mask: torch.Tensor,
last_hidden_state: torch.Tensor,
expand_size: int = 1,
) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
expanded_return_idx = (
torch.arange(input_ids.shape[0]).view(-1, 1).repeat(1, expand_size).view(-1).to(input_ids.device)
)
input_ids = input_ids.index_select(0, expanded_return_idx)
attention_mask = attention_mask.index_select(0, expanded_return_idx)
last_hidden_state = last_hidden_state.index_select(0, expanded_return_idx.to(last_hidden_state.device))
return input_ids, attention_mask, last_hidden_state
def adjust_logits_during_generation(self, logits, cur_len: int, max_length: int):
if cur_len == 1 and self.config.force_bos_token_to_be_generated:
logits = self._force_token_id_to_be_generated(logits, self.config.bos_token_id)
elif cur_len == max_length - 1 and self.config.eos_token_id is not None:
logits = self._force_token_id_to_be_generated(logits, self.config.eos_token_id)
return logits
@staticmethod
def _force_token_id_to_be_generated(scores, token_id: int):
"""force one of token_ids to be generated by setting prob of all other tokens to 0 (logprob=-float("inf"))"""
mask = torch.full_like(scores, 1, dtype=torch.bool)
mask[:, token_id] = False
return scores.masked_fill(mask, -float("inf"))
def _reorder_cache(self, past: List[torch.Tensor], beam_idx):
# if decoder past is not included in output
# speedy decoding is disabled and no need to reorder
reordered_decoder_past = []
for state in past:
reordered_decoder_past.append(state.index_select(0, beam_idx))
return reordered_decoder_past
def beam_search(
self, input_ids, encoder_output, attention_mask, num_beams, max_length, pad_token_id: int, eos_token_id: int
):
batch_size = self.beam_scorer.batch_size
num_beams = self.beam_scorer.num_beams
batch_beam_size, cur_len = input_ids.shape
assert (
num_beams * batch_size == batch_beam_size
), f"Batch dimension of `input_ids` should be {num_beams * batch_size}, but is {batch_beam_size}."
beam_scores = torch.zeros((batch_size, num_beams), dtype=torch.float, device=input_ids.device)
beam_scores[:, 1:] = -1e9
beam_scores = beam_scores.view((batch_size * num_beams,))
next_tokens = torch.zeros((batch_size, num_beams), dtype=torch.long, device=input_ids.device)
next_indices = torch.zeros((batch_size, num_beams), dtype=torch.long, device=input_ids.device)
past: List[torch.Tensor] = []
while cur_len < max_length:
logits, past = self._decoder_forward(input_ids, encoder_output, attention_mask, past)
next_token_logits = logits[:, -1, :]
# adjust tokens for Bart, *e.g.*
next_token_logits = self.adjust_logits_during_generation(
next_token_logits, cur_len=cur_len, max_length=max_length
)
next_token_scores = F.log_softmax(next_token_logits, dim=-1) # (batch_size * num_beams, vocab_size)
# pre-process distribution
next_token_scores = self.logits_processor(input_ids, next_token_scores)
next_token_scores = next_token_scores + beam_scores[:, None].expand_as(next_token_scores)
# reshape for beam search
vocab_size = next_token_scores.shape[-1]
next_token_scores = next_token_scores.view(batch_size, num_beams * vocab_size)
next_token_scores, next_tokens = torch.topk(
next_token_scores, 2 * num_beams, dim=1, largest=True, sorted=True
)
next_indices = next_tokens // vocab_size
next_tokens = next_tokens % vocab_size
beam_scores, beam_next_tokens, beam_idx = self.beam_scorer.process(
input_ids,
next_token_scores,
next_tokens,
next_indices,
pad_token_id=pad_token_id,
eos_token_id=eos_token_id,
)
input_ids = torch.cat([input_ids[beam_idx, :], beam_next_tokens.unsqueeze(-1)], dim=-1)
cur_len = cur_len + 1
if len(past) > 0:
past = self._reorder_cache(past, beam_idx)
if self.beam_scorer.is_done():
break
sequences, sequence_scores = self.beam_scorer.finalize(
input_ids,
beam_scores,
next_tokens,
next_indices,
pad_token_id=pad_token_id,
eos_token_id=eos_token_id,
)
return sequences
def forward(self, input_ids, attention_mask, num_beams, max_length, decoder_start_token_id):
pad_token_id = self.config.pad_token_id
bos_token_id = self.config.bos_token_id
eos_token_id = self.config.eos_token_id
# special case if pad_token_id is not defined
if pad_token_id is None and eos_token_id is not None:
# logger.warning(f"Setting `pad_token_id` to `eos_token_id`:{eos_token_id} for open-end generation.")
pad_token_id = eos_token_id
encoder_output = self._encoder_forward(input_ids, attention_mask)
input_ids = self._prepare_decoder_input_ids_for_generation(
input_ids,
decoder_start_token_id=decoder_start_token_id,
bos_token_id=bos_token_id,
)
batch_size = input_ids.shape[0]
length_penalty = self.config.length_penalty
num_return_sequences = self.config.num_return_sequences
early_stopping = True
self.beam_scorer.init(
batch_size=batch_size,
max_length=max_length,
num_beams=num_beams,
device=self.device,
length_penalty=length_penalty,
do_early_stopping=early_stopping,
num_beam_hyps_to_keep=num_return_sequences,
)
input_ids, attention_mask, encoder_output = self._expand_inputs_for_generation(
input_ids,
attention_mask,
encoder_output,
expand_size=num_beams,
)
return self.beam_search(
input_ids=input_ids,
encoder_output=encoder_output,
attention_mask=attention_mask,
num_beams=num_beams,
max_length=max_length,
pad_token_id=pad_token_id,
eos_token_id=eos_token_id,
)
| 0
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization
|
hf_public_repos/transformers/examples/research_projects/onnx/summarization/bart_onnx/reduce_onnx_size.py
|
"""
Code to remove duplicate initializers to reduce ONNX model size.
"""
import os
import numpy
import onnx
def _is_equal_tensor_proto(a, b):
name_a = a.name
name_b = b.name
a.name = ""
b.name = ""
res = a == b
a.name = name_a
b.name = name_b
return res
def _node_replace_input_with(node_proto, name, new_name):
for i, input_name in enumerate(node_proto.input):
if input_name == name:
node_proto.input.insert(i, new_name)
node_proto.input.pop(i + 1)
if node_proto.op_type == "If":
_graph_replace_input_with(node_proto.attribute[0].g, name, new_name)
_graph_replace_input_with(node_proto.attribute[1].g, name, new_name)
if node_proto.op_type == "Loop":
_graph_replace_input_with(node_proto.attribute[0].g, name, new_name)
def _graph_replace_input_with(graph_proto, name, new_name):
for n in graph_proto.node:
_node_replace_input_with(n, name, new_name)
def _remove_dup_initializers_from_model(model, model_without_ext, ind_to_replace):
inits_with_data = list(model.graph.initializer)
inits = list(model_without_ext.graph.initializer)
for i, ref_i in ind_to_replace:
assert inits_with_data[i].name == inits[i].name
assert inits_with_data[ref_i].name == inits[ref_i].name
assert i > ref_i
name_i = inits[i].name
name_ref = inits[ref_i].name
model_without_ext.graph.initializer.remove(inits[i])
# for n in model.graph.node:
_graph_replace_input_with(model_without_ext.graph, name_i, name_ref)
def remove_dup_initializers(onnx_file_path):
"""
Removes duplicate initializers from the model to reduce its size.
Writes a new file in the same directory as onnx_file_path and returns the path to that file.
"""
model_file_folder = os.path.dirname(onnx_file_path)
model_file_name = os.path.basename(onnx_file_path)
model = onnx.load(os.path.join(model_file_folder, model_file_name))
inits = list(model.graph.initializer)
dup_set = set()
dup_map = {}
ind_to_replace = []
total_reduced_size = 0
for i in range(len(inits)):
if i in dup_set:
continue
for j in range(i + 1, len(inits)):
if j in dup_set:
continue
if _is_equal_tensor_proto(inits[i], inits[j]):
dup_set.add(i)
dup_set.add(j)
dtype = inits[j].data_type
mem_size = numpy.prod(inits[j].dims)
if dtype == 1:
mem_size *= 4
elif dtype == 6:
mem_size *= 4
elif dtype == 7 or dtype == 11:
mem_size *= 8
else:
print("unexpected data type: ", dtype)
total_reduced_size += mem_size
name_i = inits[i].name
name_j = inits[j].name
if name_i in dup_map:
dup_map[name_i].append(name_j)
else:
dup_map[name_i] = [name_j]
ind_to_replace.append((j, i))
print("total reduced size: ", total_reduced_size / 1024 / 1024 / 1024, "GB")
ind_to_replace = sorted(ind_to_replace)
_remove_dup_initializers_from_model(model, model, ind_to_replace)
optimized_model_file_name = "optimized_" + model_file_name
new_model = os.path.join(model_file_folder, optimized_model_file_name)
onnx.save(model, new_model)
return new_model
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/adversarial/README.md
|
## Adversarial evaluation of model performances
Here is an example on evaluating a model using adversarial evaluation of natural language inference with the Heuristic Analysis for NLI Systems (HANS) dataset [McCoy et al., 2019](https://arxiv.org/abs/1902.01007). The example was gracefully provided by [Nafise Sadat Moosavi](https://github.com/ns-moosavi).
The HANS dataset can be downloaded from [this location](https://github.com/tommccoy1/hans).
This is an example of using test_hans.py:
```bash
export HANS_DIR=path-to-hans
export MODEL_TYPE=type-of-the-model-e.g.-bert-roberta-xlnet-etc
export MODEL_PATH=path-to-the-model-directory-that-is-trained-on-NLI-e.g.-by-using-run_glue.py
python run_hans.py \
--task_name hans \
--model_type $MODEL_TYPE \
--do_eval \
--data_dir $HANS_DIR \
--model_name_or_path $MODEL_PATH \
--max_seq_length 128 \
--output_dir $MODEL_PATH \
```
This will create the hans_predictions.txt file in MODEL_PATH, which can then be evaluated using hans/evaluate_heur_output.py from the HANS dataset.
The results of the BERT-base model that is trained on MNLI using batch size 8 and the random seed 42 on the HANS dataset is as follows:
```bash
Heuristic entailed results:
lexical_overlap: 0.9702
subsequence: 0.9942
constituent: 0.9962
Heuristic non-entailed results:
lexical_overlap: 0.199
subsequence: 0.0396
constituent: 0.118
```
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/adversarial/run_hans.py
|
# 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 sequence classification on HANS."""
import logging
import os
from dataclasses import dataclass, field
from typing import Dict, List, Optional
import numpy as np
import torch
from utils_hans import HansDataset, InputFeatures, hans_processors, hans_tasks_num_labels
import transformers
from transformers import (
AutoConfig,
AutoModelForSequenceClassification,
AutoTokenizer,
HfArgumentParser,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import is_main_process
logger = logging.getLogger(__name__)
@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"},
)
@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 selected in the list: " + ", ".join(hans_processors.keys())}
)
data_dir: str = field(
metadata={"help": "The input data dir. Should contain the .tsv files (or other 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 hans_data_collator(features: List[InputFeatures]) -> Dict[str, torch.Tensor]:
"""
Data collator that removes the "pairID" key if present.
"""
batch = default_data_collator(features)
_ = batch.pop("pairID", None)
return batch
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))
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 training_args.local_rank in [-1, 0] else logging.WARN,
)
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
training_args.local_rank,
training_args.device,
training_args.n_gpu,
bool(training_args.local_rank != -1),
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
set_seed(training_args.seed)
try:
num_labels = hans_tasks_num_labels[data_args.task_name]
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,
)
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,
)
# Get datasets
train_dataset = (
HansDataset(
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,
)
if training_args.do_train
else None
)
eval_dataset = (
HansDataset(
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,
evaluate=True,
)
if training_args.do_eval
else None
)
# Initialize our Trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
data_collator=hans_data_collator,
)
# Training
if training_args.do_train:
trainer.train(
model_path=model_args.model_name_or_path if os.path.isdir(model_args.model_name_or_path) else None
)
trainer.save_model()
# For convenience, we also re-save the tokenizer to the same directory,
# so that you can share your model easily on huggingface.co/models =)
if trainer.is_world_master():
tokenizer.save_pretrained(training_args.output_dir)
# Evaluation
if training_args.do_eval:
logger.info("*** Evaluate ***")
output = trainer.predict(eval_dataset)
preds = output.predictions
preds = np.argmax(preds, axis=1)
pair_ids = [ex.pairID for ex in eval_dataset]
output_eval_file = os.path.join(training_args.output_dir, "hans_predictions.txt")
label_list = eval_dataset.get_labels()
if trainer.is_world_master():
with open(output_eval_file, "w") as writer:
writer.write("pairID,gold_label\n")
for pid, pred in zip(pair_ids, preds):
writer.write("ex" + str(pid) + "," + label_list[int(pred)] + "\n")
trainer._log(output.metrics)
def _mp_fn(index):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/adversarial/utils_hans.py
|
# 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.
import logging
import os
from dataclasses import dataclass
from typing import List, Optional, Union
import tqdm
from filelock import FileLock
from transformers import (
BartTokenizer,
BartTokenizerFast,
DataProcessor,
PreTrainedTokenizer,
RobertaTokenizer,
RobertaTokenizerFast,
XLMRobertaTokenizer,
is_tf_available,
is_torch_available,
)
logger = logging.getLogger(__name__)
@dataclass(frozen=True)
class InputExample:
"""
A single training/test example for simple sequence classification.
Args:
guid: Unique id for the example.
text_a: string. The untokenized text of the first sequence. For single
sequence tasks, only this sequence must be specified.
text_b: (Optional) string. The untokenized text of the second sequence.
Only must be specified for sequence pair tasks.
label: (Optional) string. The label of the example. This should be
specified for train and dev examples, but not for test examples.
pairID: (Optional) string. Unique identifier for the pair of sentences.
"""
guid: str
text_a: str
text_b: Optional[str] = None
label: Optional[str] = None
pairID: Optional[str] = None
@dataclass(frozen=True)
class InputFeatures:
"""
A single set of features of data.
Property names are the same names as the corresponding inputs to a model.
Args:
input_ids: Indices of input sequence tokens in the vocabulary.
attention_mask: Mask to avoid performing attention on padding token indices.
Mask values selected in ``[0, 1]``:
Usually ``1`` for tokens that are NOT MASKED, ``0`` for MASKED (padded) tokens.
token_type_ids: (Optional) Segment token indices to indicate first and second
portions of the inputs. Only some models use them.
label: (Optional) Label corresponding to the input. Int for classification problems,
float for regression problems.
pairID: (Optional) Unique identifier for the pair of sentences.
"""
input_ids: List[int]
attention_mask: Optional[List[int]] = None
token_type_ids: Optional[List[int]] = None
label: Optional[Union[int, float]] = None
pairID: Optional[int] = None
if is_torch_available():
import torch
from torch.utils.data import Dataset
class HansDataset(Dataset):
"""
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] = None,
overwrite_cache=False,
evaluate: bool = False,
):
processor = hans_processors[task]()
cached_features_file = os.path.join(
data_dir,
"cached_{}_{}_{}_{}".format(
"dev" if evaluate else "train",
tokenizer.__class__.__name__,
str(max_seq_length),
task,
),
)
label_list = processor.get_labels()
if tokenizer.__class__ in (
RobertaTokenizer,
RobertaTokenizerFast,
XLMRobertaTokenizer,
BartTokenizer,
BartTokenizerFast,
):
# HACK(label indices are swapped in RoBERTa pretrained model)
label_list[1], label_list[2] = label_list[2], label_list[1]
self.label_list = label_list
# Make sure only the first process in distributed training processes the dataset,
# and the others will use the cache.
lock_path = cached_features_file + ".lock"
with FileLock(lock_path):
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}")
examples = (
processor.get_dev_examples(data_dir) if evaluate else processor.get_train_examples(data_dir)
)
logger.info("Training examples: %s", len(examples))
self.features = hans_convert_examples_to_features(examples, label_list, max_seq_length, tokenizer)
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 __getitem__(self, i) -> InputFeatures:
return self.features[i]
def get_labels(self):
return self.label_list
if is_tf_available():
import tensorflow as tf
class TFHansDataset:
"""
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,
evaluate: bool = False,
):
processor = hans_processors[task]()
label_list = processor.get_labels()
if tokenizer.__class__ in (
RobertaTokenizer,
RobertaTokenizerFast,
XLMRobertaTokenizer,
BartTokenizer,
BartTokenizerFast,
):
# HACK(label indices are swapped in RoBERTa pretrained model)
label_list[1], label_list[2] = label_list[2], label_list[1]
self.label_list = label_list
examples = processor.get_dev_examples(data_dir) if evaluate else processor.get_train_examples(data_dir)
self.features = hans_convert_examples_to_features(examples, label_list, max_seq_length, tokenizer)
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]
def get_labels(self):
return self.label_list
class HansProcessor(DataProcessor):
"""Processor for the HANS data set."""
def get_train_examples(self, data_dir):
"""See base class."""
return self._create_examples(self._read_tsv(os.path.join(data_dir, "heuristics_train_set.txt")), "train")
def get_dev_examples(self, data_dir):
"""See base class."""
return self._create_examples(self._read_tsv(os.path.join(data_dir, "heuristics_evaluation_set.txt")), "dev")
def get_labels(self):
"""See base class.
Note that we follow the standard three labels for MNLI
(see :class:`~transformers.data.processors.utils.MnliProcessor`)
but the HANS evaluation groups `contradiction` and `neutral` into `non-entailment` (label 0) while
`entailment` is label 1."""
return ["contradiction", "entailment", "neutral"]
def _create_examples(self, lines, set_type):
"""Creates examples for the training and dev sets."""
examples = []
for i, line in enumerate(lines):
if i == 0:
continue
guid = "%s-%s" % (set_type, line[0])
text_a = line[5]
text_b = line[6]
pairID = line[7][2:] if line[7].startswith("ex") else line[7]
label = line[0]
examples.append(InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label, pairID=pairID))
return examples
def hans_convert_examples_to_features(
examples: List[InputExample],
label_list: List[str],
max_length: int,
tokenizer: PreTrainedTokenizer,
):
"""
Loads a data file into a list of ``InputFeatures``
Args:
examples: List of ``InputExamples`` containing the examples.
label_list: List of labels. Can be obtained from the processor using the ``processor.get_labels()`` method.
max_length: Maximum example length.
tokenizer: Instance of a tokenizer that will tokenize the examples.
Returns:
A list of task-specific ``InputFeatures`` which can be fed to the model.
"""
label_map = {label: i for i, label in enumerate(label_list)}
features = []
for ex_index, example in tqdm.tqdm(enumerate(examples), desc="convert examples to features"):
if ex_index % 10000 == 0:
logger.info("Writing example %d" % (ex_index))
inputs = tokenizer(
example.text_a,
example.text_b,
add_special_tokens=True,
max_length=max_length,
padding="max_length",
truncation=True,
return_overflowing_tokens=True,
)
label = label_map[example.label] if example.label in label_map else 0
pairID = int(example.pairID)
features.append(InputFeatures(**inputs, label=label, pairID=pairID))
for i, example in enumerate(examples[:5]):
logger.info("*** Example ***")
logger.info(f"guid: {example}")
logger.info(f"features: {features[i]}")
return features
hans_tasks_num_labels = {
"hans": 3,
}
hans_processors = {
"hans": HansProcessor,
}
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/adversarial/requirements.txt
|
transformers == 3.5.1
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/README.md
|
# Flax/JAX community week 🤗
Welcome to the Flax/JAX community week! The goal of this week is to make compute-intensive NLP and CV projects (like pre-training BERT, GPT2, CLIP, ViT)
practicable for a wider audience of engineers and researchers.
To do so, we will try to teach **you** how to effectively use JAX/Flax on TPU and help you to complete a fun NLP and/or CV project in JAX/Flax during the community week.
Free access to a TPUv3-8 will kindly be provided by the Google Cloud team!
In this document, we list all the important information that you will need during the Flax/JAX community week.
Don't forget to sign up [here](https://forms.gle/tVGPhjKXyEsSgUcs8)!
## Table of Contents
- [Organization](#organization)
- [Important dates](#important-dates)
- [Communication](#communication)
- [Projects](#projects)
- [How to propose](#how-to-propose-a-project)
- [How to form a team](#how-to-form-a-team-around-a-project)
- [Tips & Tricks for project](#tips-on-how-to-organize-the-project)
- [How to install flax, jax, optax, transformers, datasets](#how-to-install-relevant-libraries)
- [Quickstart Flax/JAX](#quickstart-flax-and-jax)
- [Quickstart Flax/JAX in 🤗 Transformers](#quickstart-flax-and-jax-in-transformers)
- [Flax design philosophy in 🤗 Transformers](#flax-design-philosophy-in-transformers)
- [How to use flax models & scripts](#how-to-use-flax-models-and-example-scripts)
- [Talks](#talks)
- [How to use the 🤗 Hub for training](#how-to-use-the-hub-for-collaboration)
- [How to setup TPU VM](#how-to-setup-tpu-vm)
- [How to build a demo](#how-to-build-a-demo)
- [Using the Hugging Face Widgets](#using-the-hugging-face-widgets)
- [Using a Streamlit demo](#using-a-streamlit-demo)
- [Using a Gradio demo](#using-a-gradio-demo)
- [Project evaluation](#project-evaluation)
- [General Tips & Tricks](#general-tips-and-tricks)
- [FAQ](#faq)
## Organization
Participants can propose ideas for an interesting NLP and/or CV project. Teams of 3 to 5 will then be formed around the most promising and interesting projects. Make sure to read through the [Projects](#projects) section on how to propose projects, comment on other participants' project ideas, and create a team.
To help each team successfully finish their project, we have organized talks by leading scientists and engineers from Google, Hugging Face, and the open-source NLP & CV community. The talks will take place before the community week from June 30th to July 2nd. Make sure to attend the talks to get the most out of your participation! Check out the [Talks](#talks) section to get an overview of the talks, including the speaker and the time of the talk.
Each team is then given **free access to a TPUv3-8 VM** from July 7th to July 14th. In addition, we will provide training examples in JAX/Flax for a variety of NLP and Vision models to kick-start your project. During the week, we'll make sure to answer any questions you might have about JAX/Flax and Transformers and help each team as much as possible to complete their project!
At the end of the community week, each team should submit a demo of their project. All demonstrations will be evaluated by a jury and the top-3 demos will be awarded a prize. Check out the [How to submit a demo](#how-to-submit-a-demo) section for more information and suggestions on how to submit your project.
## Important dates
- **23.06.** Official announcement of the community week. Make sure to sign-up in [this google form](https://forms.gle/tVGPhjKXyEsSgUcs8).
- **23.06. - 30.06.** Participants will be added to an internal Slack channel. Project ideas can be proposed here and groups of 3-5 are formed. Read this document for more information.
- **30.06.** Release of all relevant training scripts in JAX/Flax as well as other documents on how to set up a TPU, how to use the training scripts, how to submit a demo, tips & tricks for JAX/Flax, tips & tricks for efficient use of the hub.
- **30.06. - 2.07.** Talks about JAX/Flax, TPU, Transformers, Computer Vision & NLP will be held.
- **7.07.** Start of the community week! Access to TPUv3-8 will be given to each team.
- **7.07. - 14.07.** The Hugging Face & JAX/Flax & Cloud team will be available for any questions, problems the teams might run into.
- **15.07.** Access to TPU is deactivated and community week officially ends.
- **16.07.** Deadline for each team to submit a demo.
## Communication
All important communication will take place in an internal Slack channel, called `#flax-jax-community-week`.
Important announcements of the Hugging Face, Flax/JAX, and Google Cloud team will be posted there.
Such announcements include general information about the community week (Dates, Rules, ...), release of relevant training scripts (Flax/JAX example scripts for NLP and Vision), release of other important documents (How to access the TPU), etc.
The Slack channel will also be the central place for participants to post about their results, share their learning experiences, ask questions, etc.
For issues with Flax/JAX, Transformers, Datasets or for questions that are specific to your project we would be **very happy** if you could use the following public repositories and forums:
- Flax: [Issues](https://github.com/google/flax/issues), [Questions](https://github.com/google/flax/discussions)
- JAX: [Issues](https://github.com/google/jax/issues), [Questions](https://github.com/google/jax/discussions)
- 🤗 Transformers: [Issues](https://github.com/huggingface/transformers/issues), [Questions](https://discuss.huggingface.co/c/transformers/9)
- 🤗 Datasets: [Issues](https://github.com/huggingface/datasets/issues), [Questions](https://discuss.huggingface.co/c/datasets/10)
- Project specific questions: [Forum](https://discuss.huggingface.co/c/flax-jax-projects/22)
- TPU related questions: [TODO]()
Please do **not** post the complete issue/project-specific question in the Slack channel, but instead a link to your issue/question that we will try to answer as soon as possible.
This way, we make sure that the everybody in the community can benefit from your questions - even after the community week - and that the same question is not answered twice.
To be invited to the Slack channel, please make sure you have signed up [on the Google form](https://forms.gle/tVGPhjKXyEsSgUcs8).
**Note**: If you have signed up on the google form, but you are not in the Slack channel, please leave a message on [(TODO) the official forum announcement]( ) and ping `@Suzana` and `@patrickvonplaten`.
## Projects
During the first week after the community week announcement, **23.06. - 30.06.**, teams will be formed around the most promising and interesting project ideas. Each team can consist of 2 to 10 participants. Projects can be accessed [here](https://discuss.huggingface.co/c/flax-jax-projects/22).
All officially defined projects can be seen [here](https://docs.google.com/spreadsheets/d/1GpHebL7qrwJOc9olTpIPgjf8vOS0jNb6zR_B8x_Jtik/edit?usp=sharing).
### How to propose a project
Some default project ideas are given by the organizers. **However, we strongly encourage participants to submit their own project ideas!**
Check out the [HOW_TO_PROPOSE_PROJECT.md](https://github.com/huggingface/transformers/tree/main/examples/research_projects/jax-projects/HOW_TO_PROPOSE_PROJECT.md) for more information on how to propose a new project.
### How to form a team around a project
You can check out all existing projects ideas on the forum under [Flax/JAX projects category](https://discuss.huggingface.co/c/flax-jax-projects/22).
Make sure to quickly check out each project idea and leave a ❤️ if you like an idea.
Feel free to leave comments, suggestions for improvement, or questions about more details directly on the discussion thread.
If you have found the project that you ❤️ the most, leave a message "I would like to join this project" on the discussion thread.
We strongly advise you to also shortly state who you are, which time zone you are in and why you would like to work on this project, how you can contribute to the project and what your vision is for the project.
For projects that see a lot of interest and for which enough participants have expressed interest in joining, an official team will be created by the organizers.
One of the organizers (`@Suzana`, `@valhalla`, `@osanseviero`, `@patrickvonplaten`) will leave a message "For this project the team: `<team_name>`, `<team_members>` , is officially created" on the thread and note down the teams on [this google sheet](https://docs.google.com/spreadsheets/d/1GpHebL7qrwJOc9olTpIPgjf8vOS0jNb6zR_B8x_Jtik/edit?usp=sharing).
Once created, the team can start refining their project:
- What is the goal of the project? *E.g.*, Present a language model that writes poetry in Russian.
- What model will we use? *E.g.*, FlaxGPT2
- What data will we use? *E.g.* Russian dataset of OSCAR & publicly available book on poetry
- Should we use a pre-trained model or train a model from scratch? E.g. Train a model from scratch
- What training scripts do we need? *E.g.* `transformers/examples/flax/run_clm_flax.py` can be used
- What kind of demo would we like to present? E.g. Text-generation API of the 🤗 Hub in combination with a Streamlit demo that lets the user generate a poem of a given length
- How will the work be divided? *E.g.* Team member 1 works on data preprocessing, Team member 2 works on adapting the Flax script, ...
We highly recommend that each team discusses all relevant ideas for their project directly on the forum thread.
This way valuable learning experiences are shared and accessible by the whole community in the future.
Additionally, the organizers, other participants, or anybody in the community really can read through your discussions and leave comments/tips for improvement. Obviously, you can also create private chats, ... to discuss more sensitive topics, etc.
**Important**:
- For project ideas that see a lot of interest, we are more than happy to create more than one team.
- Participants are welcome to join multiple teams, even though we encourage them to only work on a single project.
- Under special circumstances, participants can change/create new teams. Please note that we would like to keep this the exception. If however, you would like to change/leave existing teams, please leave a post on the project's thread where you ping the corresponding organizer that created the group.
- It is often easy to propose/join a project that is done in your native language. Feel free to reach out to existing [language-specific groups](https://discuss.huggingface.co/c/languages-at-hugging-face/15) to look for community members that might be interested in joining your project.
## Tips on how to organize the project
This section gives you some tips on how to most efficiently & effectively
work as a team to achieve your goal. It is by no means a strict recipe to follow,
but rather a collection of tips from the 🤗 team.
Once your team is defined, you can start working on the project as soon as possible.
### Communication
At first, it is always useful to get to know each other and to set up a means of communication.
While we recommend that all technical aspects of work can be discussed directly on the [forum](https://discuss.huggingface.co/c/flax-jax-projects/22) under your project thread,
it can be very helpful to have a more direct way of communicating, *e.g.* in a channel.
For this we have created a discord that you can access [here](https://discord.com/channels/858019234139602994/858019234139602997).
This discord will not be managed by anybody and is just there so that you can communicate more effectively with your team members.
Feel free to create a new channel for you and your team where you can discuss everything. If you and your team have already set up other ways of communicating, it is absolutely not required to make use of the discord. However, we do recommend each team to set up some kind
of channel or group for quick discussions.
### Project definition
In the very beginning, you should make sure your project is well-defined and that
everybody in the team understands the goal of the project and the work that needs to be
done in order to achieve the goal. A well-defined project:
- has defined the task on which the model will be trained
- has defined the model that will be trained
- has defined the datasets that will be used for training
- has defined the type of training scripts that need to be written
- has defined the desired outcome of the project
- has defined the workflows
By "has defined" we don't meant that the corresponding code already has to be written and ready
to be used, but that everybody in team is on the same page on what type of model, data and training script should be used.
To give an example, a well-defined project would be the following:
- task: summarization
- model: [t5-small](https://huggingface.co/t5-small)
- dataset: [CNN/Daily mail](https://huggingface.co/datasets/cnn_dailymail)
- training script: [run_summarization_flax.py](https://github.com/huggingface/transformers/blob/main/examples/flax/summarization/run_summarization_flax.py)
- outcome: t5 model that can summarize news
- work flow: adapt `run_summarization_flax.py` to work with `t5-small`.
This example is a very easy and not the most interesting project since a `t5-small`
summarization model exists already for CNN/Daily mail and pretty much no code has to be
written.
A well-defined project does not need to have the dataset be part of
the `datasets` library and the training script already be pre-written, however it should
be clear how the desired dataset can be accessed and how the training script can be
written.
It is also important to have a clear plan regarding the workflow. Usually, the
data processing is done in a first step. Once the data is in a format that the model can
work with, the training script can be written, etc. These steps should be more detailed
once the team has a clearly defined project. It can be helpful to set deadlines for each step.
### Workload division
To effectively work as a team, it is crucial to divide the workload among everybody.
Some team members will be more motivated and experienced than others and
some team members simply want to participate to learn more and cannot contribute that
much to the team. This is totally fine! One cannot expect everybody in the team to have the same level of experience and time/motivation during the community week.
As a conclusion, being honest about one's expected involvement is crucial so that
the workload can be divided accordingly. If someone doesn't think her/his tasks are feasible - let
the team know early on so that someone else can take care of it!
It is recommended that the motivated and experienced team members take the lead in dividing the work and are ready to take over the tasks of another team member if necessary.
The workload can often be divided according to:
- data preprocessing (load the data and preprocess data in the correct format)
- data tokenization / data collator (process data samples into tokens or images)
- model configuration (writing the code that defines the model)
- model forward pass (make sure input / output work correctly)
- loss function (define the loss function)
- putting the pieces together in a training script
Many of the steps above require other steps to be finished, so it often makes sense
to use dummy data in the expected format to start, *e.g.*, with the model forward pass
before the data preprocessing is done.
### Expectations
It is also very important to stay realistic with the scope of your project. Each team
has access to a TPUv3-8 for only *ca.* 10 days, so it's important to keep the scope of
the project reasonable. While we do want each team to work on interesting projects, each
team should make sure that the project goals can be achieved within the provided compute
time on TPU. For instance, pretraining a 11 billion parameters T5 model is not really a realistic
task with just 10 days of TPUv3-8 compute.
Also, it might be difficult to finish a project where the whole modeling, dataset and training code has to be written from scratch.
Having defined your project, feel free to reach out on Slack or the forum for feedback from the organizers. We can surely give you our opinion on whether the project is feasible and what can be done to improve it.
the project is feasible.
### Other tips
Here is a collection of some more tips:
- We strongly recommend to work as publicly and collaboratively as possible during the week so that other teams
and the organizers can best help you. This includes publishing important discussions on
the forum and making use of the [🤗 hub](http://huggingface.co/) to have a version
control for your models and training logs.
- When debugging, it is important that the debugging cycle is kept as short as possible to
be able to effectively debug. *E.g.* if there is a problem with your training script,
you should run it with just a couple of hundreds of examples and not the whole dataset script. This can be done by either making use of [datasets streaming](https://huggingface.co/docs/datasets/master/dataset_streaming?highlight=streaming) or by selecting just the first
X number of data samples after loading:
```python
datasets["train"] = datasets["train"].select(range(1000))
```
- Ask for help. If you are stuck, use the public Slack channel or the [forum](https://discuss.huggingface.co/c/flax-jax-projects/22) to ask for help.
## How to install relevant libraries
In the following we will explain how to install all relevant libraries on your local computer and on TPU VM.
It is recommended to install all relevant libraries both on your local machine
and on the TPU virtual machine. This way, quick prototyping and testing can be done on
your local machine and the actual training can be done on the TPU VM.
### Local computer
The following libraries are required to train a JAX/Flax model with 🤗 Transformers and 🤗 Datasets:
- [JAX](https://github.com/google/jax/)
- [Flax](https://github.com/google/flax)
- [Optax](https://github.com/deepmind/optax)
- [Transformers](https://github.com/huggingface/transformers)
- [Datasets](https://github.com/huggingface/datasets)
You should install the above libraries in a [virtual environment](https://docs.python.org/3/library/venv.html).
If you're unfamiliar with Python virtual environments, check out the [user guide](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/). Create a virtual environment with the version of Python you're going
to use and activate it.
You should be able to run the command:
```bash
python3 -m venv <your-venv-name>
```
You can activate your venv by running
```bash
source ~/<your-venv-name>/bin/activate
```
We strongly recommend to make use of the provided JAX/Flax examples scripts in [transformers/examples/flax](https://github.com/huggingface/transformers/tree/main/examples/flax) even if you want to train a JAX/Flax model of another github repository that is not integrated into 🤗 Transformers.
In all likelihood, you will need to adapt one of the example scripts, so we recommend forking and cloning the 🤗 Transformers repository as follows.
Doing so will allow you to share your fork of the Transformers library with your team members so that the team effectively works on the same code base. It will also automatically install the newest versions of `flax`, `jax` and `optax`.
1. Fork the [repository](https://github.com/huggingface/transformers) by
clicking on the 'Fork' button on the repository's page. This creates a copy of the code
under your GitHub user account.
2. Clone your fork to your local disk, and add the base repository as a remote:
```bash
$ git clone https://github.com/<your Github handle>/transformers.git
$ cd transformers
$ git remote add upstream https://github.com/huggingface/transformers.git
```
3. Create a new branch to hold your development changes. This is especially useful to share code changes with your team:
```bash
$ git checkout -b a-descriptive-name-for-my-project
```
4. Set up a flax environment by running the following command in a virtual environment:
```bash
$ pip install -e ".[flax]"
```
(If transformers was already installed in the virtual environment, remove
it with `pip uninstall transformers` before reinstalling it in editable
mode with the `-e` flag.)
If you have already cloned that repo, you might need to `git pull` to get the most recent changes in the `datasets`
library.
Running this command will automatically install `flax`, `jax` and `optax`.
Next, you should also install the 🤗 Datasets library. We strongly recommend installing the
library from source to profit from the most current additions during the community week.
Simply run the following steps:
```
$ cd ~/
$ git clone https://github.com/huggingface/datasets.git
$ cd datasets
$ pip install -e ".[streaming]"
```
If you plan on contributing a specific dataset during
the community week, please fork the datasets repository and follow the instructions
[here](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-create-a-pull-request).
To verify that all libraries are correctly installed, you can run the following command.
It assumes that both `transformers` and `datasets` were installed from main - otherwise
datasets streaming will not work correctly.
```python
from transformers import FlaxRobertaModel, RobertaTokenizerFast
from datasets import load_dataset
import jax
dataset = load_dataset('oscar', "unshuffled_deduplicated_en", split='train', streaming=True)
dummy_input = next(iter(dataset))["text"]
tokenizer = RobertaTokenizerFast.from_pretrained("roberta-base")
input_ids = tokenizer(dummy_input, return_tensors="np").input_ids[:, :10]
model = FlaxRobertaModel.from_pretrained("julien-c/dummy-unknown")
# run a forward pass, should return an object `FlaxBaseModelOutputWithPooling`
model(input_ids)
```
### TPU VM
**VERY IMPORTANT** - Only one process can access the TPU cores at a time. This means that if multiple team members
are trying to connect to the TPU cores errors, such as:
```
libtpu.so already in used by another process. Not attempting to load libtpu.so in this process.
```
are thrown. As a conclusion, we recommend every team member to create her/his own virtual environment, but only one
person should run the heavy training processes. Also, please take turns when setting up the TPUv3-8 so that everybody
can verify that JAX is correctly installed.
The following libraries are required to train a JAX/Flax model with 🤗 Transformers and 🤗 Datasets on TPU VM:
- [JAX](https://github.com/google/jax/)
- [Flax](https://github.com/google/flax)
- [Optax](https://github.com/deepmind/optax)
- [Transformers](https://github.com/huggingface/transformers)
- [Datasets](https://github.com/huggingface/datasets)
You should install the above libraries in a [virtual environment](https://docs.python.org/3/library/venv.html).
If you're unfamiliar with Python virtual environments, check out the [user guide](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/). Create a virtual environment with the version of Python you're going
to use and activate it.
You should be able to run the command:
```bash
python3 -m venv <your-venv-name>
```
If this doesn't work, you first might to have install `python3-venv`. You can do this as follows:
```bash
sudo apt-get install python3-venv
```
You can activate your venv by running
```bash
source ~/<your-venv-name>/bin/activate
```
Next you should install JAX's TPU version on TPU by running the following command:
```
$ pip install requests
```
and then:
```
$ pip install "jax[tpu]>=0.2.16" -f https://storage.googleapis.com/jax-releases/libtpu_releases.html
```
**Note**: Running this command might actually throw an error, such as:
```
Building wheel for jax (setup.py) ... error
ERROR: Command errored out with exit status 1:
command: /home/patrick/patrick/bin/python3 -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-install-lwseckn1/jax/setup.py'"'"'; __file__='"'"'/tmp/pip-install-lwseckn1/jax/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(__file__);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' bdist_wheel -d /tmp/pip-wheel-pydotzlo
cwd: /tmp/pip-install-lwseckn1/jax/
Complete output (6 lines):
usage: setup.py [global_opts] cmd1 [cmd1_opts] [cmd2 [cmd2_opts] ...]
or: setup.py --help [cmd1 cmd2 ...]
or: setup.py --help-commands
or: setup.py cmd --help
error: invalid command 'bdist_wheel'
----------------------------------------
ERROR: Failed building wheel for jax
```
Jax should have been installed correctly nevertheless.
To verify that JAX was correctly installed, you can run the following command:
```python
import jax
jax.device_count()
```
This should display the number of TPU cores, which should be 8 on a TPUv3-8 VM.
We strongly recommend to make use of the provided JAX/Flax examples scripts in [transformers/examples/flax](https://github.com/huggingface/transformers/tree/main/examples/flax) even if you want to train a JAX/Flax model of another github repository that is not integrated into 🤗 Transformers.
In all likelihood, you will need to adapt one of the example scripts, so we recommend forking and cloning the 🤗 Transformers repository as follows.
Doing so will allow you to share your fork of the Transformers library with your team members so that the team effectively works on the same code base. It will also automatically install the newest versions of `flax`, `jax` and `optax`.
1. Fork the [repository](https://github.com/huggingface/transformers) by
clicking on the 'Fork' button on the repository's page. This creates a copy of the code
under your GitHub user account.
2. Clone your fork to your local disk, and add the base repository as a remote:
```bash
$ git clone https://github.com/<your Github handle>/transformers.git
$ cd transformers
$ git remote add upstream https://github.com/huggingface/transformers.git
```
3. Create a new branch to hold your development changes. This is especially useful to share code changes with your team:
```bash
$ git checkout -b a-descriptive-name-for-my-project
```
4. Set up a flax environment by running the following command in a virtual environment:
```bash
$ pip install -e ".[flax]"
```
(If transformers was already installed in the virtual environment, remove
it with `pip uninstall transformers` before reinstalling it in editable
mode with the `-e` flag.)
If you have already cloned that repo, you might need to `git pull` to get the most recent changes in the `datasets`
library.
Running this command will automatically install `flax`, `jax` and `optax`.
Next, you should also install the 🤗 Datasets library. We strongly recommend installing the
library from source to profit from the most current additions during the community week.
Simply run the following steps:
```
$ cd ~/
$ git clone https://github.com/huggingface/datasets.git
$ cd datasets
$ pip install -e ".[streaming]"
```
If you plan on contributing a specific dataset during
the community week, please fork the datasets repository and follow the instructions
[here](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-create-a-pull-request).
To verify that all libraries are correctly installed, you can run the following command.
It assumes that both `transformers` and `datasets` were installed from main - otherwise
datasets streaming will not work correctly.
```python
from transformers import FlaxRobertaModel, RobertaTokenizerFast
from datasets import load_dataset
import jax
dataset = load_dataset('oscar', "unshuffled_deduplicated_en", split='train', streaming=True)
dummy_input = next(iter(dataset))["text"]
tokenizer = RobertaTokenizerFast.from_pretrained("roberta-base")
input_ids = tokenizer(dummy_input, return_tensors="np").input_ids[:, :10]
model = FlaxRobertaModel.from_pretrained("julien-c/dummy-unknown")
# run a forward pass, should return an object `FlaxBaseModelOutputWithPooling`
model(input_ids)
```
## Quickstart flax and jax
[JAX](https://jax.readthedocs.io/en/latest/index.html) is Autograd and XLA, brought together for high-performance numerical computing and machine learning research. It provides composable transformations of Python+NumPy programs: differentiate, vectorize, parallelize, Just-In-Time compile to GPU/TPU, and more. A great place for getting started with JAX is the [JAX 101 Tutorial](https://jax.readthedocs.io/en/latest/jax-101/index.html).
[Flax](https://flax.readthedocs.io/en/latest/index.html) is a high-performance neural network library designed for flexibility built on top of JAX. It aims to provide users with full control of their training code and is carefully designed to work well with JAX transformations such as `grad` and `pmap` (see the [Flax philosophy](https://flax.readthedocs.io/en/latest/philosophy.html)). For an introduction to Flax see the [Flax Basics Colab](https://flax.readthedocs.io/en/latest/notebooks/flax_basics.html) or the list of curated [Flax examples](https://flax.readthedocs.io/en/latest/examples.html).
## Quickstart flax and jax in transformers
Currently, we support the following models in Flax.
Note that some models are about to be merged to `main` and will
be available in a couple of days.
- [BART](https://github.com/huggingface/transformers/blob/main/src/transformers/models/bart/modeling_flax_bart.py)
- [BERT](https://github.com/huggingface/transformers/blob/main/src/transformers/models/bert/modeling_flax_bert.py)
- [BigBird](https://github.com/huggingface/transformers/blob/main/src/transformers/models/big_bird/modeling_flax_big_bird.py)
- [CLIP](https://github.com/huggingface/transformers/blob/main/src/transformers/models/clip/modeling_flax_clip.py)
- [ELECTRA](https://github.com/huggingface/transformers/blob/main/src/transformers/models/electra/modeling_flax_electra.py)
- [GPT2](https://github.com/huggingface/transformers/blob/main/src/transformers/models/gpt2/modeling_flax_gpt2.py)
- [(TODO) MBART](https://github.com/huggingface/transformers/blob/main/src/transformers/models/mbart/modeling_flax_mbart.py)
- [RoBERTa](https://github.com/huggingface/transformers/blob/main/src/transformers/models/roberta/modeling_flax_roberta.py)
- [T5](https://github.com/huggingface/transformers/blob/main/src/transformers/models/t5/modeling_flax_t5.py)
- [ViT](https://github.com/huggingface/transformers/blob/main/src/transformers/models/vit/modeling_flax_vit.py)
- [Wav2Vec2](https://github.com/huggingface/transformers/blob/main/src/transformers/models/wav2vec2/modeling_flax_wav2vec2.py)
You can find all available training scripts for JAX/Flax under the
official [flax example folder](https://github.com/huggingface/transformers/tree/main/examples/flax). Note that a couple of training scripts will be released in the following week.
- [Causal language modeling (GPT2)](https://github.com/huggingface/transformers/blob/main/examples/flax/language-modeling/run_clm_flax.py)
- [Masked language modeling (BERT, RoBERTa, ELECTRA, BigBird)](https://github.com/huggingface/transformers/blob/main/examples/flax/language-modeling/run_mlm_flax.py)
- [Text classification (BERT, RoBERTa, ELECTRA, BigBird)](https://github.com/huggingface/transformers/blob/main/examples/flax/text-classification/run_flax_glue.py)
- [Summarization / Seq2Seq (BART, MBART, T5)](https://github.com/huggingface/transformers/blob/main/examples/flax/summarization/run_summarization_flax.py)
- [Masked Seq2Seq pret-training (T5)](https://github.com/huggingface/transformers/blob/main/examples/flax/language-modeling/run_t5_mlm_flax.py)
- [Contrastive Loss pretraining for Wav2Vec2](https://github.com/huggingface/transformers/blob/main/examples/research_projects/jax-projects/wav2vec2)
- [Fine-tuning long-range QA for BigBird](https://github.com/huggingface/transformers/blob/main/examples/research_projects/jax-projects/big_bird)
- [(TODO) Image classification (ViT)]( )
- [(TODO) CLIP pretraining, fine-tuning (CLIP)]( )
### **Flax design philosophy in Transformers**
This section will explain how Flax models are implemented in Transformers and how the design differs from PyTorch.
Let's first go over the difference between Flax and PyTorch.
In JAX, most transformations (notably `jax.jit`) require functions that are transformed to be stateless so that they have no side effects. This is because any such side-effects will only be executed once when the transformed function is run during compilation and all subsequent calls of the compiled function would re-use the same side-effects of the compiled run instead of the "actual" side-effects (see [Stateful Computations in JAX](https://jax.readthedocs.io/en/latest/jax-101/07-state.html)). As a consequence, Flax models, which are designed to work well with JAX transformations, are stateless. This means that when running a model in inference, both the inputs and the model weights are passed to the forward pass. In contrast, PyTorch model are very much stateful with the weights being stored within the model instance and the user just passing the inputs to the forward pass.
Let's illustrate the difference between stateful models in PyTorch and stateless models in Flax.
For simplicity, let's assume the language model consists simply of a single attention layer [`key_proj`, `value_proj`, `query_proj`] and a linear layer `logits_proj` to project the transformed word embeddings to the output logit vectors.
#### **Stateful models in PyTorch**
In PyTorch, the weights matrices would be stored as `torch.nn.Linear` objects alongside the model's config inside the model class `ModelPyTorch`:
```python
class ModelPyTorch:
def __init__(self, config):
self.config = config
self.key_proj = torch.nn.Linear(config)
self.value_proj = torch.nn.Linear(config)
self.query_proj = torch.nn.Linear(config)
self.logits_proj = torch.nn.Linear(config)
```
Instantiating an object `model_pytorch` of the class `ModelPyTorch` would actually allocate memory for the model weights and attach them to the attributes `self.key_proj`, `self.value_proj`, `self.query_proj`, and `self.logits.proj`. We could access the weights via:
```
key_projection_matrix = model_pytorch.key_proj.weight.data
```
Visually, we would represent an object of `model_pytorch` therefore as follows:
Executing a forward pass then simply corresponds to passing the `input_ids` to the object `model_pytorch`:
```python
sequences = model_pytorch(input_ids)
```
In a more abstract way, this can be represented as passing the word embeddings to the model function to get the output logits:
This design is called **stateful** because the output logits, the `sequences`, can change even if the word embeddings, the `input_ids`, stay the same. Hence, the function's output does not only depend on its inputs, but also on its **state**, `[self.key_proj, self.value_proj, self.query_proj, self.logits_proj]`, which makes `model_pytorch` stateful.
#### **Stateless models in Flax/JAX**
Now, let's see how the mathematically equivalent model would be written in JAX/Flax. The model class `ModelFlax` would define the self-attention and logits projection weights as [**`flax.linen.Dense`**](https://flax.readthedocs.io/en/latest/_autosummary/flax.linen.Dense.html#flax.linen.Dense) objects:
```python
class ModelFlax:
def __init__(self, config):
self.config = config
self.key_proj = flax.linen.Dense(config)
self.value_proj = flax.linen.Dense(config)
self.query_proj = flax.linen.Dense(config)
self.logits_proj = flax.linen.Dense(config)
```
At first glance the linear layer class `flax.linen.Dense` looks very similar to PyTorch's `torch.nn.Linear` class. However, instantiating an object `model_flax` only defines the linear transformation functions and does **not** allocate memory to store the linear transformation weights. In a way, the attribute `self.key_proj` tell the instantiated object `model_flax` to perform a linear transformation on some input and force it to expect a weight, called `key_proj`, as an input.
This time we would illustrate the object `model_flax` without the weight matrices:
Accordingly, the forward pass requires both `input_ids` as well as a dictionary consisting of the model's weights (called `state` here) to compute the `sequences`:
To get the initial `state` we need to explicitly do a forward pass by passing a dummy input:
```python
state = model_flax.init(rng, dummy_input_ids)
```
and then we can do the forward pass.
```python
sequences = model_flax.apply(state, input_ids)
```
Visually, the forward pass would now be represented as passing all tensors required for the computation to the model's object:
This design is called **stateless** because the output logits, the `sequences`, **cannot** change if the word embeddings, the `input_ids`, stay the same. Hence, the function's output only depends on its inputs, being the `input_ids` and the `state` dictionary consisting of the weights **state**, `[key_proj, value_proj, query_proj, logits_proj]`.
Another term which is often used to describe the design difference between Flax/JAX and PyTorch is **immutable** vs **mutable**. A instantiated Flax model, `model_flax`, is **immutable** as a logical consequence of `model_flax`'s output being fully defined by its input: If calling `model_flax` could mutate `model_flax`, then calling `model_flax` twice with the same inputs could lead to different results which would violate the "*statelessness*" of Flax models.
#### **Flax models in Transformers**
Now let us see how this is handled in `Transformers.` If you have used a Flax model in Transformers already, you might wonder how come you don't always have to pass the parameters to the function of the forward pass. This is because the `FlaxPreTrainedModel` class abstracts it away.
It is designed this way so that the Flax models in Transformers will have a similar API to PyTorch and Tensorflow models.
The `FlaxPreTrainedModel` is an abstract class that holds a Flax module, handles weights initialization, and provides a simple interface for downloading and loading pre-trained weights i.e. the `save_pretrained` and `from_pretrained` methods. Each Flax model then defines its own subclass of `FlaxPreTrainedModel`; *e.g.* the BERT model has `FlaxBertPreTrainedModel`. Each such class provides two important methods, `init_weights` and `__call__`. Let's see what each of those methods do:
- The `init_weights` method takes the expected input shape and a [`PRNGKey`](https://jax.readthedocs.io/en/latest/_autosummary/jax.random.PRNGKey.html) (and any other arguments that are required to get initial weights) and calls `module.init` by passing it a random example to get the initial weights with the given `dtype` (for ex. `fp32` or `bf16` etc). This method is called when we create an instance of the model class, so the weights are already initialized when you create a model i.e., when you do
model = FlaxBertModel(config)
- The `__call__` method defines forward pass. It takes all necessary model inputs and parameters (and any other arguments required for the forward pass). The parameters are optional; when no parameters are passed, it uses the previously initialized or loaded parameters which can be accessed using `model.params`. It then calls the `module.apply` method, passing it the parameters and inputs to do the actual forward pass. So we can do a forward pass using
output = model(inputs, params=params)
Let's look at an example to see how this works. We will write a simple two-layer MLP model.
First, write a Flax module that will declare the layers and computation.
```python
import flax.linen as nn
import jax.numpy as jnp
class MLPModule(nn.Module):
config: MLPConfig
dtype: jnp.dtype = jnp.float32
def setup(self):
self.dense1 = nn.Dense(self.config.hidden_dim, dtype=self.dtype)
self.dense2 = nn.Desne(self.config.hidden_dim, dtype=self.dtype)
def __call__(self, inputs):
hidden_states = self.dense1(inputs)
hidden_states = nn.relu(hidden_states)
hidden_states = self.dense2(hidden_states)
return hidden_states
```
Now let's define the `FlaxPreTrainedModel` model class.
```python
from transformers.modeling_flax_utils import FlaxPreTrainedModel
class FlaxMLPPreTrainedModel(FlaxPreTrainedModel):
config_class = MLPConfig
base_model_prefix = "model"
module_class: nn.Module = None
def __init__(self, config: BertConfig, input_shape: Tuple = (1, 8), seed: int = 0, dtype: jnp.dtype = jnp.float32, **kwargs):
# initialize the flax module
module = self.module_class(config=config, dtype=dtype, **kwargs)
super().__init__(config, module, input_shape=input_shape, seed=seed, dtype=dtype)
def init_weights(self, rng, input_shape):
# init input tensors
inputs = jnp.zeros(input_shape, dtype="i4")
params_rng, dropout_rng = jax.random.split(rng)
rngs = {"params": params_rng, "dropout": dropout_rng}
params = self.module.init(rngs, inputs)["params"]
return params
def __call__(self, inputs, params: dict = None):
params = {"params": params or self.params}
outputs = self.module.apply(params, jnp.array(inputs))
return outputs
```
Now we can define our model class as follows.
```python
class FlaxMLPModel(FlaxMLPPreTrainedModel):
module_class = FlaxMLPModule
```
Now the `FlaxMLPModel` will have a similar interface as PyTorch or Tensorflow models and allows us to attach loaded or randomly initialized weights to the model instance.
So the important point to remember is that the `model` is not an instance of `nn.Module`; it's an abstract class, like a container that holds a Flax module, its parameters and provides convenient methods for initialization and forward pass. The key take-away here is that an instance of `FlaxMLPModel` is very much stateful now since it holds all the model parameters, whereas the underlying Flax module `FlaxMLPModule` is still stateless. Now to make `FlaxMLPModel` fully compliant with JAX transformations, it is always possible to pass the parameters to `FlaxMLPModel` as well to make it stateless and easier to work with during training. Feel free to take a look at the code to see how exactly this is implemented for ex. [`modeling_flax_bert.py`](https://github.com/huggingface/transformers/blob/main/src/transformers/models/bert/modeling_flax_bert.py#L536)
Another significant difference between Flax and PyTorch models is that, we can pass the `labels` directly to PyTorch's forward pass to compute the loss, whereas Flax models never accept `labels` as an input argument. In PyTorch, gradient backpropagation is performed by simply calling `.backward()` on the computed loss which makes it very handy for the user to be able to pass the `labels`. In Flax however, gradient backpropagation cannot be done by simply calling `.backward()` on the loss output, but the loss function itself has to be transformed by `jax.grad` or `jax.value_and_grad` to return the gradients of all parameters. This transformation cannot happen under-the-hood when one passes the `labels` to Flax's forward function, so that in Flax, we simply don't allow `labels` to be passed by design and force the user to implement the loss function oneself. As a conclusion, you will see that all training-related code is decoupled from the modeling code and always defined in the training scripts themselves.
### **How to use flax models and example scripts**
#### **How to do a forward pass**
Let's first see how to load, save and do inference with Flax models. As explained in the above section, all Flax models in Transformers have similar API to PyTorch models, so we can use the familiar `from_pretrained` and `save_pretrained` methods to load and save Flax models.
Let's use the base `FlaxRobertaModel` without any heads as an example.
```python
from transformers import FlaxRobertaModel, RobertaTokenizerFast
import jax
tokenizer = RobertaTokenizerFast.from_pretrained("roberta-base")
inputs = tokenizer("JAX/Flax is amazing ", padding="max_length", max_length=128, return_tensors="np")
model = FlaxRobertaModel.from_pretrained("julien-c/dummy-unknown")
@jax.jit
def run_model(input_ids, attention_mask):
# run a forward pass, should return an object `FlaxBaseModelOutputWithPooling`
return model(input_ids, attention_mask)
outputs = run_model(**inputs)
```
We use `jax.jit` to compile the function to get maximum performance. Note that in the above example, we set `padding=max_length` to pad all examples to the same length. We do this because JAX's compiler has to recompile a function everytime its input shape changes - in a sense a compiled function is not only defined by its code but also by its input and output shape. It is usually much more effective to pad the input to be of a fixed static shape than having to recompile every the function multiple times.
#### **How to write a training loop**
Now let's see how we can write a simple training loop to train Flax models, we will use `FlaxGPT2ForCausalLM` as an example.
A training loop for Flax models typically consists of
- A loss function that takes the parameters and inputs, runs the forward pass and returns the loss.
- We then transform the loss function using `jax.grad` or `jax.value_and_grad` so that we get the gradients of all parameters.
- An optimizer to update the paramteres using the gradients returned by the transformed loss function.
- A train step function which combines the loss function and optimizer update, does the forward and backward pass and returns the updated parameters.
Lets see how that looks like in code:
First initialize our model
```python
import jax
import jax.numpy as jnp
from transformers import FlaxGPT2ForCausalLM
model = FlaxGPT2ForCausalLM(config)
```
As explained above we don't compute the loss inside the model, but rather in the task-specific training script.
For demonstration purposes, we write a pseudo training script for causal language modeling in the following.
```python
from flax.training.common_utils import onehot
def cross_entropy(logits, labels):
return -jnp.sum(labels * jax.nn.log_softmax(logits, axis=-1), axis=-1)
# define a function which will run the forward pass return loss
def compute_loss(params, input_ids, labels):
logits = model(input_ids, params=params, train=True)
num_classes = logits.shape[-1]
loss = cross_entropy(logits, onehot(labels, num_classes)).mean()
return loss
```
Now we transform the loss function with `jax.value_and_grad`.
```python
# transform the loss function to get the gradients
grad_fn = jax.value_and_grad(compute_loss)
```
We use the [optax](https://github.com/deepmind/optax) library to Initialize the optimizer.
```python
import optax
params = model.params
tx = optax.sgd(learning_rate=3e-3)
opt_state = tx.init(params)
```
Now we define a single training step which will do a forward and a backward pass.
```python
def _train_step(params, opt_state, input_ids, labels)
# do the forward pass and get the loss and gradients
loss, grads = grad_fn(params, input_ids, labels)
# use the gradients to update parameters
updates, opt_state = tx.update(grads, opt_state)
updated_params = optax.apply_updates(params, updates)
return updates_params, opt_state, loss
train_step = jax.jit(_train_step)
```
Finally, let's run our training loop.
```python
# train loop
for i in range(10):
params, opt_state, loss = train_step(params, opt_state, input_ids, labels)
```
Note how we always pass the `params` and `opt_state` to the `train_step` which then returns the updated `params` and `opt_state`. This is because of the staless nature of JAX/Flax models, all the state
like parameters, optimizer state is kept external.
We can now save the model with the trained parameters using
```python
model.save_pretrained("awesome-flax-model", params=params)
```
Note that, as JAX is backed by the [XLA](https://www.tensorflow.org/xla) compiler any JAX/Flax code can run on all `XLA` compliant device without code change!
That menas you could use the same training script on CPUs, GPUs, TPUs.
To know more about how to train the Flax models on different devices (GPU, multi-GPUs, TPUs) and use the example scripts, please look at the [examples README](https://github.com/huggingface/transformers/tree/main/examples/flax).
## Talks
3 days of talks around JAX / Flax, Transformers, large-scale language modeling and other great topics during our community event!
### Wednesday, June 30th
- [Watch the talks on YouTube](https://www.youtube.com/watch?v=fuAyUQcVzTY)
- [Chat history](https://docs.google.com/spreadsheets/d/1PZ5xYV2hVwlAVQSqDag65ympv5YNCSDmXyG-eWTaZ_o/edit?usp=sharing)
Speaker | Topic | Time | Video |
|-------------|---------------------------------|------------------------|------------------------|
| Skye Wanderman-Milne, Google Brain | Intro to JAX on Cloud TPUs | 6.00pm-6.45pm CEST / 9.00am-9.45am PST | [](https://www.youtube.com/watch?v=fuAyUQcVzTY) |
| Marc van Zee, Google Brain | Introduction to Flax | 6.45pm-7.30pm CEST / 9.45am-10.30am PST | [](https://youtu.be/fuAyUQcVzTY?t=2569) |
| Pablo Castro, Google Brain | Using Jax & Flax for RL with the Dopamine library | 7.30pm-8.00pm CEST / 10.30am-11.00am PST | [](https://youtu.be/fuAyUQcVzTY?t=5306) |
### Thursday, July 1st
- [Watch the talks on YouTube](https://www.youtube.com/watch?v=__eG63ZP_5g)
- [Chat history](https://docs.google.com/spreadsheets/d/1PZ5xYV2hVwlAVQSqDag65ympv5YNCSDmXyG-eWTaZ_o/edit#gid=1515796400)
Speaker | Topic | Time | Video |
|-------------|---------------------------------|------------------------|------------------------|
| Suraj Patil & Patrick von Platen, Hugging Face | How to use JAX/Flax with Transformers | 5.30pm-6.00pm CEST / 8.30am-9.00am PST | [](https://www.youtube.com/watch?v=__eG63ZP_5g) |
| Sabrina J. Mielke, Johns Hopkins University & HuggingFace | From stateful code to purified JAX: how to build your neural net framework | 6.00pm-6.30pm CEST / 9.00am-9.30am PST | [](https://youtu.be/__eG63ZP_5g?t=1576) |
| Mostafa Dehghani, Google Brain | Long Range Arena: Benchmarking Efficient Transformers | 6.30pm-7.00pm CEST / 9.30am-10.00am PST | [](https://youtu.be/__eG63ZP_5g?t=3695) |
| Rohan Anil, Google Brain | Scalable Second Order Optimization for Deep Learning | 7.00pm-7.30pm CEST / 10.00am-10.30am PST | [](https://youtu.be/__eG63ZP_5g?t=5285) |
### Friday, July 2nd
- [Watch the talks on YouTube](https://www.youtube.com/watch?v=ZCMOPkcTu3s)
- [Chat history](https://docs.google.com/spreadsheets/d/1PZ5xYV2hVwlAVQSqDag65ympv5YNCSDmXyG-eWTaZ_o/edit#gid=1166061401)
Speaker | Topic | Time | Video |
|-------------|---------------------------------|------------------------|------------------------|
| Lucas Beyer, Google Brain | Vision Transformer | 5.00pm-5.30 CEST / 8.00am-8.30 PST | [](https://www.youtube.com/watch?v=ZCMOPkcTu3s) |
| Ben Wang, EleutherAI | Multihost Training in Mesh Transformer JAX | 5.30pm-6.00 CEST / 8.30am-9.00 PST | [](https://youtu.be/ZCMOPkcTu3s?t=1803) |
| Iurii Kemaev, Soňa Mokrá, Junhyuk Oh, DeepMind | DeepMind JAX Ecosystem | 6.00pm-6.30 CEST / 9.00am-9.30am PST | [](https://youtu.be/ZCMOPkcTu3s?t=3388) |
| Siddhartha Kamalakara, Joanna Yoo & João G M Araújo, Cohere | Training large scale language models | 6:30pm-7.00pm CEST / 9:30am-10.00am PST | [](https://youtu.be/ZCMOPkcTu3s?t=5095) |
### Talks & Speakers
#### Skye Wanderman-Milne, JAX developer, Google Brain
- Talk: Intro to JAX on Cloud TPUs
- Abstract: JAX is a system for high-performance machine-learning research that combines the familiarity of Python + NumPy together with the power of hardware acceleration on CPUs, GPUs, and TPUs. It offers composable function transformations for automatic differentiation, automatic batching, end-to-end compilation, and both data and model parallelism. This talk will show you how to get up and running with JAX on a Cloud TPU VM.
- Speaker info: Skye Wanderman-Milne is a software engineer working on JAX. She has previously worked on TensorFlow and Apache Impala, a high-performance distributed database.
#### Marc van Zee, Research SWE, Google Brain (Flax team)
- Talk: Introduction to Flax
- Abstract: In this talk I will provide a high-level introduction to the neural network library Flax. I will discuss the Flax philosophy, talk about the ecosystem around Flax and provide a high-level introduction to the code. I explain the Module abstraction and how to use it to train your models.
- Speaker info: Marc is at Google Research for over 4 years. First he worked on conceptual AI, developing a next generation language understanding and reasoning prototype and he authored the CFQ dataset for compositional generalization. Currently, Marc works as a research software engineer in the Flax team.
#### Pablo Castro, Staff Research Software Developer; Google Research, Brain Team
- Talk: Using Jax & Flax for RL with the Dopamine library
- Abstract: The Dopamine library was launched with TensorFlow in 2018 and we added a Jax/Flax variant of it last year. Internally, Jax's flexibility has facilitated our RL research tremendously, and we are excited to demonstrate its potential.
- Speaker info: Pablo Samuel has been at Google for over 9 years, and is currently a researcher with the Brain team, focusing on fundamental reinforcement learning, as well as machine learning and creativity. Aside from his research, Pablo Samuel is an active musician (with a channel exploring the intersection of music and computer science), and is helping increase the representation of the LatinX community in the research world.
- Dopamine repo: https://github.com/google/dopamine
- Homepage: https://psc-g.github.io/
- Twitter: https://twitter.com/pcastr
#### Suraj Patil & Patrick von Platen, Machine Learning Engineers at Hugging Face
- Talk: How to use JAX/Flax with Transformers
- Abstract: Transformers is one of the most popular open-source ML libraries and supports PyTorch, Tensorflow, and JAX/Flax. In this talk, we will explain how JAX/Flax models should be used in Transformers and compare their design in Transformers with the design of PyTorch models in Transformers. In the second part, we will give you a hands-on presentation of how a model can be trained end-to-end with the official JAX/Flax example scripts using Transformers & Datasets. Along the way, we want to give you some tips and tricks on how to best realize your project.
- Speaker info: Suraj and Patrick are part of Hugging Face’s open source team and lead the integration of JAX/Flax into Transformers.
- GitHub: https://github.com/patil-suraj & https://github.com/patrickvonplaten
#### Sabrina J. Mielke, PhD student at The Johns Hopkins University & Part-time research intern at HuggingFace
- Talk: From stateful code to purified JAX: how to build your neural net framework
- Abstract: Moving from object-oriented (and stateful) PyTorch- or TF2-code with tape-based backprop to JAX isn't easy---and while running grad() on numpy-oneliners is cool and all, you do wonder... how do I build actual big neural nets? Libraries like flax, trax, or haiku make it easy---but how could you build machinery like that yourself?
- Speaker info: Sabrina is a PhD student at the Johns Hopkins University and a part-time research intern at HuggingFace, researching open-vocabulary language models for segmentation and tokenization. She has published and co-organized workshops and shared tasks on these topics as well as on morphology and typological analysis in ACL, NAACL, EMNLP, LREC, and AAAI. You can find her reminisce for a time when formal language theory played a bigger role in NLP on Twitter at @sjmielke.
- Links: The 2020 blogpost this talk will be based on: https://sjmielke.com/jax-purify.htm, leading to our experiment Parallax and eventually Haiku
#### Mostafa Dehghani, Research Scientist, Google Brain
- Talk: Long Range Arena: Benchmarking Efficient Transformers
- Abstract: Transformers do not scale very well to long sequence lengths largely because of quadratic self-attention complexity. In the recent months, a wide spectrum of efficient, fast Transformers have been proposed to tackle this problem, more often than not claiming superior or comparable model quality to vanilla Transformer models. So, we now need a well-established consensus on how to evaluate this class of models. Moreover, inconsistent benchmarking on a wide spectrum of tasks and datasets makes it difficult to assess relative model quality amongst many models. I'll talk about a systematic and unified benchmark, LRA, specifically focused on evaluating model quality under long-context scenarios. LRA is a suite of tasks consisting of sequences ranging from 1K to 16K tokens, encompassing a wide range of data types and modalities such as text, natural, synthetic images, and mathematical expressions requiring similarity, structural, and visual-spatial reasoning. We systematically evaluate ten well-established long-range Transformer models (Reformers, Linformers, Linear Transformers, Sinkhorn Transformers, Performers, Synthesizers, Sparse Transformers, and Longformers) on LRA. LRA paves the way towards better understanding this class of efficient Transformer models, facilitates more research in this direction, and presents new challenging tasks to tackle.
- Speaker info: https://mostafadehghani.com/
#### Rohan Anil, Senior Staff Software Engineer, Google Research, Brain Team
- Talk: Scalable Second Order Optimization for Deep Learning
- Abstract: Optimization in machine learning, both theoretical and applied, is presently dominated by first-order gradient methods such as stochastic gradient descent. Second-order optimization methods, that involve second derivatives and/or second order statistics of the data, are far less prevalent despite strong theoretical properties, due to their prohibitive computation, memory and communication costs. In an attempt to bridge this gap between theoretical and practical optimization, we present a scalable implementation of a second-order preconditioned method (concretely, a variant of full-matrix Adagrad), that along with several critical algorithmic and numerical improvements, provides significant convergence and wall-clock time improvements compared to conventional first-order methods on state-of-the-art deep models. Our novel design effectively utilizes the prevalent heterogeneous hardware architecture for training deep models, consisting of a multicore CPU coupled with multiple accelerator units. We demonstrate superior performance compared to state-of-the-art on very large learning tasks such as machine translation with Transformers, language modeling with BERT, click-through rate prediction on Criteo, and image classification on ImageNet with ResNet-50.
- Speaker info: Rohan Anil is a software engineer at Google Research, Mountain View. Lately, he has been working on scalable and practical optimization techniques for efficient training of neural networks in various regimes.
- Resources:
- https://arxiv.org/abs/2002.09018
- https://arxiv.org/abs/1901.11150
- https://arxiv.org/abs/2106.06199
#### Lucas Beyer, Senior Research Engineer, Google Brain
- Talk: Vision Transformer
- Abstract: This talk will discuss the learning of general visual representations via large-scale pre-training and few-shot transfer, with a special focus on the Vision Transformer (ViT) architecture, which popularized transformers for the visual domain.
- Speaker info: Lucas Beyer is a self-taught hacker and studied engineer. He went on to do his PhD in robotic perception at RWTH Aachen and is currently on a quest to find the ultimate visual representation at Google Brain in Zürich
#### Ben Wang, Independent AI Researcher, EleutherAI
- Talk: Multihost Training in Mesh Transformer JAX
- Abstract: As models become larger, training must be scaled across multiple nodes. This talk discusses some design decisions and tradeoffs made for scaling to multiple nodes in Mesh Transformer JAX, a library for running model parallel transformers on TPU pods.
- Speaker info: Ben is an independent AI researcher who contributes to EleutherAI, an open source research collective centered around democratizing access to powerful AI models. Recently he has released GPT-J-6B, a 6 billion parameter transformer which is the most powerful autoregressive language model in terms of zero-shot performance with public weights.
- Website: https://www.eleuther.ai/
#### Iurii Kemaev, Research Engineer, Soňa Mokrá, Research Engineer, and Junhyuk Oh, Research Scientist, DeepMind
- Talk: DeepMind JAX Ecosystem
- Abstract: The DeepMind JAX Ecosystem is an effort to build a shared substrate of components to enable all aspects of AGI Research. In this talk, our researchers and engineers will give a high-level overview of our Ecosystem goals and design philosophies, using our Haiku (neural network), Optax (optimization) and RLax (reinforcement learning) libraries as examples. We will then deep dive on two examples of recent DeepMind research that have been enabled by JAX and these libraries: generative models and meta-gradient reinforcement learning.
- Speaker info:
- Iurii Kemaev is a Research Engineer at DeepMind. He has been using JAX for 2 years advancing RL research. Iurii is one of the DM JAX ecosystem leads.
- Soňa Mokrá is a Research Engineer at DeepMind. She has a background in machine translation and has been using JAX as the main ML framework for the past 6 months.
- Junhyuk Oh is a Research Scientist at DeepMind, working on reinforcement learning and meta-learning. More information is available at https://junhyuk.com/
#### Siddhartha Kamalakara, Joanna Yoo, João G M Araújo, MLE at Cohere
- Talk: Training large scale language models
- Abstract: A journey through Cohere’s experiences with training large scale language models. Join us in our exploration of pipeline and model parallelism as strategies for efficient training of large language models. We will present and motivate our recent transition to JAX+Flax as our choice of internal tech stack.
- Speaker info:
- João G M Araújo is a Brazilian college student with a passion for mathematics and a fascination for Deep Learning. João conducted research on representation learning and spent 3 months in Japan working on NeuroEvolution. João likes reading fantasy books and spending quality time with family and friends, and also runs a YouTube series on theoretical understanding of Deep Learning where researchers talk about their findings
- Joanna Yoo is one of the founding engineers at Cohere, working on scaling language models for the last year and half. Joanna loves live concerts and rock climbing!
- Siddhartha Rao Kamalakara is an MLE at Cohere and a researcher at FOR.ai with research interests at the intersection of efficient training and empirical understanding of DL.
- Website: https://cohere.ai/
## How to use the hub for collaboration
In this section, we will explain how a team can use the 🤗 hub to collaborate on a project.
The 🤗 hub allows each team to create a repository with integrated git version control that
should be used for their project.
The advantages of using a repository on the 🤗 hub are:
- easy collaboration - each team member has write access to the model repository
- integrated git version control - code scripts as well as large model files are tracked using git version control
- easy sharing - the hub allows each team to easily share their work during and after the event
- integrated tensorboard functionality - uploaded tensorboard traces are automatically displayed on an integrated tensorboard tab
We highly recommend each team to make use of the 🤗 hub during the event.
To better understand how the repository and the hub in general functions, please take a look at the documentation and the videos [here](https://huggingface.co/docs/hub).
Now let's explain in more detail how a project can be created on the hub. Having an officially defined project on [this](https://docs.google.com/spreadsheets/d/1GpHebL7qrwJOc9olTpIPgjf8vOS0jNb6zR_B8x_Jtik/edit?usp=sharing) Google Sheet you should be part of [the Flax Community organization on the hub](https://huggingface.co/flax-community). All repositories should be created under this organization so that write access can be shared and everybody can easily access other participants'
work 🤗. Note that we are giving each team member access to all repositories created under [flax-community](https://huggingface.co/flax-community), but we encourage participants to only clone and edit repositories corresponding to one's teams. If you want to help other teams, please ask them before changing files in their repository! The integrated git version control keeps track of
all changes, so in case a file was deleted by mistake, it is trivial to re-create it.
Awesome! Now, let's first go over a simple example where most of the required we'll pre-train a RoBERTa model on a low-resource language. To begin with, we create a repository
under [the Flax Community organization on the hub](https://huggingface.co/flax-community) by logging in to the hub and going to [*"Add model"*](https://huggingface.co/new). By default
the username should be displayed under "*Owner*", which we want to change to *flax-community*. Next, we give our repository a fitting name for the project - here we'll just call it
*roberta-base-als* because we'll be pretraining a RoBERTa model on the super low-resource language *Alemannic* (`als`). We make sure that the model is a public repository and create it!
It should then be displayed on [the Flax Community organization on the hub](https://huggingface.co/flax-community).
Great, now we have a project directory with integrated git version control and a public model page, which we can access under [flax-community/roberta-base-als](https://huggingface.co/flax-community/roberta-base-als). Let's create a short README so that other participants know what this model is about. You can create the README.md directly on the model page as a markdown file.
Let's now make use of the repository for training.
We assume that the 🤗 Transformers library and [git-lfs](https://git-lfs.github.com/) are correctly installed on our machine or the TPU attributed to us.
If this is not the case, please refer to the [Installation guide](#how-to-install-relevant-libraries) and the official [git-lfs](https://git-lfs.github.com/) website.
At first we should log in:
```bash
$ huggingface-cli login
```
Next we can clone the repo:
```bash
$ git clone https://huggingface.co/flax-community/roberta-base-als
```
We have now cloned the model's repository and it should be under `roberta-base-als`. As you can see,
we have all the usual git functionalities in this repo - when adding a file, we can do `git add .`, `git commit -m "add file"` and `git push`
as usual. Let's try it out by adding the model's config.
We go into the folder:
```bash
$ cd ./roberta-base-als
```
and run the following commands in a Python shell to save a config.
```python
from transformers import RobertaConfig
config = RobertaConfig.from_pretrained("roberta-base")
config.save_pretrained("./")
```
Now we've added a `config.json` file and can upload it by running
```bash
$ git add . && git commit -m "add config" && git push
```
Cool! The file is now displayed on the model page under the [files tab](https://huggingface.co/flax-community/roberta-base-als/tree/main).
We encourage you to upload all files except maybe the actual data files to the repository. This includes training scripts, model weights,
model configurations, training logs, etc...
Next, let's create a tokenizer and save it to the model dir by following the instructions of the [official Flax MLM README](https://github.com/huggingface/transformers/tree/main/examples/flax/language-modeling#train-tokenizer). We can again use a simple Python shell.
```python
from datasets import load_dataset
from tokenizers import ByteLevelBPETokenizer
# load dataset
dataset = load_dataset("oscar", "unshuffled_deduplicated_als", split="train")
# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()
def batch_iterator(batch_size=1000):
for i in range(0, len(dataset), batch_size):
yield dataset[i: i + batch_size]["text"]
# Customized training
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
"<s>",
"<pad>",
"</s>",
"<unk>",
"<mask>",
])
# Save files to disk
tokenizer.save("./tokenizer.json")
```
This creates and saves our tokenizer directly in the cloned repository.
Finally, we can start training. For now, we'll simply use the official [`run_mlm_flax`](https://github.com/huggingface/transformers/blob/main/examples/flax/language-modeling/run_mlm_flax.py)
script, but we might make some changes later. So let's copy the script into our model repository.
```bash
$ cp ~/transformers/examples/flax/language-modeling/run_mlm_flax.py ./
```
This way we are certain to have all the code used to train the model tracked in our repository.
Let's start training by running:
```bash
./run_mlm_flax.py \
--output_dir="./" \
--model_type="roberta" \
--config_name="./" \
--tokenizer_name="./" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_als" \
--max_seq_length="128" \
--per_device_train_batch_size="4" \
--per_device_eval_batch_size="4" \
--learning_rate="3e-4" \
--warmup_steps="1000" \
--overwrite_output_dir \
--num_train_epochs="8" \
--push_to_hub
```
Since the dataset is tiny this command should actually run in less than 5 minutes. Note that we attach
the flag ``--push_to_hub`` so that both model weights and tensorboard traces are automatically uploaded to the hub.
You can see the tensorboard directly on the model page, under the [Training metrics tab](https://huggingface.co/flax-community/roberta-base-als/tensorboard).
As you can see, it is pretty simple to upload model weights and training logs to the model hub. Since the repository
has git version control, you & your team probably already have the necessary skills to collaborate. Thanks
to `git-lfs` being integrated into the hub, model weights and other larger file can just as easily be uploaded
and changed. Finally, at Hugging Face, we believe that the model hub is a great platform to share your project
while you are still working on it:
- Bugs in training scripts can be found and corrected by anybody participating in the event
- Loss curves can be analyzed directly on the model page
- Model weights can be accessed and analyzed by everybody from the model repository
If you are not using a transformers model, don't worry - you should still be able to make use of the hub's functionalities!
The [huggingface_hub](https://github.com/huggingface/huggingface_hub) allows you to upload essentially any JAX/Flax model to the hub with
just a couple of lines of code. *E.g.* assuming you want to call your model simply `flax-model-dummy`, you can upload it to the hub with
just three lines of code:
```python
from flax import serialization
from jax import random
from flax import linen as nn
from huggingface_hub import Repository
model = nn.Dense(features=5)
key1, key2 = random.split(random.PRNGKey(0))
x = random.normal(key1, (10,))
params = model.init(key2, x)
bytes_output = serialization.to_bytes(params)
repo = Repository("flax-model", clone_from="flax-community/flax-model-dummy", token=True)
with repo.commit("My cool Flax model :)"):
with open("flax_model.msgpack", "wb") as f:
f.write(bytes_output)
# Repo is created and available here: https://huggingface.co/flax-community/flax-model-dummy
```
**Note**: Make sure to have `huggingface_hub >= 0.0.13` to make this command work.
For more information, check out [this PR](https://github.com/huggingface/huggingface_hub/pull/143) on how to upload any framework to the hub.
## How to setup TPU VM
In this section we will explain how you can ssh into a TPU VM that has been given to your team.
If your username is in one of the officially defined projects [here](https://docs.google.com/spreadsheets/d/1GpHebL7qrwJOc9olTpIPgjf8vOS0jNb6zR_B8x_Jtik/edit?usp=sharing), you should have received two emails:
- one that states that you have been granted the role "Community Week Participants" for the project hf-flax, and
- one (or more if you are in multiple projects) that gives you the TPU name and the TPU zone for the TPU of your team
You should click on "Open Cloud Console" on the first mail and agree to the pop up windows that follows. It will allow you to use a TPU VM. Don't worry if you cannot access the actual project `hf-flax` visually on the google cloud console and receive an error:
```
You don't have sufficient permission to view this page
```
- this is expected!
Great, now you and your team can access your TPU VM!
In the following, we will describe how to do so using a standard console, but you should also be able to connect to the TPU VM via IDEs, like Visual Studio Code, etc.
1. You need to install the Google Cloud SDK. Please follow the instructions on [cloud.google.com/sdk](https://cloud.google.com/sdk/docs/install#linux).
2. Once you've installed the google cloud sdk, you should set your account by running the following command. Make sure that `<your-email-address>` corresponds to the gmail address you used to sign up for this event.
```bash
$ gcloud config set account <your-email-adress>
```
3. Let's also make sure the correct project is set in case your email is used for multiple gcloud projects:
```bash
$ gcloud config set project hf-flax
```
4. Next, you will need to authenticate yourself. You can do so by running:
```bash
$ gcloud auth login
```
This should give you a link to a website, where you can authenticate your gmail account.
5. Finally, you can ssh into the TPU VM! Please run the following command by setting <zone> to either `europe-west4-a` or `us-central1-a` (depending on what is stated in the second email you received) and <tpu-name> to the TPU name also sent to you in the second email.
```bash
$ gcloud alpha compute tpus tpu-vm ssh <tpu-name> --zone <zone> --project hf-flax
```
This should ssh you into the TPU VM!
Now you can follow the steps of the section [How to install relevant libraries](#how-to-install-relevant-libraries) to install all necessary
libraries. Make sure to carefully follow the explanations of the "**IMPORTANT**" statement to correctly install JAX on TPU.
Also feel free to install other `python` or `apt` packages on your machine if it helps you to work more efficiently!
## How to build a demo
### Using the Hugging Face Widgets
Hugging Face has over [15 widgets](https://huggingface-widgets.netlify.app/) for different use cases using 🤗 Transformers library. Some of them also support [3rd party libraries](https://huggingface.co/docs/hub/libraries) such as [Sentence Similarity](https://huggingface.co/sentence-transformers/paraphrase-xlm-r-multilingual-v1) with Sentence Transformers and [Text to Speech](https://huggingface.co/julien-c/ljspeech_tts_train_tacotron2_raw_phn_tacotron_g2p_en_no_space_train) with [ESPnet](https://github.com/espnet/espnet).
All the widgets are open sourced in the `huggingface_hub` [repo](https://github.com/huggingface/huggingface_hub/tree/main/widgets). Here is a summary of existing widgets:
**NLP**
* **Conversational:** To have the best conversations!. [Example](https://huggingface.co/microsoft/DialoGPT-large?).
* **Feature Extraction:** Retrieve the input embeddings. [Example](https://huggingface.co/sentence-transformers/distilbert-base-nli-mean-tokens?text=test).
* **Fill Mask:** Predict potential words for a mask token. [Example](https://huggingface.co/bert-base-uncased?).
* **Question Answering:** Given a context and a question, predict the answer. [Example](https://huggingface.co/bert-large-uncased-whole-word-masking-finetuned-squad).
* **Sentence Simmilarity:** Predict how similar a set of sentences are. Useful for Sentence Transformers.
* **Summarization:** Given a text, output a summary of it. [Example](https://huggingface.co/sshleifer/distilbart-cnn-12-6).
* **Table Question Answering:** Given a table and a question, predict the answer. [Example](https://huggingface.co/google/tapas-base-finetuned-wtq).
* **Text Generation:** Generate text based on a prompt. [Example](https://huggingface.co/gpt2)
* **Token Classification:** Useful for tasks such as Named Entity Recognition and Part of Speech. [Example](https://huggingface.co/dslim/bert-base-NER).
* **Zero-Shot Classification:** Too cool to explain with words. Here is an [example](https://huggingface.co/typeform/distilbert-base-uncased-mnli)
* ([WIP](https://github.com/huggingface/huggingface_hub/issues/99)) **Table to Text Generation**.
**Speech**
* **Audio to Audio:** For tasks such as audio source separation or speech enhancement.
* **Automatic Speech Recognition:** Convert audio to text. [Example](https://huggingface.co/facebook/wav2vec2-base-960h)
* **Text to Speech**: Convert text to audio.
**Image**
* **Image Classification:** Given an image, predict its class. [Example](https://huggingface.co/osanseviero/llamastic).
* ([WIP](https://github.com/huggingface/huggingface_hub/issues/100)) **Zero Shot Image Classification**
* ([WIP](https://github.com/huggingface/huggingface_hub/issues/112)) **Image Captioning**
* ([WIP](https://github.com/huggingface/huggingface_hub/issues/113)) **Text to Image Generation**
* ([Proposed](https://github.com/huggingface/huggingface_hub/issues/127)) **Visual Question Answering**
You can propose and implement new widgets by [opening an issue](https://github.com/huggingface/huggingface_hub/issues). Contributions are welcomed!
### Using a Streamlit demo
Sometimes you might be using different libraries or a very specific application that is not well supported by the current widgets. In this case, [Streamlit](https://streamlit.io/) can be an excellent option to build a cool visual demo. Setting up a Streamlit application is straightforward and in Python!
A common use case is how to load files you have in your model repository in the Hub from the Streamlit demo. The `huggingface_hub` library is here to help you!
```
pip install huggingface_hub
```
Here is an example downloading (and caching!) a specific file directly from the Hub
```
from huggingface_hub import hf_hub_download
filepath = hf_hub_download("flax-community/roberta-base-als", "flax_model.msgpack");
```
In many cases you will want to download the full repository. Here is an example downloading all the files from a repo. You can even specify specific revisions!
```
from huggingface_hub import snapshot_download
local_path = snapshot_download("flax-community/roberta-base-als");
```
Note that if you're using 🤗 Transformers library, you can quickly load the model and tokenizer as follows
```
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("REPO_ID")
model = AutoModelForMaskedLM.from_pretrained("REPO_ID")
```
We'll provide more examples on Streamlit demos next week. Stay tuned!
### Using a Gradio demo
You can also use [Gradio](https://gradio.app/) to share your demos! [Here](https://huggingface.co/blog/gradio) is an example using the Gradio library to create a GUI for a Hugging Face model.
More to come!
## Project evaluation
For your project to be evaluated, please fill out [this google form](https://forms.gle/jQaMkj3JJdD4Xcwn9).
Please make sure that your submitted project includes a demo as well as information about the model, data, training methods, etc.
### Criteria
* **Demo.** All projects are required to have a demo. It’s open ended, but we provide some ideas on how to build demos in the [How to build a demo](#how-to-build-a-demo) section.
* **Technical difficulty.** Difficulty has different aspects, such as working with complex architectures, obtaining better evaluation metrics than existing models, or implementing models for low-resource languages.
* **Social impact.** The project is expected to have a positive social impact, e.g. by tackling under-explored area of practical interest for minorities or under-represented group (low-ressources languages, specific focus on bias, fairness or ethical issues in ML) or by tackling general societal challenges, e.g. health or climate related challenges.
* **Innovativeness.** Projects that propose novel applications or bring new ideas will be rewarded more.
### Jury
* [Niki Parmar](https://research.google/people/NikiParmar/): Staff Research Scientist at Google.
* [Ross Wightman](https://www.linkedin.com/in/wightmanr/): Angel Investor.
* [Thomas Wolf](https://www.linkedin.com/in/thomas-wolf-a056857/): Co-founder and CSO at Hugging Face.
* [Ashish Vaswani](https://research.google/people/AshishVaswani/): Staff Research Scientist at Google Brain.
### Process
* **July 17, 12h00 CEST**: TPU VM access closes.
* **July 19, 12h00 CEST**: Project completition ends (including demo).
* **July 19-21** A group of event organizers (Suraj, Patrick, Suzana, and Omar) will do an initial filter to find the top 15 projects.
* **July 22-26** The jury will go over the 15 projects and pick the top three projects out of them.
* **July 27.** Winner projects are announced
## General tips and tricks
TODO (will be filled continuously)...
## FAQ
TODO (will be filled continuously)...
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/HOW_TO_PROPOSE_PROJECT.md
|
# How to propose a Flax/JAX + Transformers project
Great that you've opened this document!
While we at 🤗 are proposing a couple of projects, we strongly
believe that the community can come up with much more **creative**, **fun**, and
**impactful** projects on their own. This being said, we are really looking forward
to seeing your project proposal!
## What a project should be about
The proposed project should fall into the machine learning fields of **Natural Language Processing (NLP)** and/or **Computer Vision (CV)** (possibly also **Speech Recognition (ASR)** depending on whether Speech Recognition models are available in Flax in due time) and aim at solving a specific task.
Possible tasks can belong to:
* text classification
* text generation
* image recognition
* image processing
* image captioning
* audio classification
* and other tasks you can think of!
The clearer a task is defined, the better your project proposal is.
*E.g.* "Using a T5 model to learn grammar correction in French" or "Adapting a pre-trained CLIP model for zero-shot image classification in Spanish" are **well-defined and clear** project proposals, while something like "Train a language model" or "Image classification" are **too vague**.
There is no limit to your creativity as long as the project is feasible and ethical.
The more creative & specific your project proposal, the more interesting it will be,
and the more likely will you find motivated team members to work on your project!
To get an idea of how to formulate your project proposals, you can browse through
existing project proposals on the [forum](https://discuss.huggingface.co/c/flax-jax-projects/22).
## How to submit a project proposal
First, you should make sure that you are [logged in](https://huggingface.co/login?sso=bm9uY2U9OTRlNjZjZmZhYjMwMmJmMWMyYjc5MmFiMTMyMzY5ODYmcmV0dXJuX3Nzb191cmw9aHR0cHMlM0ElMkYlMkZkaXNjdXNzLmh1Z2dpbmdmYWNlLmNvJTJGc2Vzc2lvbiUyRnNzb19sb2dpbg%3D%3D&sig=429ad8924bcb33c40f9823027ea749abb55d393f4f58924f36a2dba3ab0a48da) with your Hugging Face account on the forum.
Second, make sure that your project idea doesn't already exist by checking [existing projects](https://discuss.huggingface.co/c/flax-jax-projects/22).
If your project already exists - great! This means that you can comment and improve
the existing idea and join the project to form a team! If your project idea already
exists for a different language, feel free to submit the same project idea, just in
a different language.
Third, having ensured that your project doesn't exist, click on the *"New Topic"*
button on the [Flax/JAX Projects Forum category](https://discuss.huggingface.co/c/flax-jax-projects/22) to create a new project proposal.
Fourth, make sure that your project proposal includes the following information:
1. *A clear description of the project*
2. *In which language should the project be conducted?* English, German, Chinese, ...? It can also be a multi-lingual project
3. *Which model should be used?* If you want to adapt an existing model, you can add the link to one of the 4000 available checkpoints in JAX [here](https://huggingface.co/models?filter=jax) If you want to train a model from scratch, you can simply state the model architecture to be used, *e.g.* BERT, CLIP, etc. You can also base your project on a model that is not part of transformers. For an overview of libraries based on JAX, you can take a look at [awesome-jax](https://github.com/n2cholas/awesome-jax#awesome-jax-). **Note** that for a project that is not based on Transformers it will be more difficult for the 🤗 team to help you. Also have a look at the section [Quickstart Flax & Jax in Transformers](https://github.com/huggingface/transformers/tree/main/examples/research_projects/jax-projects#quickstart-flax-and-jax-in-transformers) to see what model architectures are currently supported in 🤗 Transformers.
4. *What data should be used?* It is important to state at least what kind of data you would like to use. Ideally, you can already point to publicly available data or a dataset in the 🤗 Datasets library.
5. *Are similar training scripts available in Flax/JAX?* It would be important to find similar training scripts that already exist in Flax/JAX. *E.g.* if you are working on a Seq-to-Seq task, you can make use of the [`run_summarization_flax.py`](https://github.com/huggingface/transformers/blob/main/examples/flax/summarization/run_summarization_flax.py) script which is very similar to any seq2seq training. Also have a look at the section [Quickstart Flax & Jax in Transformers](https://github.com/huggingface/transformers/tree/main/examples/research_projects/jax-projects#quickstart-flax-and-jax-in-transformers) to see what training scripts are currently supported in 🤗 Transformers.
6. *(Optionally) What are possible challenges?* List possible difficulties with your project. *E.g.* If you know that training convergence usually takes a lot of time, it is worth stating this here!
7. *(Optionally) What is the desired project outcome?* - How would you like to demo your project? One could *e.g.* create a Streamlit application.
8. *(Optionally) Links to read upon* - Can you provide any links that would help the reader to better understand your project idea?
Feel free to copy-paste the following format for your project proposal and fill out the respective sections:
```
# <FILL ME: Name of project>
<FILL ME: A clear description of the project>
## 2. Language
The model will be trained in <FILL ME: which language?>.
## 3. Model
<FILL ME: 3. Which model should be used?>
## 4. Datasets
<FILL ME: 4. Which data should be used?>
Possible links to publicly available datasets include:
- <FILL ME: Link 1 to dataset>
- <FILL ME: Link 2 to dataset>
- <FILL ME: Link 3 to dataset>
## 5. Training scripts
<FILL ME: 5. Are there publicly available training scripts that can be used/tweaked for the project?>
We can make use of <FILL ME: link to training script> to train the model.>
## 6. (Optional) Challenges
<(Optionally) FILL ME: 6. What are possible challenges?>
## 7. (Optional) Desired project outcome
<(Optionally) FILL ME: 7. What is the desired project outcome? A demo?>
## 8. (Optional) Reads
The following links can be useful to better understand the project and
what has previously been done.
- <FILL ME: Link 1 to read>
- <FILL ME: Link 2 to read>
- <FILL ME: Link 3 to read>
```
To see how a proposed project looks like, please have a look at submitted project
proposals [here](https://discuss.huggingface.co/c/flax-jax-projects/22).
## Will my project proposal be selected?
Having submitted a project proposal, you can now promote your idea in the Slack channel `#flax-jax-community-week` to try to convince other participants to join your project!
Once other people have joined your project, one of the organizers (`@Suzana, @valhalla, @osanseviero, @patrickvonplaten`) will officially create a team for your project and add your project to [this google sheet](https://docs.google.com/spreadsheets/d/1GpHebL7qrwJOc9olTpIPgjf8vOS0jNb6zR_B8x_Jtik/edit?usp=sharing).
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/model_parallel/partitions.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The Google Research Authors and The HuggingFace Team 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.
"""Utilities for constructing PyTrees of PartitionSpecs."""
# utils adapted from https://github.com/google-research/google-research/blob/master/flax_models/t5x/partitions.py
import re
from flax.core.frozen_dict import freeze
from flax.traverse_util import flatten_dict, unflatten_dict
from jax.experimental import PartitionSpec as P
# Sentinels
_unmatched = object()
# For specifying empty leaf dict `{}`
empty_dict = object()
def _match(qs, ks):
"""Return True if regexes in qs match any window of strings in tuple ks."""
# compile regexes and force complete match
qts = tuple((re.compile(x + "$") for x in qs))
for i in range(len(ks) - len(qs) + 1):
matches = [x.match(y) for x, y in zip(qts, ks[i:])]
if matches and all(matches):
return True
return False
def _replacement_rules(rules):
def replace(key, val):
for rule, replacement in rules:
if _match(rule, key):
return replacement
return val
return replace
# PartitionSpec for GPTNeo
# replicate the hidden dim and shard feed-forward and head dim
def _get_partition_rules():
return [
# embeddings
(("transformer", "wpe", "embedding"), P("mp", None)),
(("transformer", "wte", "embedding"), P("mp", None)),
# atention
(("attention", "(q_proj|k_proj|v_proj)", "kernel"), P(None, "mp")),
(("attention", "out_proj", "kernel"), P("mp", None)),
(("attention", "out_proj", "bias"), None),
# mlp
(("mlp", "c_fc", "kernel"), P(None, "mp")),
(("mlp", "c_fc", "bias"), P("mp")),
(("mlp", "c_proj", "kernel"), P("mp", None)),
(("mlp", "c_proj", "bias"), None),
# layer norms
((r"ln_\d+", "bias"), None),
((r"\d+", r"ln_\d+", "scale"), None),
(("ln_f", "bias"), None),
(("ln_f", "scale"), None),
]
def set_partitions(in_dict):
rules = _get_partition_rules()
replace = _replacement_rules(rules)
initd = {k: _unmatched for k in flatten_dict(in_dict)}
result = {k: replace(k, v) for k, v in initd.items()}
assert _unmatched not in result.values(), "Incomplete partition spec."
return freeze(unflatten_dict(result))
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/model_parallel/README.md
|
<!---
Copyright 2021 The HuggingFace Team. 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.
-->
# Model parallel language model training example
The following example showcases how to train/fine-tune GPTNeo model with model parallelism using
the JAX/Flax backend and the [`pjit`](https://jax.readthedocs.io/en/latest/jax.experimental.pjit.html) transformation.
> Note: The example is experimental and might have bugs. Also currently it only supports single V3-8.
The `partition.py` file defines the `PyTree` of `ParitionSpec` for the GPTNeo model which describes how the model will be sharded.
The actual sharding is auto-matically handled by `pjit`. The weights are sharded across all local devices.
To adapt the script for other models, we need to also change the `ParitionSpec` accordingly.
TODO: Add more explantion.
Before training, let's prepare our model first. To be able to shard the model, the sharded dimention needs to be a multiple of devices it'll be sharded on. But GPTNeo's vocab size is 50257, so we need to resize the embeddings accordingly.
```python
from transformers import FlaxGPTNeoForCausalLM, GPTNeoConfig
model = FlaxGPTNeoForCausalLM.from_pretrained("EleutherAI/gpt-neo-1.3B")
emb = jnp.zeros((50264, model.config.hidden_size))
# update the first 50257 weights using pre-trained weights
emb = emb.at[:50257, :].set(model.params["transformer"]["wte"]["embedding"])
params = model.params
params["transformer"]["wte"]["embedding"] = emb
# initialize a random model with the right vocab_size
config = GPTNeoConfig.from_pretrained("EleutherAI/gpt-neo-1.3B", vocab_size=50264)
model = FlaxGPTNeoForCausalLM(config)
# assign the pre-trained weights and save the model.
model.params = params
model.save_pretrained("gpt-neo-1.3B")
```
### Train Model
```bash
python run_clm_mp.py \
--model_name_or_path gpt-neo-1.3B \
--tokenizer_name gpt2 \
--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 \
--do_train --do_eval \
--block_size 1024 \
--num_train_epochs 5 \
--learning_rate 4e-6 \
--per_device_train_batch_size 3 --per_device_eval_batch_size 3 \
--overwrite_output_dir --output_dir ~/tmp/flax-clm \
--cache_dir ~/datasets_cache/wikitext --dtype bfloat16 \
--logging_steps 96 --eval_steps 96
```
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/model_parallel/run_clm_mp.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Pre-training/Fine-tuning the GPTNeo model for causal language modeling on a text file or a dataset using model parallelism.
"""
import logging
import math
import os
import sys
import time
from dataclasses import dataclass, field
from itertools import chain
from pathlib import Path
from typing import Callable, Optional
import datasets
import jax
import jax.numpy as jnp
import numpy as np
import optax
from datasets import Dataset, load_dataset
from flax.core.frozen_dict import freeze, unfreeze
from flax.training.common_utils import onehot, stack_forest
from jax.experimental.maps import mesh
from jax.experimental.pjit import pjit
from partitions import set_partitions
from tqdm import tqdm
import transformers
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_CAUSAL_LM_MAPPING,
AutoConfig,
AutoTokenizer,
FlaxAutoModelForCausalLM,
HfArgumentParser,
TrainingArguments,
is_tensorboard_available,
)
from transformers.testing_utils import CaptureLogger
logger = logging.getLogger(__name__)
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_CAUSAL_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
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."
)
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
block_size: Optional[int] = field(
default=None,
metadata={
"help": (
"Optional input sequence length after tokenization. "
"The training dataset will be truncated in block of this size for training. "
"Default to the model max input length for single sentence inputs (take into account special tokens)."
)
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file."
def data_loader(rng: jax.random.PRNGKey, dataset: Dataset, batch_size: int, shuffle: bool = False):
"""
Returns batches of size `batch_size` from truncated `dataset`, sharded over all local devices.
Shuffle batches if `shuffle` is `True`.
"""
steps_per_epoch = len(dataset) // batch_size
if shuffle:
batch_idx = jax.random.permutation(rng, len(dataset))
else:
batch_idx = jnp.arange(len(dataset))
batch_idx = batch_idx[: steps_per_epoch * batch_size] # Skip incomplete batch.
batch_idx = batch_idx.reshape((steps_per_epoch, batch_size))
for idx in batch_idx:
batch = dataset[idx]
batch = {k: jnp.array(v) for k, v in batch.items()}
yield batch
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = stack_forest(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
def create_learning_rate_fn(
train_ds_size: int, train_batch_size: int, num_train_epochs: int, num_warmup_steps: int, learning_rate: float
) -> Callable[[int], jnp.ndarray]:
"""Returns a linear warmup, linear_decay learning rate function."""
steps_per_epoch = train_ds_size // train_batch_size
num_train_steps = steps_per_epoch * num_train_epochs
warmup_fn = optax.linear_schedule(init_value=0.0, end_value=learning_rate, transition_steps=num_warmup_steps)
decay_fn = optax.linear_schedule(
init_value=learning_rate, end_value=0, transition_steps=num_train_steps - num_warmup_steps
)
schedule_fn = optax.join_schedules(schedules=[warmup_fn, decay_fn], boundaries=[num_warmup_steps])
return schedule_fn
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."
)
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
# Setup logging, we only want one process per machine to log things on the screen.
logger.setLevel(logging.INFO if jax.process_index() == 0 else logging.ERROR)
if jax.process_index() == 0:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
# 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 if no column called
# 'text' is found. You can easily tweak this behavior (see below).
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
dataset = load_dataset(
data_args.dataset_name, data_args.dataset_config_name, cache_dir=model_args.cache_dir, keep_in_memory=False
)
if "validation" not in dataset.keys():
dataset["validation"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
)
dataset["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"train[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
)
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"
dataset = load_dataset(extension, data_files=data_files, cache_dir=model_args.cache_dir)
# 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.
# Load pretrained config and tokenizer
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.")
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
if training_args.do_train:
column_names = dataset["train"].column_names
else:
column_names = dataset["validation"].column_names
text_column_name = "text" if "text" in column_names else column_names[0]
# since this will be pickled to avoid _LazyModule error in Hasher force logger loading before tokenize_function
tok_logger = transformers.utils.logging.get_logger("transformers.tokenization_utils_base")
def tokenize_function(examples):
with CaptureLogger(tok_logger) as cl:
output = tokenizer(examples[text_column_name])
# clm input could be much much longer than block_size
if "Token indices sequence length is longer than the" in cl.out:
tok_logger.warning(
"^^^^^^^^^^^^^^^^ Please ignore the warning above - this long input will be chunked into smaller bits"
" before being passed to the model."
)
return output
tokenized_datasets = dataset.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.block_size is None:
block_size = tokenizer.model_max_length
if block_size > config.max_position_embeddings:
logger.warning(
f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
f"Using block_size={min(1024, config.max_position_embeddings)} instead. You can change that default value by passing --block_size xxx."
)
block_size = min(1024, config.max_position_embeddings)
else:
if data_args.block_size > tokenizer.model_max_length:
logger.warning(
f"The block_size passed ({data_args.block_size}) is larger than the maximum length for the model "
f"({tokenizer.model_max_length}). Using block_size={tokenizer.model_max_length}."
)
block_size = min(data_args.block_size, tokenizer.model_max_length)
# Main data processing function that will concatenate all texts from our dataset and generate chunks of block_size.
def group_texts(examples):
# Concatenate all texts.
concatenated_examples = {k: list(chain(*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.
if total_length >= block_size:
total_length = (total_length // block_size) * block_size
# Split by chunks of max_len.
result = {
k: [t[i : i + block_size] for i in range(0, total_length, block_size)]
for k, t in concatenated_examples.items()
}
result["labels"] = result["input_ids"].copy()
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/process#map
lm_datasets = tokenized_datasets.map(
group_texts,
batched=True,
num_proc=data_args.preprocessing_num_workers,
load_from_cache_file=not data_args.overwrite_cache,
)
if training_args.do_train:
if "train" not in tokenized_datasets:
raise ValueError("--do_train requires a train dataset")
train_dataset = lm_datasets["train"]
if data_args.max_train_samples is not None:
max_train_samples = min(len(train_dataset), data_args.max_train_samples)
train_dataset = train_dataset.select(range(max_train_samples))
if training_args.do_eval:
if "validation" not in tokenized_datasets:
raise ValueError("--do_eval requires a validation dataset")
eval_dataset = lm_datasets["validation"]
if data_args.max_eval_samples is not None:
max_eval_samples = min(len(eval_dataset), data_args.max_eval_samples)
eval_dataset = eval_dataset.select(range(max_eval_samples))
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
rng, dropout_rng = jax.random.split(rng)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
eval_batch_size = int(training_args.per_device_eval_batch_size) * jax.device_count()
steps_per_epoch = len(train_dataset) // train_batch_size
total_train_steps = steps_per_epoch * num_epochs
# TODO: weights should be initialized in pjitted fun, this won't work for REALLY large models
# TODO: when loading from pre-trained model we need to make sure the vocab is divisible by num_partitions
# GPT2's vocab is odd, we need to resize it for fine-tuning
model = FlaxAutoModelForCausalLM.from_pretrained(
model_args.model_name_or_path, seed=training_args.seed, dtype=getattr(jnp, model_args.dtype)
)
# Create learning rate schedule
linear_decay_lr_schedule_fn = create_learning_rate_fn(
len(train_dataset),
train_batch_size,
training_args.num_train_epochs,
training_args.warmup_steps,
training_args.learning_rate,
)
optimizer = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
)
def get_initial_state(params):
state = optimizer.init(params)
return tuple(state), params
# Get PartitionSpec for model params
param_spec = set_partitions(unfreeze(model.params))
# Get the PyTree for opt_state, we don't actually initialize the opt_state yet.
params_shapes = jax.tree_util.tree_map(lambda x: x.shape, model.params)
state_shapes = jax.eval_shape(get_initial_state, params_shapes)
# get PartitionSpec for opt_state, this is very specific to adamw
# TODO: optax returns different state for different optimizers, how can we handle this generically ?
# or maybe we don't since in our examples we just use adamw or adafactor
def get_opt_spec(x):
if isinstance(x, dict):
return param_spec
return None
opt_state_spec, param_spec = jax.tree_util.tree_map(
get_opt_spec, state_shapes, is_leaf=lambda x: isinstance(x, (dict, optax.EmptyState))
)
# pjit the get_initial_state function to shard params and init
# optimizer state in sharded way
p_get_initial_state = pjit(
get_initial_state,
in_axis_resources=None,
out_axis_resources=(opt_state_spec, param_spec),
)
# hack: move the inital params to CPU to free up device memory
# TODO: allow loading weights on CPU in pre-trained model
model.params = jax.tree_util.tree_map(lambda x: np.asarray(x), model.params)
# mesh defination
mesh_devices = np.array(jax.devices()).reshape(1, jax.local_device_count())
# actually initialize the opt_state
with mesh(mesh_devices, ("dp", "mp")):
opt_state, params = p_get_initial_state(freeze(model.params))
# cross-entropy with z loss
def loss_fn(logits, labels, z_loss=0):
shift_logits = logits[..., :-1, :]
shift_labels = labels[..., 1:]
shift_labels = onehot(shift_labels, shift_logits.shape[-1])
shift_logits = shift_logits - jax.lax.stop_gradient(shift_logits.max(axis=-1, keepdims=True))
log_z = jnp.log(jnp.sum(jnp.exp(shift_logits), axis=-1, keepdims=True))
log_softmax = shift_logits - log_z
loss = -jnp.sum(shift_labels * log_softmax, axis=-1)
loss += (1e-4 * jnp.square(log_z.squeeze(-1))) * z_loss
return loss.mean()
# Define gradient update step fn
# TODO: try to use TrainState instead of passing params and opt_state individually
def train_step(params, opt_state, dropout_rng, batch, step):
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
def compute_loss(params):
labels = batch.pop("labels")
logits = model(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
loss = loss_fn(logits, labels, z_loss=1.0)
return loss
grad_fn = jax.value_and_grad(compute_loss)
loss, grads = grad_fn(params)
updates, new_opt_state = optimizer.update(grads, opt_state, params)
new_params = optax.apply_updates(params, updates)
metrics = {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(step)}
return new_params, tuple(new_opt_state), new_dropout_rng, metrics, step + 1
# Define eval fn
def eval_step(input_ids, labels, params):
logits = model(input_ids=input_ids, params=params, train=False)[0]
loss = loss_fn(logits, labels)
# metrics
return {"loss": loss}
p_train_step = pjit(
train_step,
in_axis_resources=(param_spec, opt_state_spec, None, None, None),
out_axis_resources=(param_spec, opt_state_spec, None, None, None),
donate_argnums=(0, 1),
)
p_eval_step = pjit(
eval_step,
in_axis_resources=(None, None, param_spec),
out_axis_resources=None,
)
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {num_epochs}")
logger.info(f" Instantaneous batch size per device = {training_args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel & distributed) = {train_batch_size}")
logger.info(f" Total optimization steps = {total_train_steps}")
train_time = 0
train_metrics = []
epochs = tqdm(range(num_epochs), desc=f"Epoch ... (1/{num_epochs})", position=0)
global_step = 0
# we are not doing 2D parallelism (yet!), this just does model parallelism
with mesh(mesh_devices, ("dp", "mp")):
for _ in epochs:
# ======================== Training ================================
train_start = time.time()
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# Generate an epoch by shuffling sampling indices from the train dataset
train_metrics = []
train_loader = data_loader(input_rng, train_dataset, train_batch_size, shuffle=True)
steps_per_epoch = len(train_dataset) // train_batch_size
# train
for _ in tqdm(range(steps_per_epoch), desc="Training...", position=1, leave=False):
batch = next(train_loader)
params, opt_state, dropout_rng, train_metric, global_step = p_train_step(
params,
opt_state,
dropout_rng,
batch,
global_step,
)
train_metrics.append(train_metric)
cur_step = global_step
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step} | Loss: {train_metric['loss']}, Learning Rate:"
f" {train_metric['learning_rate']})"
)
train_metrics = []
if cur_step % training_args.eval_steps == 0 and cur_step > 0:
# ======================== Evaluating ==============================
eval_metrics = []
eval_loader = data_loader(input_rng, eval_dataset, eval_batch_size)
eval_steps = len(eval_dataset) // eval_batch_size
for _ in tqdm(range(eval_steps), desc="Evaluating...", position=2, leave=False):
batch = next(eval_loader)
metrics = p_eval_step(batch["input_ids"], batch["labels"], params)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = stack_forest(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.mean, eval_metrics)
try:
eval_metrics["perplexity"] = math.exp(eval_metrics["loss"])
except OverflowError:
eval_metrics["perplexity"] = float("inf")
logger.info(
f"Step... ({cur_step} | Eval loss: {eval_metrics['loss']} | Eval Perplexity:"
f" {eval_metrics['perplexity']}"
)
if cur_step % training_args.save_steps == 0 and cur_step > 0:
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(params)
model.save_pretrained(
training_args.output_dir,
params=params,
push_to_hub=training_args.push_to_hub,
commit_message=f"Saving weights and logs of step {cur_step}",
)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/hybrid_clip/configuration_hybrid_clip.py
|
import copy
from transformers.configuration_utils import PretrainedConfig
from transformers.utils import logging
logger = logging.get_logger(__name__)
class HybridCLIPConfig(PretrainedConfig):
r"""
:class:`HybridCLIPConfig` is the configuration class to store the configuration of a
:class:`~HybridCLIPModel`. It is used to instantiate HybridCLIPModel model according to the specified arguments,
defining the text model and vision model configs.
Configuration objects inherit from :class:`~transformers.PretrainedConfig` and can be used to control the model
outputs. Read the documentation from :class:`~transformers.PretrainedConfig` for more information.
Args:
text_config_dict (:obj:`dict`):
Dictionary of configuration options that defines text model config.
vision_config_dict (:obj:`dict`):
Dictionary of configuration options that defines vison model config.
projection_dim (:obj:`int`, `optional`, defaults to 512):
Dimentionality of text and vision projection layers.
kwargs (`optional`):
Dictionary of keyword arguments.
Examples::
>>> from transformers import BertConfig, CLIPConfig, HybridCLIPConfig, FlaxHybridCLIP
>>> # Initializing a BERT and CLIP configuration
>>> config_text = BertConfig()
>>> config_vision = CLIPConfig()
>>> config = HybridCLIPConfig.from_text_vision_configs(config_text, config_vision, projection_dim=512)
>>> # Initializing a BERT and CLIPVision model
>>> model = EncoderDecoderModel(config=config)
>>> # Accessing the model configuration
>>> config_text = model.config.text_config
>>> config_vision = model.config.vision_config
>>> # Saving the model, including its configuration
>>> model.save_pretrained('my-model')
>>> # loading model and config from pretrained folder
>>> encoder_decoder_config = HybridCLIPConfig.from_pretrained('my-model')
>>> model = FlaxHybridCLIP.from_pretrained('my-model', config=encoder_decoder_config)
"""
model_type = "hybrid-clip"
is_composition = True
def __init__(self, projection_dim=512, **kwargs):
super().__init__(**kwargs)
if "text_config" not in kwargs:
raise ValueError("`text_config` can not be `None`.")
if "vision_config" not in kwargs:
raise ValueError("`vision_config` can not be `None`.")
text_config = kwargs.pop("text_config")
vision_config = kwargs.pop("vision_config")
text_model_type = text_config.pop("model_type")
vision_model_type = vision_config.pop("model_type")
from transformers import AutoConfig
self.text_config = AutoConfig.for_model(text_model_type, **text_config)
if vision_model_type == "clip":
self.vision_config = AutoConfig.for_model(vision_model_type, **vision_config).vision_config
elif vision_model_type == "clip_vision_model":
from transformers import CLIPVisionConfig
self.vision_config = CLIPVisionConfig(**vision_config)
else:
self.vision_config = AutoConfig.for_model(vision_model_type, **vision_config)
self.projection_dim = projection_dim
self.initializer_factor = 1.0
@classmethod
def from_text_vision_configs(cls, text_config: PretrainedConfig, vision_config: PretrainedConfig, **kwargs):
r"""
Instantiate a :class:`HybridCLIPConfig` (or a derived class) from text model configuration and
vision model configuration.
Returns:
:class:`HybridCLIPConfig`: An instance of a configuration object
"""
return cls(text_config=text_config.to_dict(), vision_config=vision_config.to_dict(), **kwargs)
def to_dict(self):
"""
Serializes this instance to a Python dictionary. Override the default
:meth:`~transformers.PretrainedConfig.to_dict`.
Returns:
:obj:`Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
"""
output = copy.deepcopy(self.__dict__)
output["text_config"] = self.text_config.to_dict()
output["vision_config"] = self.vision_config.to_dict()
output["model_type"] = self.__class__.model_type
return output
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/hybrid_clip/README.md
|
<!---
Copyright 2021 The HuggingFace Team. 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.
-->
# Vision-Text dual encoder model training examples
> Note: This example is experimental and might not give the best possible results
The following example showcases how to train a CLIP like vision-text dual encoder model
using a pre-trained vision and text encoder using the JAX/Flax backend.
Such a model can be used for natural language image search and potentially zero-shot image classification.
The model is inspired by the [CLIP](https://openai.com/blog/clip/) approach, introduced by Alec Radford et al.
The idea is to train a vision encoder and a text encoder jointly to project the representation of images and their
captions into the same embedding space, such that the caption embeddings are located near the embeddings
of the images they describe.
JAX/Flax allows you to trace pure functions and compile them into efficient, fused accelerator code on both GPU and TPU.
Models written in JAX/Flax are **immutable** and updated in a purely functional
way which enables simple and efficient model parallelism.
In this example we will use the vision model from [CLIP](https://huggingface.co/models?filter=clip)
as the image encoder and [`roberta-base`](https://huggingface.co/roberta-base) as the text encoder.
Note that one can also use the [ViT](https://huggingface.co/models?filter=vit) model as image encoder and any other BERT or ROBERTa model as text encoder.
To train the model on languages other than English one should choose a text encoder trained on the desired
language and a image-text dataset in that language. One such dataset is [WIT](https://github.com/google-research-datasets/wit).
Let's start by creating a model repository to save the trained model and logs.
Here we call the model `"clip-roberta-base"`, but you can change the model name as you like.
You can do this either directly on [huggingface.co](https://huggingface.co/new) (assuming that
you are logged in) or via the command line:
```
huggingface-cli repo create clip-roberta-base
```
Next we clone the model repository to add the tokenizer and model files.
```
git clone https://huggingface.co/<your-username>/clip-roberta-base
```
To ensure that all tensorboard traces will be uploaded correctly, we need to
track them. You can run the following command inside your model repo to do so.
```
cd clip-roberta-base
git lfs track "*tfevents*"
```
Great, we have set up our model repository. During training, we will automatically
push the training logs and model weights to the repo.
Next, let's add a symbolic link to the `run_hybrid_clip.py`.
```bash
export MODEL_DIR="./clip-roberta-base
ln -s ~/transformers/examples/research_projects/jax-projects/hybrid_clip/run_hybrid_clip.py run_hybrid_clip.py
```
## How to use the `FlaxHybridCLIP` model:
The `FlaxHybridCLIP` class let's you load any text and vision encoder model to create a dual encoder.
Here is an example of how to load the model using pre-trained text and vision models.
```python
from modeling_hybrid_clip import FlaxHybridCLIP
model = FlaxHybridCLIP.from_text_vision_pretrained("bert-base-uncased", "openai/clip-vit-base-patch32")
# save the model
model.save_pretrained("bert-clip")
# load the saved model
model = FlaxHybridCLIP.from_pretrained("bert-clip")
```
If the checkpoints are in PyTorch then one could pass `text_from_pt=True` and `vision_from_pt=True`. This will load the model
PyTorch checkpoints convert them to flax and load the model.
```python
model = FlaxHybridCLIP.from_text_vision_pretrained("bert-base-uncased", "openai/clip-vit-base-patch32", text_from_pt=True, vision_from_pt=True)
```
This loads both the text and vision encoders using pre-trained weights, the projection layers are randomly
initialized except for CLIP's vision model. If you use CLIP to initialize the vision model then the vision projection weights are also
loaded using the pre-trained weights.
## Prepare the dataset
We will use the MS-COCO dataset to train our dual encoder model. MS-COCO contains over 82,000 images, each of which has at least 5 different caption annotations. The dataset is usually used for image captioning tasks, but we can repurpose the image-caption pairs to train our dual encoder model for image search.
### Download and extract the data.
It consists of two compressed folders: one with images, and the other—with associated image captions. Note that the compressed images folder is 13GB in size.
```bash
wget http://images.cocodataset.org/annotations/annotations_trainval2014.zip
wget http://images.cocodataset.org/zips/train2014.zip
unzip annotations_trainval2014.zip
unzip train2014.zip
mkdir coco_dataset
mv train2014 coco_dataset/
mv annotations coco_dataset/
```
### Prepare dataset files and split the dataset.
```python
import json
import collections
images_dir = "coco_dataset/train2014"
annotation_file = "coco_dataset/annotations/captions_train2014.json"
with open(annotation_file, "r") as f:
annotations = json.load(f)["annotations"]
image_path_to_caption = collections.defaultdict(list)
for element in annotations:
caption = f"{element['caption'].lower().rstrip('.')}"
image_path = images_dir + "/COCO_train2014_" + "%012d.jpg" % (element["image_id"])
image_path_to_caption[image_path].append(caption)
lines = []
for image_path, captions in image_path_to_caption.items():
lines.append(json.dumps({"image_path": image_path, "captions": captions}))
train_lines = lines[:-8000]
valid_line = lines[-8000:]
with open("coco_dataset/train_dataset.json", "w") as f:
f.write("\n".join(train_lines))
with open("coco_dataset/valid_dataset.json", "w") as f:
f.write("\n".join(valid_line))
```
> Note: The data loading and processing part of this script can still be improved for maximum performance. In particular one should decode the images beforehand and use those instead decoding them each time. If the dataset is small or if you have huge disk space the you could also pre-process all the dataset beforehand and then use it.
## Train the model
Next we can run the example script to train the model:
```bash
python run_hybrid_clip.py \
--output_dir ${MODEL_DIR} \
--text_model_name_or_path="roberta-base" \
--vision_model_name_or_path="openai/clip-vit-base-patch32" \
--tokenizer_name="roberta-base" \
--train_file="coco_dataset/train_dataset.json" \
--validation_file="coco_dataset/validation_dataset.json" \
--do_train --do_eval \
--num_train_epochs="40" --max_seq_length 96 \
--per_device_train_batch_size="64" \
--per_device_eval_batch_size="64" \
--learning_rate="5e-5" --warmup_steps="0" --weight_decay 0.1 \
--overwrite_output_dir \
--preprocessing_num_workers 32 \
--push_to_hub
```
This should finish in ~1h50 mins with min validation loss 2.43. Training statistics can be accessed on [tfhub.de](https://tensorboard.dev/experiment/RUNPYd1yRgSD5kZSb9hDig/#scalars)
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/hybrid_clip/run_hybrid_clip.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Training a CLIP like dual encoder models using text and vision encoders in the library.
The script can be used to train CLIP like models for languages other than english by using
a text encoder pre-trained in the desired language. Currently this script support the following vision
and text models:
Vision models: ViT(https://huggingface.co/models?filter=vit), CLIP (https://huggingface.co/models?filter=clip)
Text models: BERT, ROBERTa (https://huggingface.co/models?filter=fill-mask)
"""
import json
import logging
import os
import sys
import time
from dataclasses import dataclass, field
from pathlib import Path
from typing import Callable, Optional
import jax
import jax.numpy as jnp
import optax
import torch
from flax import jax_utils
from flax.jax_utils import unreplicate
from flax.training import train_state
from flax.training.common_utils import get_metrics, shard, shard_prng_key
from modeling_hybrid_clip import FlaxHybridCLIP
from torchvision.datasets import VisionDataset
from torchvision.io import ImageReadMode, read_image
from torchvision.transforms import CenterCrop, ConvertImageDtype, Normalize, Resize
from torchvision.transforms.functional import InterpolationMode
from tqdm import tqdm
import transformers
from transformers import AutoTokenizer, HfArgumentParser, TrainingArguments, is_tensorboard_available, set_seed
logger = logging.getLogger(__name__)
# Cache the result
has_tensorboard = is_tensorboard_available()
if has_tensorboard:
try:
from flax.metrics.tensorboard import SummaryWriter
except ImportError as ie:
has_tensorboard = False
print(f"Unable to display metrics through TensorBoard because some package are not installed: {ie}")
else:
print(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
text_model_name_or_path: str = field(
metadata={
"help": (
"The text model checkpoint for weights initialization. "
"Don't set if you want to train a model from scratch."
)
},
)
vision_model_name_or_path: str = field(
metadata={
"help": (
"The vision model checkpoint for weights initialization. "
"Don't set if you want to train a model from scratch."
)
},
)
from_pt: bool = field(
default=True,
metadata={"help": "whether to load the text and vision model using PyTorch checkpoints."},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
data_dir: Optional[str] = field(default=None, metadata={"help": "The data directory containing input files."})
train_file: Optional[str] = field(
default=None, metadata={"help": "The input training data file (a jsonlines file)."}
)
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file (a jsonlines file)."},
)
max_seq_length: Optional[int] = field(
default=72,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
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."
)
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
def __post_init__(self):
if self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension == "json", "`train_file` should be a json file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension == "json", "`validation_file` should be a json file."
# We use torchvision for faster image pre-processing.
# We need to ensure faster processing speed as it can become a bottleneck on TPU
class Transform(torch.nn.Module):
def __init__(self, image_size):
super().__init__()
self.transforms = torch.nn.Sequential(
Resize([image_size], interpolation=InterpolationMode.BICUBIC),
CenterCrop(image_size),
ConvertImageDtype(torch.float),
Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)),
)
def forward(self, x: torch.Tensor) -> torch.Tensor:
with torch.no_grad():
x = self.transforms(x)
return x
class ImageTextDataset(VisionDataset):
"""
Dtaset for loading image-text data for tasks like CLIP training, Image Captioning.
Args:
root: (string): The root path where the dataset is stored
file_path: (string): Path to the file containing the image_paths and associated captions.
The expected format is jsonlines where each line is a json object containing to keys.
`image_path`: The path to the image.
`captions`: An `array` of captions.
transform (callable, optional): A function/transform that takes in an PIL image
and returns a transformed version. E.g, ``transforms.ToTensor``
target_transform (callable, optional): A function/transform that takes in the
target and transforms it.
transforms (callable, optional): A function/transform that takes input sample and its target as entry
and returns a transformed version.
"""
def __init__(
self,
root: str,
file_path: str,
captions_per_image=2,
transform: Optional[Callable] = None,
target_transform: Optional[Callable] = None,
transforms: Optional[Callable] = None,
):
super().__init__(root, transforms, transform, target_transform)
with open(file_path, "r") as f:
examples = [json.loads(line) for line in f.readlines()]
self.captions = []
self.image_paths = []
for example in examples:
captions_subset = example["captions"][:captions_per_image]
self.captions.extend(captions_subset)
self.image_paths.extend([example["image_path"]] * len(captions_subset))
def _load_image(self, idx: int):
path = self.image_paths[idx]
return read_image(path, mode=ImageReadMode.RGB)
def _load_target(self, idx):
return self.captions[idx]
def __getitem__(self, index: int):
image = self._load_image(index)
target = self._load_target(index)
if self.transforms is not None:
image, target = self.transforms(image, target)
return image, target
def __len__(self) -> int:
return len(self.captions)
class TrainState(train_state.TrainState):
dropout_rng: jnp.ndarray
def replicate(self):
return jax_utils.replicate(self).replace(dropout_rng=shard_prng_key(self.dropout_rng))
def write_metric(summary_writer, train_metrics, eval_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
def create_learning_rate_fn(
train_ds_size: int, train_batch_size: int, num_train_epochs: int, num_warmup_steps: int, learning_rate: float
) -> Callable[[int], jnp.ndarray]:
"""Returns a linear warmup, linear_decay learning rate function."""
steps_per_epoch = train_ds_size // train_batch_size
num_train_steps = steps_per_epoch * num_train_epochs
warmup_fn = optax.linear_schedule(init_value=0.0, end_value=learning_rate, transition_steps=num_warmup_steps)
decay_fn = optax.linear_schedule(
init_value=learning_rate, end_value=0, transition_steps=num_train_steps - num_warmup_steps
)
schedule_fn = optax.join_schedules(schedules=[warmup_fn, decay_fn], boundaries=[num_warmup_steps])
return schedule_fn
def main():
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."
)
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
# Setup logging, we only want one process per machine to log things on the screen.
logger.setLevel(logging.INFO if jax.process_index() == 0 else logging.ERROR)
if jax.process_index() == 0:
transformers.utils.logging.set_verbosity_info()
else:
transformers.utils.logging.set_verbosity_error()
# Set the verbosity to info of the Transformers logger (on main process only):
logger.info(f"Training/evaluation parameters {training_args}")
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer
)
elif model_args.text_model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.text_model_name_or_path, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
model = FlaxHybridCLIP.from_text_vision_pretrained(
model_args.text_model_name_or_path,
model_args.vision_model_name_or_path,
seed=training_args.seed,
dtype=getattr(jnp, model_args.dtype),
text_from_pt=model_args.from_pt,
vision_from_pt=model_args.from_pt,
)
config = model.config
# set seed for torch dataloaders
set_seed(training_args.seed)
# Initialize torchvision transforms and jit them for faster processing
preprocess = Transform(config.vision_config.image_size)
preprocess = torch.jit.script(preprocess)
# Initialize the image-text dataset
train_dataset = ImageTextDataset(
data_args.data_dir,
data_args.train_file,
captions_per_image=2,
transform=preprocess,
)
eval_dataset = ImageTextDataset(
data_args.data_dir,
data_args.validation_file,
captions_per_image=1,
transform=preprocess,
)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
eval_batch_size = int(training_args.per_device_eval_batch_size) * jax.device_count()
steps_per_epoch = len(train_dataset) // train_batch_size
total_train_steps = steps_per_epoch * num_epochs
# Use collate function to tokenizer the text and convert the processed images to numpy
def collate_fn(examples):
pixel_values = torch.stack([example[0] for example in examples]).permute(0, 2, 3, 1).numpy()
captions = [example[1] for example in examples]
inputs = tokenizer(
captions, max_length=data_args.max_seq_length, padding="max_length", truncation=True, return_tensors="np"
)
batch = {
"pixel_values": pixel_values,
"input_ids": inputs["input_ids"],
"attention_mask": inputs["attention_mask"],
}
return batch
# Create data loaders
train_loader = torch.utils.data.DataLoader(
train_dataset,
batch_size=train_batch_size,
shuffle=True,
num_workers=data_args.preprocessing_num_workers,
persistent_workers=True,
drop_last=True,
collate_fn=collate_fn,
)
eval_loader = torch.utils.data.DataLoader(
eval_dataset,
batch_size=eval_batch_size,
shuffle=False,
num_workers=data_args.preprocessing_num_workers,
persistent_workers=True,
drop_last=True,
collate_fn=collate_fn,
)
# Enable tensorboard only on the master node
if has_tensorboard and jax.process_index() == 0:
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir).joinpath("logs").as_posix())
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
rng, dropout_rng = jax.random.split(rng)
# Create learning rate schedule
linear_decay_lr_schedule_fn = create_learning_rate_fn(
len(train_dataset),
train_batch_size,
training_args.num_train_epochs,
training_args.warmup_steps,
training_args.learning_rate,
)
# create adam optimizer
adamw = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
)
# Setup train state
state = TrainState.create(apply_fn=model.__call__, params=model.params, tx=adamw, dropout_rng=dropout_rng)
def cross_entropy(logits, axis):
logprobs = jax.nn.log_softmax(logits, axis=axis)
nll = jnp.diag(logprobs)
ce = -jnp.mean(nll)
return ce
def clip_loss(similarity):
loss = (cross_entropy(similarity, axis=0) + cross_entropy(similarity, axis=1)) / 2
return loss
# Define gradient update step fn
def train_step(state, batch):
dropout_rng, new_dropout_rng = jax.random.split(state.dropout_rng)
def compute_loss(params):
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
loss = clip_loss(logits)
return loss
grad_fn = jax.value_and_grad(compute_loss)
loss, grad = grad_fn(state.params)
grad = jax.lax.pmean(grad, "batch")
new_state = state.apply_gradients(grads=grad, dropout_rng=new_dropout_rng)
metrics = {"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}
metrics = jax.lax.pmean(metrics, axis_name="batch")
return new_state, metrics
# Define eval fn
def eval_step(params, batch):
logits = model(**batch, params=params, train=False)[0]
loss = clip_loss(logits)
# summarize metrics
metrics = {"loss": loss}
metrics = jax.lax.pmean(metrics, axis_name="batch")
return metrics
# Create parallel version of the train and eval step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
p_eval_step = jax.pmap(eval_step, "batch")
# Replicate the train state on each device
state = state.replicate()
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {num_epochs}")
logger.info(f" Instantaneous batch size per device = {training_args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel & distributed) = {train_batch_size}")
logger.info(f" Total optimization steps = {total_train_steps}")
train_time = 0
# Create sampling rng
rng, input_rng = jax.random.split(rng)
epochs = tqdm(range(num_epochs), desc=f"Epoch ... (1/{num_epochs})", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
# Create sampling rng
rng, input_rng = jax.random.split(rng)
train_metrics = []
steps_per_epoch = len(train_dataset) // train_batch_size
train_step_progress_bar = tqdm(total=steps_per_epoch, desc="Training...", position=1, leave=False)
# train
for batch in train_loader:
batch = shard(batch)
state, train_metric = p_train_step(state, batch)
train_metrics.append(train_metric)
train_step_progress_bar.update(1)
train_time += time.time() - train_start
train_metric = unreplicate(train_metric)
train_step_progress_bar.close()
epochs.write(
f"Epoch... ({epoch + 1}/{num_epochs} | Loss: {train_metric['loss']}, Learning Rate:"
f" {train_metric['learning_rate']})"
)
# ======================== Evaluating ==============================
eval_metrics = []
eval_steps = len(eval_dataset) // eval_batch_size
eval_step_progress_bar = tqdm(total=eval_steps, desc="Evaluating...", position=2, leave=False)
for batch in eval_loader:
# Model forward
batch = shard(batch)
metrics = p_eval_step(state.params, batch)
eval_metrics.append(metrics)
eval_step_progress_bar.update(1)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.mean, eval_metrics)
# Print metrics and update progress bar
eval_step_progress_bar.close()
desc = f"Epoch... ({epoch + 1}/{num_epochs} | Eval Loss: {eval_metrics['loss']})"
epochs.write(desc)
epochs.desc = desc
# Save metrics
if has_tensorboard and jax.process_index() == 0:
cur_step = epoch * (len(train_dataset) // train_batch_size)
write_metric(summary_writer, train_metrics, eval_metrics, train_time, cur_step)
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(unreplicate(state.params))
model.save_pretrained(
training_args.output_dir,
params=params,
push_to_hub=training_args.push_to_hub,
commit_message=f"Saving weights and logs of epoch {epoch+1}",
)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/hybrid_clip/modeling_hybrid_clip.py
|
# coding=utf-8
# Copyright 2021 The HuggingFace Team. 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.
from typing import Optional, Tuple
import flax.linen as nn
import jax
import jax.numpy as jnp
from configuration_hybrid_clip import HybridCLIPConfig
from flax.core.frozen_dict import FrozenDict
from transformers import FLAX_MODEL_MAPPING, FlaxCLIPVisionModel
from transformers.modeling_flax_utils import FlaxPreTrainedModel
from transformers.models.clip.modeling_flax_clip import FlaxCLIPOutput
from transformers.utils import logging
logger = logging.get_logger(__name__)
class FlaxHybridCLIPModule(nn.Module):
config: HybridCLIPConfig
dtype: jnp.dtype = jnp.float32
def setup(self):
text_config = self.config.text_config
vision_config = self.config.vision_config
self.projection_dim = self.config.projection_dim
self.text_embed_dim = text_config.hidden_size
self.vision_embed_dim = vision_config.hidden_size
text_module = FLAX_MODEL_MAPPING[self.config.text_config.__class__].module_class
vision_module = FLAX_MODEL_MAPPING.get(self.config.vision_config.__class__, FlaxCLIPVisionModel).module_class
self.text_model = text_module(text_config, dtype=self.dtype)
self.vision_model = vision_module(vision_config, dtype=self.dtype)
self.visual_projection = nn.Dense(
self.projection_dim,
dtype=self.dtype,
kernel_init=jax.nn.initializers.normal(0.02),
use_bias=False,
)
self.text_projection = nn.Dense(
self.projection_dim,
dtype=self.dtype,
kernel_init=jax.nn.initializers.normal(0.02),
use_bias=False,
)
self.logit_scale = self.param("logit_scale", jax.nn.initializers.ones, [])
def __call__(
self,
input_ids=None,
pixel_values=None,
attention_mask=None,
position_ids=None,
token_type_ids=None,
deterministic: bool = True,
output_attentions=None,
output_hidden_states=None,
return_dict=None,
):
return_dict = return_dict if return_dict is not None else self.config.return_dict
vision_outputs = self.vision_model(
pixel_values=pixel_values,
deterministic=deterministic,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
text_outputs = self.text_model(
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
deterministic=deterministic,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
image_embeds = vision_outputs[1]
image_embeds = self.visual_projection(image_embeds)
text_embeds = text_outputs[1]
text_embeds = self.text_projection(text_embeds)
# normalized features
image_embeds = image_embeds / jnp.linalg.norm(image_embeds, axis=-1, keepdims=True)
text_embeds = text_embeds / jnp.linalg.norm(text_embeds, axis=-1, keepdims=True)
# cosine similarity as logits
logit_scale = jnp.exp(self.logit_scale)
logits_per_text = jnp.matmul(text_embeds, image_embeds.T) * logit_scale
logits_per_image = logits_per_text.T
if not return_dict:
return (logits_per_image, logits_per_text, text_embeds, image_embeds, text_outputs, vision_outputs)
return FlaxCLIPOutput(
logits_per_image=logits_per_image,
logits_per_text=logits_per_text,
text_embeds=text_embeds,
image_embeds=image_embeds,
text_model_output=text_outputs,
vision_model_output=vision_outputs,
)
class FlaxHybridCLIP(FlaxPreTrainedModel):
config_class = HybridCLIPConfig
module_class = FlaxHybridCLIPModule
def __init__(
self,
config: HybridCLIPConfig,
input_shape: Optional[Tuple] = None,
seed: int = 0,
dtype: jnp.dtype = jnp.float32,
**kwargs,
):
if input_shape is None:
input_shape = ((1, 1), (1, config.vision_config.image_size, config.vision_config.image_size, 3))
module = self.module_class(config=config, dtype=dtype, **kwargs)
super().__init__(config, module, input_shape=input_shape, seed=seed, dtype=dtype)
def init_weights(self, rng: jax.random.PRNGKey, input_shape: Tuple, params: FrozenDict = None) -> FrozenDict:
# init input tensor
input_ids = jnp.zeros(input_shape[0], dtype="i4")
position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_shape[0])
token_type_ids = jnp.ones_like(input_ids)
attention_mask = jnp.ones_like(input_ids)
pixel_values = jax.random.normal(rng, input_shape[1])
params_rng, dropout_rng = jax.random.split(rng)
rngs = {"params": params_rng, "dropout": dropout_rng}
return self.module.init(rngs, input_ids, pixel_values, attention_mask, position_ids, token_type_ids)["params"]
def __call__(
self,
input_ids,
pixel_values,
attention_mask=None,
position_ids=None,
token_type_ids=None,
params: dict = None,
dropout_rng: jax.random.PRNGKey = None,
train: bool = False,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
):
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
output_hidden_states = (
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
)
return_dict = return_dict if return_dict is not None else self.config.return_dict
if position_ids is None:
position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_ids.shape)
if token_type_ids is None:
token_type_ids = jnp.zeros_like(input_ids)
if attention_mask is None:
attention_mask = jnp.ones_like(input_ids)
# Handle any PRNG if needed
rngs = {}
if dropout_rng is not None:
rngs["dropout"] = dropout_rng
return self.module.apply(
{"params": params or self.params},
jnp.array(input_ids, dtype="i4"),
jnp.array(pixel_values, dtype=jnp.float32),
jnp.array(attention_mask, dtype="i4"),
jnp.array(position_ids, dtype="i4"),
jnp.array(token_type_ids, dtype="i4"),
not train,
output_attentions,
output_hidden_states,
return_dict,
rngs=rngs,
)
def get_text_features(
self,
input_ids,
attention_mask=None,
position_ids=None,
token_type_ids=None,
params: dict = None,
dropout_rng: jax.random.PRNGKey = None,
train=False,
):
r"""
Args:
input_ids (:obj:`numpy.ndarray` of shape :obj:`(batch_size, sequence_length)`):
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you
provide it.
Indices can be obtained using :class:`~transformers.PreTrainedTokenizer`. See
:meth:`transformers.PreTrainedTokenizer.encode` and :meth:`transformers.PreTrainedTokenizer.__call__`
for details.
`What are input IDs? <../glossary.html#input-ids>`__
Returns:
text_features (:obj:`jnp.ndarray` of shape :obj:`(batch_size, output_dim`): The text embeddings
obtained by applying the projection layer to the pooled output of text model.
"""
if position_ids is None:
position_ids = jnp.broadcast_to(jnp.arange(jnp.atleast_2d(input_ids).shape[-1]), input_ids.shape)
if token_type_ids is None:
token_type_ids = jnp.zeros_like(input_ids)
if attention_mask is None:
attention_mask = jnp.ones_like(input_ids)
# Handle any PRNG if needed
rngs = {}
if dropout_rng is not None:
rngs["dropout"] = dropout_rng
def _get_features(module, input_ids, attention_mask, position_ids, token_type_ids, deterministic):
text_outputs = module.text_model(
input_ids=input_ids,
attention_mask=attention_mask,
position_ids=position_ids,
token_type_ids=token_type_ids,
deterministic=deterministic,
)
pooled_output = text_outputs[1]
text_features = module.text_projection(pooled_output)
return text_features
return self.module.apply(
{"params": params or self.params},
jnp.array(input_ids, dtype="i4"),
jnp.array(attention_mask, dtype="i4"),
jnp.array(position_ids, dtype="i4"),
jnp.array(token_type_ids, dtype="i4"),
not train,
method=_get_features,
rngs=rngs,
)
def get_image_features(
self, pixel_values, params: dict = None, dropout_rng: jax.random.PRNGKey = None, train=False
):
r"""
Args:
pixel_values (:obj:`numpy.ndarray` of shape :obj:`(batch_size, num_channels, height, width)`):
Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained
using :class:`~transformers.ImageFeatureExtractionMixin`. See
:meth:`transformers.ImageFeatureExtractionMixin.__call__` for details.
Returns:
image_features (:obj:`jnp.ndarray` of shape :obj:`(batch_size, output_dim`): The image embeddings
obtained by applying the projection layer to the pooled output of vision model.
"""
# Handle any PRNG if needed
rngs = {}
if dropout_rng is not None:
rngs["dropout"] = dropout_rng
def _get_features(module, pixel_values, deterministic):
vision_outputs = module.vision_model(pixel_values=pixel_values, deterministic=deterministic)
pooled_output = vision_outputs[1] # pooled_output
image_features = module.visual_projection(pooled_output)
return image_features
return self.module.apply(
{"params": params or self.params},
jnp.array(pixel_values, dtype=jnp.float32),
not train,
method=_get_features,
rngs=rngs,
)
@classmethod
def from_text_vision_pretrained(
cls,
text_model_name_or_path: str = None,
vision_model_name_or_path: str = None,
*model_args,
**kwargs,
) -> FlaxPreTrainedModel:
"""
Params:
text_model_name_or_path (:obj: `str`, `optional`):
Information necessary to initiate the text model. Can be either:
- A string, the `model id` of a pretrained model hosted inside a model repo on huggingface.co.
Valid model ids can be located at the root-level, like ``bert-base-uncased``, or namespaced under
a user or organization name, like ``dbmdz/bert-base-german-cased``.
- A path to a `directory` containing model weights saved using
:func:`~transformers.FlaxPreTrainedModel.save_pretrained`, e.g., ``./my_model_directory/``.
- A path or url to a `PyTorch checkpoint folder` (e.g, ``./pt_model``). In
this case, ``from_pt`` should be set to :obj:`True` and a configuration object should be provided
as ``config`` argument. This loading path is slower than converting the PyTorch checkpoint in
a Flax model using the provided conversion scripts and loading the Flax model afterwards.
vision_model_name_or_path (:obj: `str`, `optional`, defaults to `None`):
Information necessary to initiate the vision model. Can be either:
- A string, the `model id` of a pretrained model hosted inside a model repo on huggingface.co.
Valid model ids can be located at the root-level, like ``bert-base-uncased``, or namespaced under
a user or organization name, like ``dbmdz/bert-base-german-cased``.
- A path to a `directory` containing model weights saved using
:func:`~transformers.FlaxPreTrainedModel.save_pretrained`, e.g., ``./my_model_directory/``.
- A path or url to a `PyTorch checkpoint folder` (e.g, ``./pt_model``). In
this case, ``from_pt`` should be set to :obj:`True` and a configuration object should be provided
as ``config`` argument. This loading path is slower than converting the PyTorch checkpoint in
a Flax model using the provided conversion scripts and loading the Flax model afterwards.
model_args (remaining positional arguments, `optional`):
All remaning positional arguments will be passed to the underlying model's ``__init__`` method.
kwargs (remaining dictionary of keyword arguments, `optional`):
Can be used to update the configuration object (after it being loaded) and initiate the model (e.g.,
:obj:`output_attentions=True`).
- To update the text configuration, use the prefix `text_` for each configuration parameter.
- To update the vision configuration, use the prefix `vision_` for each configuration parameter.
- To update the parent model configuration, do not use a prefix for each configuration parameter.
Behaves differently depending on whether a :obj:`config` is provided or automatically loaded.
Example::
>>> from transformers import FlaxHybridCLIP
>>> # initialize a model from pretrained BERT and CLIP models. Note that the projection layers will be randomly initialized.
>>> # If using CLIP's vision model the vision projection layer will be initialized using pre-trained weights
>>> model = FlaxHybridCLIP.from_text_vision_pretrained('bert-base-uncased', 'openai/clip-vit-base-patch32')
>>> # saving model after fine-tuning
>>> model.save_pretrained("./bert-clip")
>>> # load fine-tuned model
>>> model = FlaxHybridCLIP.from_pretrained("./bert-clip")
"""
kwargs_text = {
argument[len("text_") :]: value for argument, value in kwargs.items() if argument.startswith("text_")
}
kwargs_vision = {
argument[len("vision_") :]: value for argument, value in kwargs.items() if argument.startswith("vision_")
}
# remove text, vision kwargs from kwargs
for key in kwargs_text.keys():
del kwargs["text_" + key]
for key in kwargs_vision.keys():
del kwargs["vision_" + key]
# Load and initialize the text and vision model
text_model = kwargs_text.pop("model", None)
if text_model is None:
assert (
text_model_name_or_path is not None
), "If `model` is not defined as an argument, a `text_model_name_or_path` has to be defined"
from transformers import FlaxAutoModel
if "config" not in kwargs_text:
from transformers import AutoConfig
text_config = AutoConfig.from_pretrained(text_model_name_or_path)
kwargs_text["config"] = text_config
text_model = FlaxAutoModel.from_pretrained(text_model_name_or_path, *model_args, **kwargs_text)
vision_model = kwargs_vision.pop("model", None)
if vision_model is None:
assert (
vision_model_name_or_path is not None
), "If `model` is not defined as an argument, a `vision_model_name_or_path` has to be defined"
from transformers import FlaxAutoModel
if "config" not in kwargs_vision:
from transformers import AutoConfig
vision_config = AutoConfig.from_pretrained(vision_model_name_or_path)
kwargs_vision["config"] = vision_config
vision_model = FlaxAutoModel.from_pretrained(vision_model_name_or_path, *model_args, **kwargs_vision)
# instantiate config with corresponding kwargs
dtype = kwargs.pop("dtype", jnp.float32)
config = HybridCLIPConfig.from_text_vision_configs(text_model.config, vision_model.config, **kwargs)
# init model
model = cls(config, *model_args, dtype=dtype, **kwargs)
if vision_config.model_type == "clip":
model.params["vision_model"]["vision_model"] = vision_model.params["vision_model"]
model.params["visual_projection"]["kernel"] = vision_model.params["visual_projection"]["kernel"]
else:
model.params["vision_model"] = vision_model.params
model.params["text_model"] = text_model.params
return model
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/hybrid_clip/requirements.txt
|
jax>=0.2.8
jaxlib>=0.1.59
flax>=0.3.5
optax>=0.0.8
-f https://download.pytorch.org/whl/torch_stable.html
torch==1.9.0+cpu
-f https://download.pytorch.org/whl/torch_stable.html
torchvision==0.10.0+cpu
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/wav2vec2/README.md
|
# Wav2Vec2 Contrastive Loss PreTraining examples
The following example showcases how to pretrain a wav2vec2 model using the JAX/Flax backend.
Pretraining Wav2Vec2 is rather complex, so it is highly recommended to read the
[official paper](https://arxiv.org/abs/2006.11477).
JAX/Flax allows you to trace pure functions and compile them into efficient, fused accelerator code on both GPU and TPU.
Models written in JAX/Flax are **immutable** and updated in a purely functional
way which enables simple and efficient model parallelism.
`run_wav2vec2_pretrain_flax.py` is a lightweight example of how to download and preprocess a dataset from the 🤗 Datasets library or use your own files (jsonlines or csv), then pretrain the wav2vec2 architectures above on it.
For custom datasets in `jsonlines` format please see: [the Datasets documentation](https://huggingface.co/docs/datasets/loading_datasets#json-files) and you also will find examples of these below.
Let's start by creating a model repository to save the trained model and logs.
Here we call the model `"wav2vec2-base-robust"`, but you can change the model name as you like.
You can do this either directly on [huggingface.co](https://huggingface.co/new) (assuming that
you are logged in) or via the command line:
```
huggingface-cli repo create wav2vec2-base-robust
```
Next we clone the model repository to add the tokenizer and model files.
```
git clone https://huggingface.co/<your-username>/wav2vec2-base-robust
```
To ensure that all tensorboard traces will be uploaded correctly, we need to
track them. You can run the following command inside your model repo to do so.
```
cd wav2vec2-base-robust
git lfs track "*tfevents*"
```
Great, we have set up our model repository. During training, we will automatically
push the training logs and model weights to the repo.
Next, let's add a symbolic link to the `run_wav2vec2_pretrain_flax`.
```bash
export MODEL_DIR="./wav2vec2-base-robust"
ln -s ~/transformers/examples/research_projects/jax-projects/wav2vec2/run_wav2vec2_pretrain_flax.py ./
```
### Create the model configuration
Let's first create the model configuration and store it in the model repository.
Note that many training parameters can be set in the model configuration including
the configuration about the masking distribution (`mask_time_length`, `mask_time_prob`),
dropout (`attention_dropout`, ...), the trade-off between the contrastive loss and
the diversity loss, etc...
Mostly likely you will need to change these parameters depending on your use case.
Again, we highly recommend to read the [official paper](https://arxiv.org/abs/2006.11477)
to better understand which parameters can be set for pretraining.
For this example, we will be using a `"base"`-sized model of Wav2Vec2 with robust
layer norm and keep most of the default settings.
```python
model_dir="./wav2vec2-base-robust"
from transformers import Wav2Vec2Config
config = Wav2Vec2Config.from_pretrained(
"facebook/wav2vec2-base",
mask_time_length=10,
mask_time_prob=0.05,
diversity_loss_weight=0.1,
num_negatives=100,
do_stable_layer_norm=True,
feat_extract_norm="layer",
)
config.save_pretrained(model_dir)
```
### Create a feature extractor configuration
Before we can start the training, we need to define
a feature extractor that takes care of normalization, etc...
Here we can also re-use the feature extractor of [wav2vec2-base-960h](https://huggingface.co/facebook/wav2vec2-base) while making sure that padding is allowed.
```python
model_dir="./wav2vec2-base-robust"
from transformers import Wav2Vec2FeatureExtractor
config = Wav2Vec2FeatureExtractor.from_pretrained("facebook/wav2vec2-base", return_attention_mask=True)
config.save_pretrained(model_dir)
```
### Train the model
Finally, we can run the example script to train the model:
```bash
./run_wav2vec2_pretrain_flax.py \
--output_dir=${MODEL_DIR} \
--num_train_epochs="5" \
--per_device_train_batch_size="32" \
--per_device_eval_batch_size="32" \
--learning_rate="5e-4" \
--weight_decay="0.01" \
--warmup_steps="2000" \
--model_name_or_path=${MODEL_DIR} \
--dataset_name="librispeech_asr" \
--dataset_config_name="clean" \
--train_split_name="train.100" \
--preprocessing_num_workers="4" \
--max_duration_in_seconds="10.0" \
--adam_beta1="0.9" \
--adam_beta2="0.98" \
--pad_to_multiple_of="16384" \
--push_to_hub
```
Note that this script is not fully tested yet, so we cannot ensure that
the above script leads to satisfying results.
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/wav2vec2/run_wav2vec2_pretrain_flax.py
|
#!/usr/bin/env python3
import logging
import sys
import time
from dataclasses import field
from pathlib import Path
from typing import Dict, List, Optional, Union
import flax
import jax
import jax.numpy as jnp
import librosa
import numpy as np
import optax
from datasets import DatasetDict, load_dataset
from flax import jax_utils, traverse_util
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard
from tqdm import tqdm
from transformers import (
FlaxWav2Vec2ForPreTraining,
HfArgumentParser,
TrainingArguments,
Wav2Vec2Config,
Wav2Vec2FeatureExtractor,
is_tensorboard_available,
)
from transformers.models.wav2vec2.modeling_flax_wav2vec2 import _compute_mask_indices, _sample_negative_indices
logger = logging.getLogger(__name__)
@flax.struct.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"}
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
)
freeze_feature_extractor: Optional[bool] = field(
default=True, metadata={"help": "Whether to freeze the feature extractor layers of the model."}
)
verbose_logging: Optional[bool] = field(
default=False,
metadata={"help": "Whether to log verbose messages or not."},
)
max_gumbel_temperature: Optional[float] = field(
default=2.0, metadata={"help": "Maximum temperature for gumbel softmax."}
)
min_gumbel_temperature: Optional[float] = field(
default=0.1, metadata={"help": "Minimum temperature for gumbel softmax."}
)
gumbel_temperature_decay: Optional[float] = field(
default=0.999995, metadata={"help": "Decay of gumbel temperature during training."}
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
@flax.struct.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.
"""
dataset_name: 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)."}
)
train_split_name: Optional[str] = field(
default="train",
metadata={
"help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
},
)
validation_split_name: Optional[str] = field(
default="validation",
metadata={
"help": (
"The name of the validation data set split to use (via the datasets library). Defaults to 'validation'"
)
},
)
speech_file_column: Optional[str] = field(
default="file",
metadata={"help": "Column in the dataset that contains speech file path. Defaults to 'file'"},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
max_duration_in_seconds: Optional[float] = field(
default=20.0, metadata={"help": "Filter audio files that are longer than `max_duration_in_seconds` seconds"}
)
pad_to_multiple_of: Optional[int] = field(
default=1024,
metadata={
"help": (
"If set will pad the sequence to a multiple of the provided value. This is important to avoid"
" triggering recompilations on TPU"
)
},
)
@flax.struct.dataclass
class FlaxDataCollatorForWav2Vec2Pretraining:
"""
Data collator that will dynamically pad the inputs received and prepare masked indices
for self-supervised pretraining.
Args:
model (:class:`~transformers.FlaxWav2Vec2ForPreTraining`):
The Wav2Vec2 model used for pretraining. The data collator needs to have access
to config and ``_get_feat_extract_output_lengths`` function for correct padding.
feature_extractor (:class:`~transformers.Wav2Vec2FeatureExtractor`):
The processor used for proccessing the data.
padding (:obj:`bool`, :obj:`str` or :class:`~transformers.tokenization_utils_base.PaddingStrategy`, `optional`, defaults to :obj:`True`):
Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
among:
* :obj:`True` or :obj:`'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
sequence if provided).
* :obj:`'max_length'`: Pad to a maximum length specified with the argument :obj:`max_length` or to the
maximum acceptable input length for the model if that argument is not provided.
* :obj:`False` or :obj:`'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of
different lengths).
max_length (:obj:`int`, `optional`):
Maximum length of the ``input_values`` of the returned list and optionally padding length (see above).
pad_to_multiple_of (:obj:`int`, `optional`):
If set will pad the sequence to a multiple of the provided value.
This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
7.5 (Volta).
"""
model: FlaxWav2Vec2ForPreTraining
feature_extractor: Wav2Vec2FeatureExtractor
padding: Union[bool, str] = "longest"
pad_to_multiple_of: Optional[int] = None
max_length: Optional[int] = None
def __call__(self, features: List[Dict[str, Union[List[int], np.ndarray]]]) -> Dict[str, np.ndarray]:
# reformat list to dict and set to pytorch format
batch = self.feature_extractor.pad(
features,
max_length=self.max_length,
padding=self.padding,
pad_to_multiple_of=self.pad_to_multiple_of,
return_tensors="np",
)
mask_indices_seq_length = self.model._get_feat_extract_output_lengths(batch["input_values"].shape[-1])
batch_size = batch["input_values"].shape[0]
attention_mask = None
if batch["attention_mask"] is not None:
output_lengths = self.model._get_feat_extract_output_lengths(batch["attention_mask"].sum(-1))
attention_mask = np.zeros((batch_size, mask_indices_seq_length), dtype=np.int8)
# these two operations makes sure that all values
# before the output lengths indices are attended to
attention_mask[(np.arange(attention_mask.shape[0]), output_lengths - 1)] = 1
attention_mask = jnp.flip(jnp.flip(attention_mask, -1).cumsum(-1), -1).astype("bool")
# sample randomly masked indices
batch["mask_time_indices"] = _compute_mask_indices(
(batch_size, mask_indices_seq_length),
self.model.config.mask_time_prob,
self.model.config.mask_time_length,
attention_mask=attention_mask,
min_masks=2,
)
# sample indices to take for negative vectors
batch["sampled_negative_indices"] = _sample_negative_indices(
(batch["mask_time_indices"].shape + (self.model.config.proj_codevector_dim,)),
self.model.config.num_negatives,
attention_mask=attention_mask,
)
return batch
def configure_logger(model_args: ModelArguments, training_args: TrainingArguments):
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
handlers=[logging.StreamHandler(sys.stdout)],
)
logging_level = logging.WARNING
if model_args.verbose_logging:
logging_level = logging.DEBUG
logger.setLevel(logging_level)
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
def generate_batch_splits(samples_idx: np.ndarray, batch_size: int) -> np.ndarray:
num_samples = len(samples_idx)
samples_to_remove = num_samples % batch_size
if samples_to_remove != 0:
samples_idx = samples_idx[:-samples_to_remove]
sections_split = num_samples // batch_size
batch_idx = np.split(samples_idx, sections_split)
return batch_idx
def compute_contrastive_loss(
quantized_features, transformer_features, negative_indices, mask_time_indices, logits_temp, num_negatives
):
batch_size, sequence_length, hidden_size = quantized_features.shape
# take negative vectors from sampled indices
quantized_negatives = quantized_features.reshape(-1, hidden_size)[negative_indices.reshape(-1)]
quantized_negatives = quantized_negatives.reshape(
batch_size, sequence_length, num_negatives, hidden_size
).transpose(2, 0, 1, 3)
target_features = jnp.concatenate([quantized_features[None, :], quantized_negatives], axis=0)
loss_logits = optax.cosine_similarity(transformer_features, target_features)
loss_logits = loss_logits / logits_temp
neg_is_pos = (quantized_features == quantized_negatives).all(-1)
neg_is_pos = jnp.concatenate([jnp.full((1,) + loss_logits.shape[1:], False), neg_is_pos], axis=0)
# make sure incorrectly sampled vectors don't contribute to loss
loss_logits = jnp.where(neg_is_pos, -1e9, loss_logits)
predictions = loss_logits.transpose(2, 1, 0).reshape(-1, loss_logits.shape[0])
targets = ((1 - mask_time_indices) * -100).transpose(1, 0).flatten()
target_mask = jnp.where(targets >= 0, 1.0, 0.0)
contrastive_loss = optax.softmax_cross_entropy(predictions, onehot(targets, predictions.shape[-1])) * target_mask
contrastive_loss = contrastive_loss.sum()
return contrastive_loss
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))
model_args, data_args, training_args = parser.parse_args_into_dataclasses()
configure_logger(model_args, training_args)
# Downloading and loading a dataset from the hub.
datasets = load_dataset(data_args.dataset_name, data_args.dataset_config_name, cache_dir=model_args.cache_dir)
if "validation" not in datasets.keys():
# make sure only "validation" and "train" keys remain"
datasets = DatasetDict()
datasets["validation"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"{data_args.train_split_name}[:{data_args.validation_split_percentage}%]",
cache_dir=model_args.cache_dir,
)
datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"{data_args.train_split_name}[{data_args.validation_split_percentage}%:]",
cache_dir=model_args.cache_dir,
)
else:
# make sure only "validation" and "train" keys remain"
datasets = DatasetDict()
datasets["validation"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split="validation",
cache_dir=model_args.cache_dir,
)
datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=f"{data_args.train_split_name}",
cache_dir=model_args.cache_dir,
)
# only normalized-inputs-training is supported
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, do_normalize=True
)
def prepare_dataset(batch):
# check that all files have the correct sampling rate
batch["speech"], _ = librosa.load(batch[data_args.speech_file_column], sr=feature_extractor.sampling_rate)
return batch
# load audio files into numpy arrays
vectorized_datasets = datasets.map(
prepare_dataset, num_proc=data_args.preprocessing_num_workers, remove_columns=datasets["train"].column_names
)
# filter audio files that are too long
vectorized_datasets = vectorized_datasets.filter(
lambda data: len(data["speech"]) < int(data_args.max_duration_in_seconds * feature_extractor.sampling_rate)
)
def normalize(batch):
return feature_extractor(batch["speech"], sampling_rate=feature_extractor.sampling_rate)
# normalize and transform to `BatchFeatures`
vectorized_datasets = vectorized_datasets.map(
normalize,
batched=True,
num_proc=data_args.preprocessing_num_workers,
load_from_cache_file=not data_args.overwrite_cache,
remove_columns=vectorized_datasets["train"].column_names,
)
# pretraining is only supported for "newer" stable layer norm architecture
# apply_spec_augment has to be True, mask_feature_prob has to be 0.0
config = Wav2Vec2Config.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
)
if not config.do_stable_layer_norm or config.feat_extract_norm != "layer":
raise ValueError(
"PreTraining is only supported for ``config.do_stable_layer_norm=True`` and"
" ``config.feat_extract_norm='layer'"
)
model = FlaxWav2Vec2ForPreTraining(config, seed=training_args.seed, dtype=getattr(jnp, model_args.dtype))
# Activate gradient checkpointing if needed
if training_args.gradient_checkpointing:
model.gradient_checkpointing_enable()
data_collator = FlaxDataCollatorForWav2Vec2Pretraining(
model=model, feature_extractor=feature_extractor, pad_to_multiple_of=data_args.pad_to_multiple_of
)
# Enable tensorboard only on the master node
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
else:
logger.warning(
"Unable to display metrics through TensorBoard because the package is not installed: "
"Please run pip install tensorboard to enable."
)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
dropout_rngs = jax.random.split(rng, jax.local_device_count())
gumbel_rngs = jax.random.split(rng, jax.local_device_count())
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
eval_batch_size = int(training_args.per_device_eval_batch_size) * jax.device_count()
num_train_steps = len(vectorized_datasets["train"]) // train_batch_size * num_epochs
# Create learning rate schedule
warmup_fn = optax.linear_schedule(
init_value=0.0, end_value=training_args.learning_rate, transition_steps=training_args.warmup_steps
)
decay_fn = optax.linear_schedule(
init_value=training_args.learning_rate,
end_value=0,
transition_steps=num_train_steps - training_args.warmup_steps,
)
linear_decay_lr_schedule_fn = optax.join_schedules(
schedules=[warmup_fn, decay_fn], boundaries=[training_args.warmup_steps]
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
flat_mask = {
path: (path[-1] != "bias" and path[-2:] not in [("layer_norm", "scale"), ("final_layer_norm", "scale")])
for path in flat_params
}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
adamw = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state and define training hyper-parameters
state = train_state.TrainState.create(apply_fn=model.__call__, params=model.params, tx=adamw)
num_negatives = model.config.num_negatives
contrastive_logits_temperature = model.config.contrastive_logits_temperature
num_codevectors = model.config.num_codevectors_per_group * model.config.num_codevector_groups
diversity_loss_weight = model.config.diversity_loss_weight
# Define gradient update step fn
def train_step(state, batch, dropout_rng, gumbel_rng):
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
gumbel_rng, new_gumbel_rng = jax.random.split(gumbel_rng)
def loss_fn(params):
negative_indices = batch.pop("sampled_negative_indices")
gumbel_temperature = jnp.clip(
model_args.max_gumbel_temperature * model_args.gumbel_temperature_decay**state.step,
a_min=model_args.min_gumbel_temperature,
)
outputs = state.apply_fn(
**batch,
gumbel_temperature=gumbel_temperature,
params=params,
dropout_rng=dropout_rng,
gumbel_rng=gumbel_rng,
train=True,
)
contrastive_loss = compute_contrastive_loss(
outputs.projected_quantized_states,
outputs.projected_states,
negative_indices,
batch["mask_time_indices"],
contrastive_logits_temperature,
num_negatives,
)
diversity_loss = (num_codevectors - outputs.codevector_perplexity) / num_codevectors
loss = contrastive_loss + diversity_loss_weight * diversity_loss
return loss
grad_fn = jax.value_and_grad(loss_fn)
loss, grad = grad_fn(state.params)
grad = jax.lax.pmean(grad, "batch")
new_state = state.apply_gradients(grads=grad)
metrics = jax.lax.pmean(
{"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}, axis_name="batch"
)
return new_state, metrics, new_dropout_rng, new_gumbel_rng
# Create parallel version of the train step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
# Define eval fn
def eval_step(params, batch):
negative_indices = batch.pop("sampled_negative_indices")
outputs = model(**batch, params=params, train=False)
contrastive_loss = compute_contrastive_loss(
outputs.projected_quantized_states,
outputs.projected_states,
negative_indices,
batch["mask_time_indices"],
contrastive_logits_temperature,
num_negatives,
)
diversity_loss = (num_codevectors - outputs.codevector_perplexity) / num_codevectors
loss = contrastive_loss + diversity_loss_weight * diversity_loss
# summarize metrics
metrics = {"loss": loss.mean(), "codevector_perplexity": outputs.codevector_perplexity}
metrics = jax.lax.pmean(metrics, axis_name="batch")
return metrics
p_eval_step = jax.pmap(eval_step, "batch", donate_argnums=(0,))
# Replicate the train state on each device
state = jax_utils.replicate(state)
train_time = 0
train_metrics = []
epochs = tqdm(range(num_epochs), desc=f"Epoch ... (1/{num_epochs})", position=0)
for epoch in epochs:
# ======================== Training ================================
train_start = time.time()
# Create sampling rng
rng, input_rng = jax.random.split(rng)
# Generate an epoch by shuffling sampling indices from the train dataset
num_train_samples = len(vectorized_datasets["train"])
# Avoid using jax.numpy here in case of TPU training
train_samples_idx = np.random.permutation(np.arange(num_train_samples))
train_batch_idx = generate_batch_splits(train_samples_idx, train_batch_size)
# Gather the indexes for creating the batch and do a training step
for step, batch_idx in enumerate(tqdm(train_batch_idx, desc="Training...", position=1)):
samples = [vectorized_datasets["train"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
model_inputs = shard(model_inputs.data)
# Model forward
state, train_metric, dropout_rngs, gumbel_rngs = p_train_step(
state, model_inputs, dropout_rngs, gumbel_rngs
)
train_metrics.append(train_metric)
cur_step = epoch * (num_train_samples // train_batch_size) + step
if cur_step % training_args.logging_steps == 0 and cur_step > 0:
# Save metrics
train_metric = jax_utils.unreplicate(train_metric)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, cur_step)
epochs.write(
f"Step... ({cur_step} | Loss: {train_metric['loss'].mean()}, Learning Rate:"
f" {train_metric['learning_rate'].mean()})"
)
train_metrics = []
# ======================== Evaluating ==============================
num_eval_samples = len(vectorized_datasets["validation"])
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size)
eval_metrics = []
for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=2)):
samples = [vectorized_datasets["validation"][int(idx)] for idx in batch_idx]
model_inputs = data_collator(samples)
# Model forward
model_inputs = shard(model_inputs.data)
metrics = p_eval_step(state.params, model_inputs)
eval_metrics.append(metrics)
# get eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.mean, eval_metrics)
# Update progress bar
epochs.write(
f"Epoch... ({epoch + 1}/{num_epochs} | Loss: {eval_metrics['loss']}, Perplexity:"
f" {eval_metrics['codevector_perplexity']})"
)
# Save metrics
if has_tensorboard and jax.process_index() == 0:
cur_step = epoch * (len(vectorized_datasets["train"]) // train_batch_size)
write_eval_metric(summary_writer, eval_metrics, cur_step)
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(jax.tree_util.tree_map(lambda x: x[0], state.params))
model.save_pretrained(training_args.output_dir, params=params, push_to_hub=training_args.push_to_hub)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/dataset-streaming/run_mlm_flax_stream.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Team 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.
"""
Fine-tuning the library models for masked language modeling (BERT, ALBERT, RoBERTa...) with whole word masking on a
text file or a dataset.
Here is the full list of checkpoints on the hub that can be fine-tuned by this script:
https://huggingface.co/models?filter=fill-mask
"""
import logging
import os
import sys
import time
from collections import defaultdict
from dataclasses import dataclass, field
# You can also adapt this script on your own masked language modeling task. Pointers for this are left as comments.
from pathlib import Path
from typing import Dict, List, Optional, Tuple
import datasets
import flax
import jax
import jax.numpy as jnp
import numpy as np
import optax
from datasets import load_dataset
from flax import jax_utils, traverse_util
from flax.training import train_state
from flax.training.common_utils import get_metrics, onehot, shard
from tqdm import tqdm
from transformers import (
CONFIG_MAPPING,
FLAX_MODEL_FOR_MASKED_LM_MAPPING,
AutoConfig,
AutoTokenizer,
FlaxAutoModelForMaskedLM,
HfArgumentParser,
PreTrainedTokenizerBase,
TensorType,
TrainingArguments,
is_tensorboard_available,
set_seed,
)
if datasets.__version__ <= "1.8.0":
raise ValueError("Make sure to upgrade `datasets` to a version >= 1.9.0 to use dataset streaming")
MODEL_CONFIG_CLASSES = list(FLAX_MODEL_FOR_MASKED_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
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"}
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
dtype: Optional[str] = field(
default="float32",
metadata={
"help": (
"Floating-point format in which the model weights should be initialized and trained. Choose one of"
" `[float32, float16, bfloat16]`."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
train_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input train ref data file for whole word masking in Chinese."},
)
validation_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input validation ref data file for whole word masking in Chinese."},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
max_seq_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated. Default to the max input length of the model."
)
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
mlm_probability: float = field(
default=0.15, metadata={"help": "Ratio of tokens to mask for masked language modeling loss"}
)
pad_to_max_length: bool = field(
default=False,
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."
)
},
)
line_by_line: bool = field(
default=False,
metadata={"help": "Whether distinct lines of text in the dataset are to be handled as distinct sequences."},
)
text_column_name: str = field(
default="text", metadata={"help": "The name of the column to retrieve the training text."}
)
shuffle_buffer_size: int = field(
default=10000, metadata={"help": "The number of examples to pre-load for shuffling."}
)
num_train_steps: int = field(default=50000, metadata={"help": "The number of training steps."})
num_eval_samples: int = field(default=50000, metadata={"help": "The number of samples to be used for evaluation"})
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file."
@flax.struct.dataclass
class FlaxDataCollatorForLanguageModeling:
"""
Data collator used for language modeling. Inputs are dynamically padded to the maximum length of a batch if they
are not all of the same length.
Args:
tokenizer (:class:`~transformers.PreTrainedTokenizer` or :class:`~transformers.PreTrainedTokenizerFast`):
The tokenizer used for encoding the data.
mlm_probability (:obj:`float`, `optional`, defaults to 0.15):
The probability with which to (randomly) mask tokens in the input.
.. note::
For best performance, this data collator should be used with a dataset having items that are dictionaries or
BatchEncoding, with the :obj:`"special_tokens_mask"` key, as returned by a
:class:`~transformers.PreTrainedTokenizer` or a :class:`~transformers.PreTrainedTokenizerFast` with the
argument :obj:`return_special_tokens_mask=True`.
"""
tokenizer: PreTrainedTokenizerBase
mlm_probability: float = 0.15
def __post_init__(self):
if self.tokenizer.mask_token is None:
raise ValueError(
"This tokenizer does not have a mask token which is necessary for masked language modeling. "
"You should pass `mlm=False` to train on causal language modeling instead."
)
def __call__(self, examples: List[Dict[str, np.ndarray]]) -> Dict[str, np.ndarray]:
# Handle dict or lists with proper padding and conversion to tensor.
batch = self.tokenizer.pad(examples, return_tensors=TensorType.NUMPY)
# If special token mask has been preprocessed, pop it from the dict.
special_tokens_mask = batch.pop("special_tokens_mask", None)
batch["input_ids"], batch["labels"] = self.mask_tokens(
batch["input_ids"], special_tokens_mask=special_tokens_mask
)
return batch
def mask_tokens(
self, inputs: np.ndarray, special_tokens_mask: Optional[np.ndarray]
) -> Tuple[jnp.ndarray, jnp.ndarray]:
"""
Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.
"""
labels = inputs.copy()
# We sample a few tokens in each sequence for MLM training (with probability `self.mlm_probability`)
probability_matrix = np.full(labels.shape, self.mlm_probability)
special_tokens_mask = special_tokens_mask.astype("bool")
probability_matrix[special_tokens_mask] = 0.0
masked_indices = np.random.binomial(1, probability_matrix).astype("bool")
labels[~masked_indices] = -100 # We only compute loss on masked tokens
# 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
indices_replaced = np.random.binomial(1, np.full(labels.shape, 0.8)).astype("bool") & masked_indices
inputs[indices_replaced] = self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token)
# 10% of the time, we replace masked input tokens with random word
indices_random = np.random.binomial(1, np.full(labels.shape, 0.5)).astype("bool")
indices_random &= masked_indices & ~indices_replaced
random_words = np.random.randint(self.tokenizer.vocab_size, size=labels.shape, dtype="i4")
inputs[indices_random] = random_words[indices_random]
# The rest of the time (10% of the time) we keep the masked input tokens unchanged
return inputs, labels
def generate_batch_splits(samples_idx: np.ndarray, batch_size: int) -> np.ndarray:
num_samples = len(samples_idx)
samples_to_remove = num_samples % batch_size
if samples_to_remove != 0:
samples_idx = samples_idx[:-samples_to_remove]
sections_split = num_samples // batch_size
batch_idx = np.split(samples_idx, sections_split)
return batch_idx
def advance_iter_and_group_samples(train_iterator, num_samples, max_seq_length):
"""
The training iterator is advanced so that after groupifying the samples,
`num_samples` of length `max_seq_length` are returned.
"""
num_total_tokens = max_seq_length * num_samples
samples = defaultdict(list)
i = 0
while i < num_total_tokens:
tokenized_samples = next(train_iterator)
i += len(tokenized_samples["input_ids"])
# concatenate tokenized samples to list (excluding "id" and "text")
samples = {
k: samples[k] + tokenized_samples[k] for k in ["input_ids", "attention_mask", "special_tokens_mask"]
}
# Concatenated tokens are split to lists of length `max_seq_length`.
# Note that remainedr of % max_seq_length are thrown away.
def group_texts(examples):
result = {
k: [t[i : i + max_seq_length] for i in range(0, num_total_tokens, max_seq_length)]
for k, t in examples.items()
}
return result
grouped_samples = group_texts(samples)
return grouped_samples
def write_train_metric(summary_writer, train_metrics, train_time, step):
summary_writer.scalar("train_time", train_time, step)
train_metrics = get_metrics(train_metrics)
for key, vals in train_metrics.items():
tag = f"train_{key}"
for i, val in enumerate(vals):
summary_writer.scalar(tag, val, step - len(vals) + i + 1)
def write_eval_metric(summary_writer, eval_metrics, step):
for metric_name, value in eval_metrics.items():
summary_writer.scalar(f"eval_{metric_name}", value, step)
if __name__ == "__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",
level="INFO",
datefmt="[%X]",
)
# Log on each process the small summary:
logger = logging.getLogger(__name__)
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):
logger.info(f"Training/evaluation parameters {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 if no column called
# 'text' is found. You can easily tweak this behavior (see below).
if data_args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
dataset = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
cache_dir=model_args.cache_dir,
streaming=True,
split="train",
)
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.")
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer
)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
# 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[data_args.text_column_name], return_special_tokens_mask=True)
tokenized_datasets = dataset.map(tokenize_function, batched=True, remove_columns=list(dataset.features.keys()))
shuffle_seed = training_args.seed
tokenized_datasets = tokenized_datasets.shuffle(buffer_size=data_args.shuffle_buffer_size, seed=shuffle_seed)
has_tensorboard = is_tensorboard_available()
if has_tensorboard and jax.process_index() == 0:
try:
from flax.metrics.tensorboard import SummaryWriter
except ImportError as ie:
has_tensorboard = False
logger.warning(
f"Unable to display metrics through TensorBoard because some package are not installed: {ie}"
)
summary_writer = SummaryWriter(log_dir=Path(training_args.output_dir))
# Data collator
# This one will take care of randomly masking the tokens.
data_collator = FlaxDataCollatorForLanguageModeling(tokenizer=tokenizer, mlm_probability=data_args.mlm_probability)
# Initialize our training
rng = jax.random.PRNGKey(training_args.seed)
dropout_rngs = jax.random.split(rng, jax.local_device_count())
if model_args.model_name_or_path:
model = FlaxAutoModelForMaskedLM.from_pretrained(
model_args.model_name_or_path, config=config, seed=training_args.seed, dtype=getattr(jnp, model_args.dtype)
)
else:
model = FlaxAutoModelForMaskedLM.from_config(
config, seed=training_args.seed, dtype=getattr(jnp, model_args.dtype)
)
# Store some constant
num_epochs = int(training_args.num_train_epochs)
train_batch_size = int(training_args.per_device_train_batch_size) * jax.device_count()
eval_batch_size = int(training_args.per_device_eval_batch_size) * jax.device_count()
# define number steps per stream epoch
num_train_steps = data_args.num_train_steps
# Create learning rate schedule
warmup_fn = optax.linear_schedule(
init_value=0.0, end_value=training_args.learning_rate, transition_steps=training_args.warmup_steps
)
decay_fn = optax.linear_schedule(
init_value=training_args.learning_rate,
end_value=0,
transition_steps=num_train_steps - training_args.warmup_steps,
)
linear_decay_lr_schedule_fn = optax.join_schedules(
schedules=[warmup_fn, decay_fn], boundaries=[training_args.warmup_steps]
)
# We use Optax's "masking" functionality to not apply weight decay
# to bias and LayerNorm scale parameters. decay_mask_fn returns a
# mask boolean with the same structure as the parameters.
# The mask is True for parameters that should be decayed.
# Note that this mask is specifically adapted for FlaxBERT-like models.
# For other models, one should correct the layer norm parameter naming
# accordingly.
def decay_mask_fn(params):
flat_params = traverse_util.flatten_dict(params)
flat_mask = {path: (path[-1] != "bias" and path[-2:] != ("LayerNorm", "scale")) for path in flat_params}
return traverse_util.unflatten_dict(flat_mask)
# create adam optimizer
adamw = optax.adamw(
learning_rate=linear_decay_lr_schedule_fn,
b1=training_args.adam_beta1,
b2=training_args.adam_beta2,
eps=training_args.adam_epsilon,
weight_decay=training_args.weight_decay,
mask=decay_mask_fn,
)
# Setup train state
state = train_state.TrainState.create(apply_fn=model.__call__, params=model.params, tx=adamw)
# Define gradient update step fn
def train_step(state, batch, dropout_rng):
dropout_rng, new_dropout_rng = jax.random.split(dropout_rng)
def loss_fn(params):
labels = batch.pop("labels")
logits = state.apply_fn(**batch, params=params, dropout_rng=dropout_rng, train=True)[0]
# compute loss, ignore padded input tokens
label_mask = jnp.where(labels > 0, 1.0, 0.0)
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
# take average
loss = loss.sum() / label_mask.sum()
return loss
grad_fn = jax.value_and_grad(loss_fn)
loss, grad = grad_fn(state.params)
grad = jax.lax.pmean(grad, "batch")
new_state = state.apply_gradients(grads=grad)
metrics = jax.lax.pmean(
{"loss": loss, "learning_rate": linear_decay_lr_schedule_fn(state.step)}, axis_name="batch"
)
return new_state, metrics, new_dropout_rng
# Create parallel version of the train step
p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
# Define eval fn
def eval_step(params, batch):
labels = batch.pop("labels")
logits = model(**batch, params=params, train=False)[0]
# compute loss, ignore padded input tokens
label_mask = jnp.where(labels > 0, 1.0, 0.0)
loss = optax.softmax_cross_entropy(logits, onehot(labels, logits.shape[-1])) * label_mask
# compute accuracy
accuracy = jnp.equal(jnp.argmax(logits, axis=-1), labels) * label_mask
# summarize metrics
metrics = {"loss": loss.sum(), "accuracy": accuracy.sum(), "normalizer": label_mask.sum()}
metrics = jax.lax.psum(metrics, axis_name="batch")
return metrics
p_eval_step = jax.pmap(eval_step, "batch", donate_argnums=(0,))
# Replicate the train state on each device
state = jax_utils.replicate(state)
train_time = 0
train_start = time.time()
train_metrics = []
eval_metrics = []
training_iter = iter(tokenized_datasets)
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
eval_samples = advance_iter_and_group_samples(training_iter, data_args.num_eval_samples, max_seq_length)
steps = tqdm(range(num_train_steps), desc="Training...", position=0)
for step in range(num_train_steps):
# ======================== Training ================================
try:
samples = advance_iter_and_group_samples(training_iter, train_batch_size, max_seq_length)
except StopIteration:
# Once the end of the dataset stream is reached, the training iterator
# is reinitialized and reshuffled and a new eval dataset is randomly chosen.
shuffle_seed += 1
tokenized_datasets.set_epoch(shuffle_seed)
training_iter = iter(tokenized_datasets)
eval_dataset = advance_iter_and_group_samples(training_iter, data_args.num_eval_samples, max_seq_length)
samples = advance_iter_and_group_samples(training_iter, train_batch_size, max_seq_length)
# process input samples
model_inputs = data_collator(samples)
# Model forward
model_inputs = shard(model_inputs.data)
state, train_metric, dropout_rngs = p_train_step(state, model_inputs, dropout_rngs)
train_metrics.append(train_metric)
if step % training_args.logging_steps == 0 and step > 0:
steps.write(
f"Step... ({step} | Loss: {train_metric['loss'].mean()}, Learning Rate:"
f" {train_metric['learning_rate'].mean()})"
)
train_time += time.time() - train_start
if has_tensorboard and jax.process_index() == 0:
write_train_metric(summary_writer, train_metrics, train_time, step)
train_metrics = []
# ======================== Evaluating ==============================
if step % training_args.eval_steps == 0 and step > 0:
# Avoid using jax.numpy here in case of TPU training
eval_samples_idx = np.arange(data_args.num_eval_samples)
eval_batch_idx = generate_batch_splits(eval_samples_idx, eval_batch_size)
for i, batch_idx in enumerate(tqdm(eval_batch_idx, desc="Evaluating ...", position=1)):
# process input samples
batch_eval_samples = {k: [v[idx] for idx in batch_idx] for k, v in eval_samples.items()}
model_inputs = data_collator(batch_eval_samples)
# Model forward
model_inputs = shard(model_inputs.data)
metrics = p_eval_step(state.params, model_inputs)
eval_metrics.append(metrics)
# normalize eval metrics
eval_metrics = get_metrics(eval_metrics)
eval_metrics = jax.tree_util.tree_map(jnp.sum, eval_metrics)
eval_normalizer = eval_metrics.pop("normalizer")
eval_metrics = jax.tree_util.tree_map(lambda x: x / eval_normalizer, eval_metrics)
# Update progress bar
steps.desc = (
f"Step... ({step + 1}/{num_train_steps} | Loss: {eval_metrics['loss']}, Acc:"
f" {eval_metrics['accuracy']})"
)
if has_tensorboard and jax.process_index() == 0:
write_eval_metric(summary_writer, eval_metrics, step)
eval_metrics = []
# save checkpoint after each epoch and push checkpoint to the hub
if jax.process_index() == 0:
params = jax.device_get(jax.tree_util.tree_map(lambda x: x[0], state.params))
model.save_pretrained(
training_args.output_dir,
params=params,
push_to_hub=training_args.push_to_hub,
commit_message=f"Saving weights and logs of step {step+1}",
)
# update tqdm bar
steps.update(1)
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/dataset-streaming/README.md
|
<!---
Copyright 2021 The HuggingFace Team. 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.
-->
# Language model training examples in streaming mode
The following examples showcase how to train a language model from scratch
using the JAX/Flax backend.
JAX/Flax allows you to trace pure functions and compile them into efficient, fused accelerator code on both GPU and TPU.
Models written in JAX/Flax are **immutable** and updated in a purely functional
way which enables simple and efficient model parallelism.
All of the following examples make use of [dataset streaming](https://huggingface.co/docs/datasets/master/dataset_streaming), therefore allowing to train models on massive datasets\
without ever having to download the full dataset.
## Masked language modeling
In the following, we demonstrate how to train a bi-directional transformer model
using masked language modeling objective as introduced in [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://arxiv.org/abs/1810.04805).
More specifically, we demonstrate how JAX/Flax and dataset streaming can be leveraged
to pre-train [**`roberta-base`**](https://huggingface.co/roberta-base)
in English on a single TPUv3-8 pod for 10000 update steps.
The example script uses the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.
Let's start by creating a model repository to save the trained model and logs.
Here we call the model `"english-roberta-base-dummy"`, but you can change the model name as you like.
You can do this either directly on [huggingface.co](https://huggingface.co/new) (assuming that
you are logged in) or via the command line:
```
huggingface-cli repo create english-roberta-base-dummy
```
Next we clone the model repository to add the tokenizer and model files.
```
git clone https://huggingface.co/<your-username>/english-roberta-base-dummy
```
To ensure that all tensorboard traces will be uploaded correctly, we need to
track them. You can run the following command inside your model repo to do so.
```
cd english-roberta-base-dummy
git lfs track "*tfevents*"
```
Great, we have set up our model repository. During training, we will automatically
push the training logs and model weights to the repo.
Next, let's add a symbolic link to the `run_mlm_flax.py`.
```bash
export MODEL_DIR="./english-roberta-base-dummy"
ln -s ~/transformers/examples/research_projects/jax-projects/dataset-streaming/run_mlm_flax_stream.py ./
```
### Copy config and tokenizer of existing model
In this example, we will simply copy an existing config and tokenizer in English.
You can run the following code in a Python shell to do so.
```python
from transformers import RobertaTokenizerFast, RobertaConfig
model_dir = "./english-roberta-base-dummy"
tokenizer = RobertaTokenizerFast.from_pretrained("roberta-base")
config = RobertaConfig.from_pretrained("roberta-base")
tokenizer.save_pretrained(model_dir)
config.save_pretrained(model_dir)
```
### Train model
Next we can run the example script to pretrain the model.
Compared to the default [`run_mlm_flax`](https://github.com/huggingface/transformers/blob/main/examples/flax/language-modeling/run_mlm_flax.py), we introduced 4 new training settings:
- `num_train_steps` - how many update steps should be run.
- `num_eval_samples` - how many training samples should be taken for evaluation.
- `logging_steps` - at what rate should the training loss be logged.
- `eval_steps` - at what rate should evaluation be run.
10K update steps
```bash
./run_mlm_flax_stream.py \
--output_dir="${MODEL_DIR}" \
--model_type="roberta" \
--config_name="${MODEL_DIR}" \
--tokenizer_name="${MODEL_DIR}" \
--dataset_name="oscar" \
--dataset_config_name="unshuffled_deduplicated_en" \
--max_seq_length="128" \
--per_device_train_batch_size="128" \
--per_device_eval_batch_size="128" \
--learning_rate="3e-4" \
--warmup_steps="1000" \
--overwrite_output_dir \
--adam_beta1="0.9" \
--adam_beta2="0.98" \
--num_train_steps="10000" \
--num_eval_samples="5000" \
--logging_steps="250" \
--eval_steps="1000" \
--push_to_hub
```
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/README.md
|
Author: [@vasudevgupta7](https://github.com/thevasudevgupta/)
## Intro
In this project, we fine-tuned [**BigBird**](https://arxiv.org/abs/2007.14062) on [**natural-questions**](https://huggingface.co/datasets/natural_questions) dataset for **question-answering** task on long documents. **BigBird**, is a **sparse-attention based transformer** which extends Transformer based models, such as BERT to much **longer sequences**.
Read more about BigBird at https://huggingface.co/blog/big-bird
## Fine-tuning
**Setup**
You need to install jax yourself by following the official docs ([refer this](https://github.com/google/jax#installation)). Other requirements for this project can be installed by running following command:
```shell
pip3 install -qr requirements.txt
```
**Download & prepare dataset**
The Natural Questions corpus contains questions from real users, and it requires QA systems to read and comprehend an entire Wikipedia article that may or may not contain the answer to the question. This corpus takes ~100 GB on disk. We have used HuggingFace datasets to download & process the dataset.
```shell
# just run following CMD
python3 prepare_natural_questions.py
# this will download the whole dataset from HuggingFace Hub & will make it ready for training
# this script takes ~3 hours to process the dataset
```
**Launch Training**
We have trained on Cloud's TPU v3-8. Each epoch took around 4.5 hours and the model got converged in just 2 epochs. You can see complete training args in [this script](bigbird_flax.py).
```shell
# just run following CMD
python3 train.py
# In case, you want to try hparams tuning, you can run wandb sweep
wandb sweep --project=bigbird sweep_flax.yaml
wandb agent <agent-id-obtained-by-above-CMD>
```
## Evaluation
Our evaluation script is different from the original script and we are evaluating sequences with length up to 4096 for simplicity. We managed to get the **EM score of ~55.2** using our evaluation script.
```shell
# download validation-dataset first
mkdir natural-questions-validation
wget https://huggingface.co/datasets/vasudevgupta/natural-questions-validation/resolve/main/natural_questions-validation.arrow -P natural-questions-validation
wget https://huggingface.co/datasets/vasudevgupta/natural-questions-validation/resolve/main/dataset_info.json -P natural-questions-validation
wget https://huggingface.co/datasets/vasudevgupta/natural-questions-validation/resolve/main/state.json -P natural-questions-validation
# simply run following command
python3 evaluate.py
```
You can find our checkpoint on HuggingFace Hub ([see this](https://huggingface.co/vasudevgupta/flax-bigbird-natural-questions)). In case you are interested in PyTorch BigBird fine-tuning, you can refer to [this repositary](https://github.com/thevasudevgupta/bigbird).
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/prepare_natural_questions.py
|
import os
import jsonlines
import numpy as np
from tqdm import tqdm
DOC_STRIDE = 2048
MAX_LENGTH = 4096
SEED = 42
PROCESS_TRAIN = os.environ.pop("PROCESS_TRAIN", "false")
CATEGORY_MAPPING = {"null": 0, "short": 1, "long": 2, "yes": 3, "no": 4}
def _get_single_answer(example):
def choose_first(answer, is_long_answer=False):
assert isinstance(answer, list)
if len(answer) == 1:
answer = answer[0]
return {k: [answer[k]] for k in answer} if is_long_answer else answer
for a in answer:
if is_long_answer:
a = {k: [a[k]] for k in a}
if len(a["start_token"]) > 0:
break
return a
answer = {"id": example["id"]}
annotation = example["annotations"]
yes_no_answer = annotation["yes_no_answer"]
if 0 in yes_no_answer or 1 in yes_no_answer:
answer["category"] = ["yes"] if 1 in yes_no_answer else ["no"]
answer["start_token"] = answer["end_token"] = []
answer["start_byte"] = answer["end_byte"] = []
answer["text"] = ["<cls>"]
else:
answer["category"] = ["short"]
out = choose_first(annotation["short_answers"])
if len(out["start_token"]) == 0:
# answer will be long if short is not available
answer["category"] = ["long"]
out = choose_first(annotation["long_answer"], is_long_answer=True)
out["text"] = []
answer.update(out)
# disregard some samples
if len(answer["start_token"]) > 1 or answer["start_token"] == answer["end_token"]:
answer["remove_it"] = True
else:
answer["remove_it"] = False
cols = ["start_token", "end_token", "start_byte", "end_byte", "text"]
if not all(isinstance(answer[k], list) for k in cols):
raise ValueError("Issue in ID", example["id"])
return answer
def get_context_and_ans(example, assertion=False):
"""Gives new context after removing <html> & new answer tokens as per new context"""
answer = _get_single_answer(example)
# bytes are of no use
del answer["start_byte"]
del answer["end_byte"]
# handle yes_no answers explicitly
if answer["category"][0] in ["yes", "no"]: # category is list with one element
doc = example["document"]["tokens"]
context = []
for i in range(len(doc["token"])):
if not doc["is_html"][i]:
context.append(doc["token"][i])
return {
"context": " ".join(context),
"answer": {
"start_token": -100, # ignore index in cross-entropy
"end_token": -100, # ignore index in cross-entropy
"category": answer["category"],
"span": answer["category"], # extra
},
}
# later, help in removing all no answers
if answer["start_token"] == [-1]:
return {
"context": "None",
"answer": {
"start_token": -1,
"end_token": -1,
"category": "null",
"span": "None", # extra
},
}
# handling normal samples
cols = ["start_token", "end_token"]
answer.update({k: answer[k][0] if len(answer[k]) > 0 else answer[k] for k in cols}) # e.g. [10] == 10
doc = example["document"]["tokens"]
start_token = answer["start_token"]
end_token = answer["end_token"]
context = []
for i in range(len(doc["token"])):
if not doc["is_html"][i]:
context.append(doc["token"][i])
else:
if answer["start_token"] > i:
start_token -= 1
if answer["end_token"] > i:
end_token -= 1
new = " ".join(context[start_token:end_token])
# checking above code
if assertion:
"""checking if above code is working as expected for all the samples"""
is_html = doc["is_html"][answer["start_token"] : answer["end_token"]]
old = doc["token"][answer["start_token"] : answer["end_token"]]
old = " ".join([old[i] for i in range(len(old)) if not is_html[i]])
if new != old:
print("ID:", example["id"])
print("New:", new, end="\n")
print("Old:", old, end="\n\n")
return {
"context": " ".join(context),
"answer": {
"start_token": start_token,
"end_token": end_token - 1, # this makes it inclusive
"category": answer["category"], # either long or short
"span": new, # extra
},
}
def get_strided_contexts_and_ans(example, tokenizer, doc_stride=2048, max_length=4096, assertion=True):
# overlap will be of doc_stride - q_len
out = get_context_and_ans(example, assertion=assertion)
answer = out["answer"]
# later, removing these samples
if answer["start_token"] == -1:
return {
"example_id": example["id"],
"input_ids": [[-1]],
"labels": {
"start_token": [-1],
"end_token": [-1],
"category": ["null"],
},
}
input_ids = tokenizer(example["question"]["text"], out["context"]).input_ids
q_len = input_ids.index(tokenizer.sep_token_id) + 1
# return yes/no
if answer["category"][0] in ["yes", "no"]: # category is list with one element
inputs = []
category = []
q_indices = input_ids[:q_len]
doc_start_indices = range(q_len, len(input_ids), max_length - doc_stride)
for i in doc_start_indices:
end_index = i + max_length - q_len
slice = input_ids[i:end_index]
inputs.append(q_indices + slice)
category.append(answer["category"][0])
if slice[-1] == tokenizer.sep_token_id:
break
return {
"example_id": example["id"],
"input_ids": inputs,
"labels": {
"start_token": [-100] * len(category),
"end_token": [-100] * len(category),
"category": category,
},
}
splitted_context = out["context"].split()
complete_end_token = splitted_context[answer["end_token"]]
answer["start_token"] = len(
tokenizer(
" ".join(splitted_context[: answer["start_token"]]),
add_special_tokens=False,
).input_ids
)
answer["end_token"] = len(
tokenizer(" ".join(splitted_context[: answer["end_token"]]), add_special_tokens=False).input_ids
)
answer["start_token"] += q_len
answer["end_token"] += q_len
# fixing end token
num_sub_tokens = len(tokenizer(complete_end_token, add_special_tokens=False).input_ids)
if num_sub_tokens > 1:
answer["end_token"] += num_sub_tokens - 1
old = input_ids[answer["start_token"] : answer["end_token"] + 1] # right & left are inclusive
start_token = answer["start_token"]
end_token = answer["end_token"]
if assertion:
"""This won't match exactly because of extra gaps => visaully inspect everything"""
new = tokenizer.decode(old)
if answer["span"] != new:
print("ISSUE IN TOKENIZATION")
print("OLD:", answer["span"])
print("NEW:", new, end="\n\n")
if len(input_ids) <= max_length:
return {
"example_id": example["id"],
"input_ids": [input_ids],
"labels": {
"start_token": [answer["start_token"]],
"end_token": [answer["end_token"]],
"category": answer["category"],
},
}
q_indices = input_ids[:q_len]
doc_start_indices = range(q_len, len(input_ids), max_length - doc_stride)
inputs = []
answers_start_token = []
answers_end_token = []
answers_category = [] # null, yes, no, long, short
for i in doc_start_indices:
end_index = i + max_length - q_len
slice = input_ids[i:end_index]
inputs.append(q_indices + slice)
assert len(inputs[-1]) <= max_length, "Issue in truncating length"
if start_token >= i and end_token <= end_index - 1:
start_token = start_token - i + q_len
end_token = end_token - i + q_len
answers_category.append(answer["category"][0]) # ["short"] -> "short"
else:
start_token = -100
end_token = -100
answers_category.append("null")
new = inputs[-1][start_token : end_token + 1]
answers_start_token.append(start_token)
answers_end_token.append(end_token)
if assertion:
"""checking if above code is working as expected for all the samples"""
if new != old and new != [tokenizer.cls_token_id]:
print("ISSUE in strided for ID:", example["id"])
print("New:", tokenizer.decode(new))
print("Old:", tokenizer.decode(old), end="\n\n")
if slice[-1] == tokenizer.sep_token_id:
break
return {
"example_id": example["id"],
"input_ids": inputs,
"labels": {
"start_token": answers_start_token,
"end_token": answers_end_token,
"category": answers_category,
},
}
def prepare_inputs(example, tokenizer, doc_stride=2048, max_length=4096, assertion=False):
example = get_strided_contexts_and_ans(
example,
tokenizer,
doc_stride=doc_stride,
max_length=max_length,
assertion=assertion,
)
return example
def save_to_disk(hf_data, file_name):
with jsonlines.open(file_name, "a") as writer:
for example in tqdm(hf_data, total=len(hf_data), desc="Saving samples ... "):
labels = example["labels"]
for ids, start, end, cat in zip(
example["input_ids"],
labels["start_token"],
labels["end_token"],
labels["category"],
):
if start == -1 and end == -1:
continue # leave waste samples with no answer
if cat == "null" and np.random.rand() < 0.6:
continue # removing 50 % samples
writer.write(
{
"input_ids": ids,
"start_token": start,
"end_token": end,
"category": CATEGORY_MAPPING[cat],
}
)
if __name__ == "__main__":
"""Running area"""
from datasets import load_dataset
from transformers import BigBirdTokenizer
data = load_dataset("natural_questions")
tokenizer = BigBirdTokenizer.from_pretrained("google/bigbird-roberta-base")
data = data["train" if PROCESS_TRAIN == "true" else "validation"]
fn_kwargs = {
"tokenizer": tokenizer,
"doc_stride": DOC_STRIDE,
"max_length": MAX_LENGTH,
"assertion": False,
}
data = data.map(prepare_inputs, fn_kwargs=fn_kwargs)
data = data.remove_columns(["annotations", "document", "id", "question"])
print(data)
np.random.seed(SEED)
cache_file_name = "nq-training.jsonl" if PROCESS_TRAIN == "true" else "nq-validation.jsonl"
save_to_disk(data, file_name=cache_file_name)
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/sweep_flax.yaml
|
command:
- python3
- train.py
method: random
parameters:
lr:
values: [4e-5, 3e-5]
warmup_steps:
values: [20000, 15000, 10000, 5000]
weight_decay:
distribution: normal
mu: 1e-2
sigma: 2e-3
metric:
name: eval_loss
goal: minimize
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/train.py
|
import os
from dataclasses import replace
import jax
import wandb
from bigbird_flax import Args, DataCollator, FlaxBigBirdForNaturalQuestions, Trainer, build_tx, train_step, val_step
from datasets import load_dataset
from flax import jax_utils
from transformers import BigBirdTokenizerFast
if __name__ == "__main__":
print("#################### AVAILABLE DEVICES ####################")
print(jax.devices())
print("###########################################################")
# setup for wandb sweep
args = Args()
logger = wandb.init(project="bigbird-natural-questions", config=args.__dict__)
wandb_args = dict(logger.config)
del wandb_args["batch_size"]
args = replace(args, **wandb_args)
base_dir = args.base_dir + "-" + wandb.run.id
args = replace(args, base_dir=base_dir)
print(args)
tr_dataset = load_dataset("json", data_files=args.tr_data_path)["train"]
val_dataset = load_dataset("json", data_files=args.val_data_path)["train"]
# drop extra batch for now
indices = range(len(tr_dataset) - len(tr_dataset) % args.batch_size)
tr_dataset = tr_dataset.shuffle().select(indices)
indices = range(len(val_dataset) - len(val_dataset) % args.batch_size)
val_dataset = val_dataset.shuffle().select(indices)
if os.environ.get("TRAIN_ON_SMALL", "false") == "true":
tr_dataset = tr_dataset.shuffle().select(range(80000))
val_dataset = val_dataset.shuffle().select(range(8000))
print(tr_dataset)
print(val_dataset)
model = FlaxBigBirdForNaturalQuestions.from_pretrained(
args.model_id, block_size=args.block_size, num_random_blocks=args.num_random_blocks
)
tokenizer = BigBirdTokenizerFast.from_pretrained(args.model_id)
data_collator = DataCollator(pad_id=tokenizer.pad_token_id, max_length=4096)
tx_args = {
"lr": args.lr,
"init_lr": args.init_lr,
"warmup_steps": args.warmup_steps,
"num_train_steps": args.max_epochs * (len(tr_dataset) // args.batch_size),
"weight_decay": args.weight_decay,
}
tx, lr = build_tx(**tx_args)
trainer = Trainer(
args=args,
data_collator=data_collator,
model_save_fn=model.save_pretrained,
train_step_fn=train_step,
val_step_fn=val_step,
logger=logger,
scheduler_fn=lr,
)
ckpt_dir = None
state = trainer.create_state(model, tx, num_train_steps=tx_args["num_train_steps"], ckpt_dir=ckpt_dir)
try:
trainer.train(state, tr_dataset, val_dataset)
except KeyboardInterrupt:
print("Oooops; TRAINING STOPPED UNFORTUNATELY")
print("SAVING WEIGHTS IN `final-weights`")
params = jax_utils.unreplicate(state.params)
model.save_pretrained(os.path.join(args.base_dir, "final-weights"), params=params)
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/bigbird_flax.py
|
import json
import os
from dataclasses import dataclass
from functools import partial
from typing import Callable
import flax.linen as nn
import jax
import jax.numpy as jnp
import joblib
import optax
import wandb
from flax import jax_utils, struct, traverse_util
from flax.serialization import from_bytes, to_bytes
from flax.training import train_state
from flax.training.common_utils import shard
from tqdm.auto import tqdm
from transformers import BigBirdConfig, FlaxBigBirdForQuestionAnswering
from transformers.models.big_bird.modeling_flax_big_bird import FlaxBigBirdForQuestionAnsweringModule
class FlaxBigBirdForNaturalQuestionsModule(FlaxBigBirdForQuestionAnsweringModule):
"""
BigBirdForQuestionAnswering with CLS Head over the top for predicting category
This way we can load its weights with FlaxBigBirdForQuestionAnswering
"""
config: BigBirdConfig
dtype: jnp.dtype = jnp.float32
add_pooling_layer: bool = True
def setup(self):
super().setup()
self.cls = nn.Dense(5, dtype=self.dtype)
def __call__(self, *args, **kwargs):
outputs = super().__call__(*args, **kwargs)
cls_out = self.cls(outputs[2])
return outputs[:2] + (cls_out,)
class FlaxBigBirdForNaturalQuestions(FlaxBigBirdForQuestionAnswering):
module_class = FlaxBigBirdForNaturalQuestionsModule
def calculate_loss_for_nq(start_logits, start_labels, end_logits, end_labels, pooled_logits, pooler_labels):
def cross_entropy(logits, labels, reduction=None):
"""
Args:
logits: bsz, seqlen, vocab_size
labels: bsz, seqlen
"""
vocab_size = logits.shape[-1]
labels = (labels[..., None] == jnp.arange(vocab_size)[None]).astype("f4")
logits = jax.nn.log_softmax(logits, axis=-1)
loss = -jnp.sum(labels * logits, axis=-1)
if reduction is not None:
loss = reduction(loss)
return loss
cross_entropy = partial(cross_entropy, reduction=jnp.mean)
start_loss = cross_entropy(start_logits, start_labels)
end_loss = cross_entropy(end_logits, end_labels)
pooled_loss = cross_entropy(pooled_logits, pooler_labels)
return (start_loss + end_loss + pooled_loss) / 3
@dataclass
class Args:
model_id: str = "google/bigbird-roberta-base"
logging_steps: int = 3000
save_steps: int = 10500
block_size: int = 128
num_random_blocks: int = 3
batch_size_per_device: int = 1
max_epochs: int = 5
# tx_args
lr: float = 3e-5
init_lr: float = 0.0
warmup_steps: int = 20000
weight_decay: float = 0.0095
save_dir: str = "bigbird-roberta-natural-questions"
base_dir: str = "training-expt"
tr_data_path: str = "data/nq-training.jsonl"
val_data_path: str = "data/nq-validation.jsonl"
def __post_init__(self):
os.makedirs(self.base_dir, exist_ok=True)
self.save_dir = os.path.join(self.base_dir, self.save_dir)
self.batch_size = self.batch_size_per_device * jax.device_count()
@dataclass
class DataCollator:
pad_id: int
max_length: int = 4096 # no dynamic padding on TPUs
def __call__(self, batch):
batch = self.collate_fn(batch)
batch = jax.tree_util.tree_map(shard, batch)
return batch
def collate_fn(self, features):
input_ids, attention_mask = self.fetch_inputs(features["input_ids"])
batch = {
"input_ids": jnp.array(input_ids, dtype=jnp.int32),
"attention_mask": jnp.array(attention_mask, dtype=jnp.int32),
"start_labels": jnp.array(features["start_token"], dtype=jnp.int32),
"end_labels": jnp.array(features["end_token"], dtype=jnp.int32),
"pooled_labels": jnp.array(features["category"], dtype=jnp.int32),
}
return batch
def fetch_inputs(self, input_ids: list):
inputs = [self._fetch_inputs(ids) for ids in input_ids]
return zip(*inputs)
def _fetch_inputs(self, input_ids: list):
attention_mask = [1 for _ in range(len(input_ids))]
while len(input_ids) < self.max_length:
input_ids.append(self.pad_id)
attention_mask.append(0)
return input_ids, attention_mask
def get_batched_dataset(dataset, batch_size, seed=None):
if seed is not None:
dataset = dataset.shuffle(seed=seed)
for i in range(len(dataset) // batch_size):
batch = dataset[i * batch_size : (i + 1) * batch_size]
yield dict(batch)
@partial(jax.pmap, axis_name="batch")
def train_step(state, drp_rng, **model_inputs):
def loss_fn(params):
start_labels = model_inputs.pop("start_labels")
end_labels = model_inputs.pop("end_labels")
pooled_labels = model_inputs.pop("pooled_labels")
outputs = state.apply_fn(**model_inputs, params=params, dropout_rng=drp_rng, train=True)
start_logits, end_logits, pooled_logits = outputs
return state.loss_fn(
start_logits,
start_labels,
end_logits,
end_labels,
pooled_logits,
pooled_labels,
)
drp_rng, new_drp_rng = jax.random.split(drp_rng)
grad_fn = jax.value_and_grad(loss_fn)
loss, grads = grad_fn(state.params)
metrics = jax.lax.pmean({"loss": loss}, axis_name="batch")
grads = jax.lax.pmean(grads, "batch")
state = state.apply_gradients(grads=grads)
return state, metrics, new_drp_rng
@partial(jax.pmap, axis_name="batch")
def val_step(state, **model_inputs):
start_labels = model_inputs.pop("start_labels")
end_labels = model_inputs.pop("end_labels")
pooled_labels = model_inputs.pop("pooled_labels")
outputs = state.apply_fn(**model_inputs, params=state.params, train=False)
start_logits, end_logits, pooled_logits = outputs
loss = state.loss_fn(start_logits, start_labels, end_logits, end_labels, pooled_logits, pooled_labels)
metrics = jax.lax.pmean({"loss": loss}, axis_name="batch")
return metrics
class TrainState(train_state.TrainState):
loss_fn: Callable = struct.field(pytree_node=False)
@dataclass
class Trainer:
args: Args
data_collator: Callable
train_step_fn: Callable
val_step_fn: Callable
model_save_fn: Callable
logger: wandb
scheduler_fn: Callable = None
def create_state(self, model, tx, num_train_steps, ckpt_dir=None):
params = model.params
state = TrainState.create(
apply_fn=model.__call__,
params=params,
tx=tx,
loss_fn=calculate_loss_for_nq,
)
if ckpt_dir is not None:
params, opt_state, step, args, data_collator = restore_checkpoint(ckpt_dir, state)
tx_args = {
"lr": args.lr,
"init_lr": args.init_lr,
"warmup_steps": args.warmup_steps,
"num_train_steps": num_train_steps,
"weight_decay": args.weight_decay,
}
tx, lr = build_tx(**tx_args)
state = train_state.TrainState(
step=step,
apply_fn=model.__call__,
params=params,
tx=tx,
opt_state=opt_state,
)
self.args = args
self.data_collator = data_collator
self.scheduler_fn = lr
model.params = params
state = jax_utils.replicate(state)
return state
def train(self, state, tr_dataset, val_dataset):
args = self.args
total = len(tr_dataset) // args.batch_size
rng = jax.random.PRNGKey(0)
drp_rng = jax.random.split(rng, jax.device_count())
for epoch in range(args.max_epochs):
running_loss = jnp.array(0, dtype=jnp.float32)
tr_dataloader = get_batched_dataset(tr_dataset, args.batch_size, seed=epoch)
i = 0
for batch in tqdm(tr_dataloader, total=total, desc=f"Running EPOCH-{epoch}"):
batch = self.data_collator(batch)
state, metrics, drp_rng = self.train_step_fn(state, drp_rng, **batch)
running_loss += jax_utils.unreplicate(metrics["loss"])
i += 1
if i % args.logging_steps == 0:
state_step = jax_utils.unreplicate(state.step)
tr_loss = running_loss.item() / i
lr = self.scheduler_fn(state_step - 1)
eval_loss = self.evaluate(state, val_dataset)
logging_dict = {
"step": state_step.item(),
"eval_loss": eval_loss.item(),
"tr_loss": tr_loss,
"lr": lr.item(),
}
tqdm.write(str(logging_dict))
self.logger.log(logging_dict, commit=True)
if i % args.save_steps == 0:
self.save_checkpoint(args.save_dir + f"-e{epoch}-s{i}", state=state)
def evaluate(self, state, dataset):
dataloader = get_batched_dataset(dataset, self.args.batch_size)
total = len(dataset) // self.args.batch_size
running_loss = jnp.array(0, dtype=jnp.float32)
i = 0
for batch in tqdm(dataloader, total=total, desc="Evaluating ... "):
batch = self.data_collator(batch)
metrics = self.val_step_fn(state, **batch)
running_loss += jax_utils.unreplicate(metrics["loss"])
i += 1
return running_loss / i
def save_checkpoint(self, save_dir, state):
state = jax_utils.unreplicate(state)
print(f"SAVING CHECKPOINT IN {save_dir}", end=" ... ")
self.model_save_fn(save_dir, params=state.params)
with open(os.path.join(save_dir, "opt_state.msgpack"), "wb") as f:
f.write(to_bytes(state.opt_state))
joblib.dump(self.args, os.path.join(save_dir, "args.joblib"))
joblib.dump(self.data_collator, os.path.join(save_dir, "data_collator.joblib"))
with open(os.path.join(save_dir, "training_state.json"), "w") as f:
json.dump({"step": state.step.item()}, f)
print("DONE")
def restore_checkpoint(save_dir, state):
print(f"RESTORING CHECKPOINT FROM {save_dir}", end=" ... ")
with open(os.path.join(save_dir, "flax_model.msgpack"), "rb") as f:
params = from_bytes(state.params, f.read())
with open(os.path.join(save_dir, "opt_state.msgpack"), "rb") as f:
opt_state = from_bytes(state.opt_state, f.read())
args = joblib.load(os.path.join(save_dir, "args.joblib"))
data_collator = joblib.load(os.path.join(save_dir, "data_collator.joblib"))
with open(os.path.join(save_dir, "training_state.json"), "r") as f:
training_state = json.load(f)
step = training_state["step"]
print("DONE")
return params, opt_state, step, args, data_collator
def scheduler_fn(lr, init_lr, warmup_steps, num_train_steps):
decay_steps = num_train_steps - warmup_steps
warmup_fn = optax.linear_schedule(init_value=init_lr, end_value=lr, transition_steps=warmup_steps)
decay_fn = optax.linear_schedule(init_value=lr, end_value=1e-7, transition_steps=decay_steps)
lr = optax.join_schedules(schedules=[warmup_fn, decay_fn], boundaries=[warmup_steps])
return lr
def build_tx(lr, init_lr, warmup_steps, num_train_steps, weight_decay):
def weight_decay_mask(params):
params = traverse_util.flatten_dict(params)
mask = {k: (v[-1] != "bias" and v[-2:] != ("LayerNorm", "scale")) for k, v in params.items()}
return traverse_util.unflatten_dict(mask)
lr = scheduler_fn(lr, init_lr, warmup_steps, num_train_steps)
tx = optax.adamw(learning_rate=lr, weight_decay=weight_decay, mask=weight_decay_mask)
return tx, lr
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/evaluate.py
|
import jax
import jax.numpy as jnp
from bigbird_flax import FlaxBigBirdForNaturalQuestions
from datasets import load_from_disk
from transformers import BigBirdTokenizerFast
CATEGORY_MAPPING = {0: "null", 1: "short", 2: "long", 3: "yes", 4: "no"}
PUNCTUATION_SET_TO_EXCLUDE = set("".join(["‘", "’", "´", "`", ".", ",", "-", '"']))
def get_sub_answers(answers, begin=0, end=None):
return [" ".join(x.split(" ")[begin:end]) for x in answers if len(x.split(" ")) > 1]
def expand_to_aliases(given_answers, make_sub_answers=False):
if make_sub_answers:
# if answers are longer than one word, make sure a predictions is correct if it coresponds to the complete 1: or :-1 sub word
# *e.g.* if the correct answer contains a prefix such as "the", or "a"
given_answers = (
given_answers + get_sub_answers(given_answers, begin=1) + get_sub_answers(given_answers, end=-1)
)
answers = []
for answer in given_answers:
alias = answer.replace("_", " ").lower()
alias = "".join(c if c not in PUNCTUATION_SET_TO_EXCLUDE else " " for c in alias)
answers.append(" ".join(alias.split()).strip())
return set(answers)
def get_best_valid_start_end_idx(start_scores, end_scores, top_k=1, max_size=100):
best_start_scores, best_start_idx = jax.lax.top_k(start_scores, top_k)
best_end_scores, best_end_idx = jax.lax.top_k(end_scores, top_k)
widths = best_end_idx[:, None] - best_start_idx[None, :]
mask = jnp.logical_or(widths < 0, widths > max_size)
scores = (best_end_scores[:, None] + best_start_scores[None, :]) - (1e8 * mask)
best_score = jnp.argmax(scores).item()
return best_start_idx[best_score % top_k], best_end_idx[best_score // top_k]
def format_dataset(sample):
question = sample["question"]["text"]
context = sample["document"]["tokens"]["token"]
is_html = sample["document"]["tokens"]["is_html"]
long_answers = sample["annotations"]["long_answer"]
short_answers = sample["annotations"]["short_answers"]
context_string = " ".join([context[i] for i in range(len(context)) if not is_html[i]])
# 0 - No ; 1 - Yes
for answer in sample["annotations"]["yes_no_answer"]:
if answer == 0 or answer == 1:
return {
"question": question,
"context": context_string,
"short": [],
"long": [],
"category": "no" if answer == 0 else "yes",
}
short_targets = []
for s in short_answers:
short_targets.extend(s["text"])
short_targets = list(set(short_targets))
long_targets = []
for s in long_answers:
if s["start_token"] == -1:
continue
answer = context[s["start_token"] : s["end_token"]]
html = is_html[s["start_token"] : s["end_token"]]
new_answer = " ".join([answer[i] for i in range(len(answer)) if not html[i]])
if new_answer not in long_targets:
long_targets.append(new_answer)
category = "long_short" if len(short_targets + long_targets) > 0 else "null"
return {
"question": question,
"context": context_string,
"short": short_targets,
"long": long_targets,
"category": category,
}
def main():
dataset = load_from_disk("natural-questions-validation")
dataset = dataset.map(format_dataset).remove_columns(["annotations", "document", "id"])
print(dataset)
short_validation_dataset = dataset.filter(lambda x: (len(x["question"]) + len(x["context"])) < 4 * 4096)
short_validation_dataset = short_validation_dataset.filter(lambda x: x["category"] != "null")
short_validation_dataset
model_id = "vasudevgupta/flax-bigbird-natural-questions"
model = FlaxBigBirdForNaturalQuestions.from_pretrained(model_id)
tokenizer = BigBirdTokenizerFast.from_pretrained(model_id)
@jax.jit
def forward(*args, **kwargs):
start_logits, end_logits, pooled_logits = model(*args, **kwargs)
return start_logits, end_logits, jnp.argmax(pooled_logits, axis=-1)
def evaluate(example):
# encode question and context so that they are separated by a tokenizer.sep_token and cut at max_length
inputs = tokenizer(
example["question"],
example["context"],
return_tensors="np",
max_length=4096,
padding="max_length",
truncation=True,
)
start_scores, end_scores, category = forward(**inputs)
predicted_category = CATEGORY_MAPPING[category.item()]
example["targets"] = example["long"] + example["short"]
if example["category"] in ["yes", "no", "null"]:
example["targets"] = [example["category"]]
example["has_tgt"] = example["category"] != "null"
# Now target can be: "yes", "no", "null", "list of long & short answers"
if predicted_category in ["yes", "no", "null"]:
example["output"] = [predicted_category]
example["match"] = example["output"] == example["targets"]
example["has_pred"] = predicted_category != "null"
return example
max_size = 38 if predicted_category == "short" else 1024
start_score, end_score = get_best_valid_start_end_idx(
start_scores[0], end_scores[0], top_k=8, max_size=max_size
)
input_ids = inputs["input_ids"][0].tolist()
example["output"] = [tokenizer.decode(input_ids[start_score : end_score + 1])]
answers = expand_to_aliases(example["targets"], make_sub_answers=True)
predictions = expand_to_aliases(example["output"])
# some preprocessing to both prediction and answer
answers = {"".join(a.split()) for a in answers}
predictions = {"".join(p.split()) for p in predictions}
predictions = {s for s in predictions if s not in ["``", "''", "`", "'"]}
# if there is a common element, it's a exact match
example["match"] = len(list(answers & predictions)) > 0
example["has_pred"] = predicted_category != "null" and len(predictions) > 0
return example
short_validation_dataset = short_validation_dataset.map(evaluate)
total = len(short_validation_dataset)
matched = len(short_validation_dataset.filter(lambda x: x["match"] == 1))
print("EM score:", (matched / total) * 100, "%")
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/jax-projects
|
hf_public_repos/transformers/examples/research_projects/jax-projects/big_bird/requirements.txt
|
git+https://github.com/huggingface/transformers@main
datasets
sentencepiece
wandb
flax
jsonlines
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/longform-qa/README.md
|
# Long Form Question Answering
Author: @yjernite
This folder contains the code for the Long Form Question answering [demo](http://35.226.96.115:8080/) as well as methods to train and use a fully end-to-end Long Form Question Answering system using the [🤗transformers](https://github.com/huggingface/transformers) and [🤗datasets](https://github.com/huggingface/datasets) libraries.
You can use these methods to train your own system by following along the associate [notebook](https://github.com/huggingface/notebooks/blob/master/longform-qa/Long_Form_Question_Answering_with_ELI5_and_Wikipedia.ipynb) or [blog post](https://yjernite.github.io/lfqa.html).
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/longform-qa/eli5_utils.py
|
import functools
import math
import os # noqa: F401
from random import choice, randint
from time import time
import datasets # noqa: F401
import faiss # noqa: F401
import numpy as np
import pandas as pd
import torch
import torch.utils.checkpoint as checkpoint
from elasticsearch import Elasticsearch # noqa: F401
from elasticsearch.helpers import bulk, streaming_bulk # noqa: F401
from torch import nn
from torch.utils.data import DataLoader, Dataset, RandomSampler, SequentialSampler
from tqdm import tqdm
from transformers import AdamW, AutoModel, AutoModelForSeq2SeqLM, AutoTokenizer, get_linear_schedule_with_warmup
pd.set_option("display.max_colwidth", None)
###############
# Sparse index
###############
def make_es_index_snippets(es_client, passages_dset, index_name="english_wiki_kilt_snippets_100w"):
index_config = {
"settings": {
"number_of_shards": 1,
"analysis": {"analyzer": {"stop_standard": {"type": "standard", " stopwords": "_english_"}}},
},
"mappings": {
"properties": {
"article_title": {"type": "text", "analyzer": "standard", "similarity": "BM25"},
"section_title": {"type": "text", "analyzer": "standard", "similarity": "BM25"},
"passage_text": {"type": "text", "analyzer": "standard", "similarity": "BM25"},
}
},
}
es_client.indices.create(index=index_name, body=index_config)
number_of_docs = passages_dset.num_rows
progress = tqdm(unit="docs", total=number_of_docs)
successes = 0
def passage_generator():
for passage in passages_dset:
yield passage
# create the ES index
for ok, action in streaming_bulk(
client=es_client,
index=index_name,
actions=passage_generator(),
):
progress.update(1)
successes += ok
print("Indexed %d documents" % (successes,))
def query_es_index(question, es_client, index_name="english_wiki_kilt_snippets_100w", n_results=10, min_length=20):
q = question.lower()
banned = ["how", "why", "what", "where", "which", "do", "does", "is", "?", "eli5", "eli5:"]
q = " ".join([w for w in q.split() if w not in banned])
response = es_client.search(
index=index_name,
body={
"query": {
"multi_match": {
"query": q,
"fields": ["article_title", "section_title", "passage_text^2"],
"type": "cross_fields",
}
},
"size": 2 * n_results,
},
)
hits = response["hits"]["hits"]
support_doc = "<P> " + " <P> ".join([hit["_source"]["passage_text"] for hit in hits])
res_list = [{k: hit["_source"][k] for k in hit["_source"] if k != "passage_text"} for hit in hits]
for r, hit in zip(res_list, hits):
r["passage_id"] = hit["_id"]
r["score"] = hit["_score"]
r["passage_text"] = hit["_source"]["passage_text"]
res_list = [res for res in res_list if len(res["passage_text"].split()) > min_length][:n_results]
return support_doc, res_list
###############
# ELI5 retriever training
###############
class ELI5DatasetQARetriver(Dataset):
def __init__(self, examples_array, extra_answer_threshold=3, min_answer_length=64, training=True, n_samples=None):
self.data = examples_array
self.answer_thres = extra_answer_threshold
self.min_length = min_answer_length
self.training = training
self.n_samples = self.data.num_rows if n_samples is None else n_samples
def __len__(self):
return self.n_samples
def make_example(self, idx):
example = self.data[idx]
question = example["title"]
if self.training:
answers = [a for i, (a, sc) in enumerate(zip(example["answers"]["text"], example["answers"]["score"]))]
answer_tab = choice(answers).split(" ")
start_idx = randint(0, max(0, len(answer_tab) - self.min_length))
answer_span = " ".join(answer_tab[start_idx:])
else:
answer_span = example["answers"]["text"][0]
return (question, answer_span)
def __getitem__(self, idx):
return self.make_example(idx % self.data.num_rows)
class RetrievalQAEmbedder(nn.Module):
def __init__(self, sent_encoder, dim):
super(RetrievalQAEmbedder, self).__init__()
self.sent_encoder = sent_encoder
self.output_dim = 128
self.project_q = nn.Linear(dim, self.output_dim, bias=False)
self.project_a = nn.Linear(dim, self.output_dim, bias=False)
self.ce_loss = nn.CrossEntropyLoss(reduction="mean")
def embed_sentences_checkpointed(self, input_ids, attention_mask, checkpoint_batch_size=-1):
# reproduces BERT forward pass with checkpointing
if checkpoint_batch_size < 0 or input_ids.shape[0] < checkpoint_batch_size:
return self.sent_encoder(input_ids, attention_mask=attention_mask)[1]
else:
# prepare implicit variables
device = input_ids.device
input_shape = input_ids.size()
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
head_mask = [None] * self.sent_encoder.config.num_hidden_layers
extended_attention_mask: torch.Tensor = self.sent_encoder.get_extended_attention_mask(
attention_mask, input_shape
)
# define function for checkpointing
def partial_encode(*inputs):
encoder_outputs = self.sent_encoder.encoder(
inputs[0],
attention_mask=inputs[1],
head_mask=head_mask,
)
sequence_output = encoder_outputs[0]
pooled_output = self.sent_encoder.pooler(sequence_output)
return pooled_output
# run embedding layer on everything at once
embedding_output = self.sent_encoder.embeddings(
input_ids=input_ids, position_ids=None, token_type_ids=token_type_ids, inputs_embeds=None
)
# run encoding and pooling on one mini-batch at a time
pooled_output_list = []
for b in range(math.ceil(input_ids.shape[0] / checkpoint_batch_size)):
b_embedding_output = embedding_output[b * checkpoint_batch_size : (b + 1) * checkpoint_batch_size]
b_attention_mask = extended_attention_mask[b * checkpoint_batch_size : (b + 1) * checkpoint_batch_size]
pooled_output = checkpoint.checkpoint(partial_encode, b_embedding_output, b_attention_mask)
pooled_output_list.append(pooled_output)
return torch.cat(pooled_output_list, dim=0)
def embed_questions(self, q_ids, q_mask, checkpoint_batch_size=-1):
q_reps = self.embed_sentences_checkpointed(q_ids, q_mask, checkpoint_batch_size)
return self.project_q(q_reps)
def embed_answers(self, a_ids, a_mask, checkpoint_batch_size=-1):
a_reps = self.embed_sentences_checkpointed(a_ids, a_mask, checkpoint_batch_size)
return self.project_a(a_reps)
def forward(self, q_ids, q_mask, a_ids, a_mask, checkpoint_batch_size=-1):
device = q_ids.device
q_reps = self.embed_questions(q_ids, q_mask, checkpoint_batch_size)
a_reps = self.embed_answers(a_ids, a_mask, checkpoint_batch_size)
compare_scores = torch.mm(q_reps, a_reps.t())
loss_qa = self.ce_loss(compare_scores, torch.arange(compare_scores.shape[1]).to(device))
loss_aq = self.ce_loss(compare_scores.t(), torch.arange(compare_scores.shape[0]).to(device))
loss = (loss_qa + loss_aq) / 2
return loss
def make_qa_retriever_model(model_name="google/bert_uncased_L-8_H-512_A-8", from_file=None, device="cuda:0"):
tokenizer = AutoTokenizer.from_pretrained(model_name)
bert_model = AutoModel.from_pretrained(model_name).to(device)
# run bert_model on a dummy batch to get output dimension
d_ids = torch.LongTensor(
[[bert_model.config.bos_token_id if bert_model.config.bos_token_id is not None else 1]]
).to(device)
d_mask = torch.LongTensor([[1]]).to(device)
sent_dim = bert_model(d_ids, attention_mask=d_mask)[1].shape[-1]
qa_embedder = RetrievalQAEmbedder(bert_model, sent_dim).to(device)
if from_file is not None:
param_dict = torch.load(from_file) # has model weights, optimizer, and scheduler states
qa_embedder.load_state_dict(param_dict["model"])
return tokenizer, qa_embedder
def make_qa_retriever_batch(qa_list, tokenizer, max_len=64, device="cuda:0"):
q_ls = [q for q, a in qa_list]
a_ls = [a for q, a in qa_list]
q_toks = tokenizer(q_ls, max_length=max_len, padding="max_length", truncation=True)
q_ids, q_mask = (
torch.LongTensor(q_toks["input_ids"]).to(device),
torch.LongTensor(q_toks["attention_mask"]).to(device),
)
a_toks = tokenizer(a_ls, max_length=max_len, padding="max_length", truncation=True)
a_ids, a_mask = (
torch.LongTensor(a_toks["input_ids"]).to(device),
torch.LongTensor(a_toks["attention_mask"]).to(device),
)
return (q_ids, q_mask, a_ids, a_mask)
def train_qa_retriever_epoch(model, dataset, tokenizer, optimizer, scheduler, args, e=0):
model.train()
# make iterator
train_sampler = RandomSampler(dataset)
model_collate_fn = functools.partial(
make_qa_retriever_batch, tokenizer=tokenizer, max_len=args.max_length, device="cuda:0"
)
data_loader = DataLoader(dataset, batch_size=args.batch_size, sampler=train_sampler, collate_fn=model_collate_fn)
epoch_iterator = tqdm(data_loader, desc="Iteration", disable=True)
# accumulate loss since last print
loc_steps = 0
loc_loss = 0.0
st_time = time()
for step, batch in enumerate(epoch_iterator):
q_ids, q_mask, a_ids, a_mask = batch
pre_loss = model(q_ids, q_mask, a_ids, a_mask, checkpoint_batch_size=args.checkpoint_batch_size)
loss = pre_loss.sum()
# optimizer
loss.backward()
optimizer.step()
scheduler.step()
model.zero_grad()
# some printing within the epoch
loc_loss += loss.item()
loc_steps += 1
if step % args.print_freq == 0 or step == 1:
print(
"{:2d} {:5d} of {:5d} \t L: {:.3f} \t -- {:.3f}".format(
e,
step,
len(dataset) // args.batch_size,
loc_loss / loc_steps,
time() - st_time,
)
)
loc_loss = 0
loc_steps = 0
def train_qa_retriever_joint_epoch(model, dataset_list, tokenizer, optimizer, scheduler, args, e=0):
model.train()
model_collate_fn = functools.partial(
make_qa_retriever_batch, tokenizer=tokenizer, max_len=args.max_length, device="cuda:0"
)
# make iterator
train_samplers = [RandomSampler(dataset) for dataset in dataset_list]
data_loaders = [
DataLoader(dataset, batch_size=args.batch_size, sampler=train_sampler, collate_fn=model_collate_fn)
for dataset, train_sampler in zip(dataset_list, train_samplers)
]
iterators = [iter(dloader) for dloader in data_loaders]
joint_iter = zip(*iterators)
# accumulate loss since last print
loc_steps = 0
loc_loss = 0.0
st_time = time()
for step, (batches,) in enumerate(zip(joint_iter)):
for batch in batches:
q_ids, q_mask, a_ids, a_mask = batch
loss = model(q_ids, q_mask, a_ids, a_mask, checkpoint_batch_size=args.checkpoint_batch_size)
# optimizer
loss.backward()
optimizer.step()
scheduler.step()
model.zero_grad()
# some printing within the epoch
loc_loss += loss.item()
loc_steps += 1
if step % args.print_freq == 0:
print(
"{:2d} {:5d} of {:5d} \t L: {:.3f} \t -- {:.3f}".format(
e,
step,
len(dataset_list[0]) // args.batch_size,
loc_loss / loc_steps,
time() - st_time,
)
)
loc_loss = 0
loc_steps = 0
def evaluate_qa_retriever(model, dataset, tokenizer, args):
model.eval()
# make iterator
eval_sampler = SequentialSampler(dataset)
model_collate_fn = functools.partial(
make_qa_retriever_batch, tokenizer=tokenizer, max_len=args.max_length, device="cuda:0"
)
data_loader = DataLoader(dataset, batch_size=args.batch_size, sampler=eval_sampler, collate_fn=model_collate_fn)
epoch_iterator = tqdm(data_loader, desc="Iteration", disable=True)
tot_loss = 0.0
with torch.no_grad():
for step, batch in enumerate(epoch_iterator):
q_ids, q_mask, a_ids, a_mask = batch
loss = model(q_ids, q_mask, a_ids, a_mask)
tot_loss += loss.item()
return tot_loss / (step + 1)
def train_qa_retriever(qar_model, qar_tokenizer, qar_train_dset, qar_valid_dset, qar_args):
qar_optimizer = AdamW(qar_model.parameters(), lr=qar_args.learning_rate, eps=1e-8)
qar_scheduler = get_linear_schedule_with_warmup(
qar_optimizer,
num_warmup_steps=100,
num_training_steps=(qar_args.num_epochs + 1) * math.ceil(len(qar_train_dset) / qar_args.batch_size),
)
for e in range(qar_args.num_epochs):
train_qa_retriever_epoch(qar_model, qar_train_dset, qar_tokenizer, qar_optimizer, qar_scheduler, qar_args, e)
m_save_dict = {
"model": qar_model.state_dict(),
"optimizer": qar_optimizer.state_dict(),
"scheduler": qar_scheduler.state_dict(),
}
print("Saving model {}".format(qar_args.model_save_name))
torch.save(m_save_dict, "{}_{}.pth".format(qar_args.model_save_name, e))
eval_loss = evaluate_qa_retriever(qar_model, qar_valid_dset, qar_tokenizer, qar_args)
print("Evaluation loss epoch {:4d}: {:.3f}".format(e, eval_loss))
###############
# ELI5 seq2seq model training
###############
class ELI5DatasetS2S(Dataset):
def __init__(
self, examples_array, make_doc_fun=None, extra_answer_threshold=3, document_cache=None, training=True
):
self.training = training
self.data = examples_array
self.make_doc_function = make_doc_fun
self.document_cache = {} if document_cache is None else document_cache
assert not (make_doc_fun is None and document_cache is None)
# make index of specific question-answer pairs from multi-answers
if self.training:
self.qa_id_list = [
(i, j)
for i, qa in enumerate(self.data)
for j, (a, sc) in enumerate(zip(qa["answers"]["text"], qa["answers"]["score"]))
if j == 0 or sc >= extra_answer_threshold
]
else:
self.qa_id_list = [(i, 0) for i in range(self.data.num_rows)]
def __len__(self):
return len(self.qa_id_list)
def make_example(self, idx):
i, j = self.qa_id_list[idx]
example = self.data[i]
question = example["title"] + " " + example["selftext"]
answer = example["answers"]["text"][j]
q_id = example["q_id"]
if self.make_doc_function is not None:
self.document_cache[q_id] = self.document_cache.get(q_id, self.make_doc_function(example["title"]))
document = self.document_cache[q_id]
in_st = "question: {} context: {}".format(
question.lower().replace(" --t--", "").strip(),
document.lower().strip(),
)
out_st = answer
return (in_st, out_st)
def __getitem__(self, idx):
return self.make_example(idx)
def make_qa_s2s_model(model_name="facebook/bart-large", from_file=None, device="cuda:0"):
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name).to(device)
if from_file is not None:
param_dict = torch.load(from_file) # has model weights, optimizer, and scheduler states
model.load_state_dict(param_dict["model"])
return tokenizer, model
def make_qa_s2s_batch(qa_list, tokenizer, max_len=64, max_a_len=360, device="cuda:0"):
q_ls = [q for q, a in qa_list]
a_ls = [a for q, a in qa_list]
q_toks = tokenizer(q_ls, max_length=max_len, padding="max_length", truncation=True)
q_ids, q_mask = (
torch.LongTensor(q_toks["input_ids"]).to(device),
torch.LongTensor(q_toks["attention_mask"]).to(device),
)
a_toks = tokenizer(a_ls, max_length=min(max_len, max_a_len), padding="max_length", truncation=True)
a_ids, a_mask = (
torch.LongTensor(a_toks["input_ids"]).to(device),
torch.LongTensor(a_toks["attention_mask"]).to(device),
)
lm_labels = a_ids[:, 1:].contiguous().clone()
lm_labels[a_mask[:, 1:].contiguous() == 0] = -100
model_inputs = {
"input_ids": q_ids,
"attention_mask": q_mask,
"decoder_input_ids": a_ids[:, :-1].contiguous(),
"lm_labels": lm_labels,
}
return model_inputs
def train_qa_s2s_epoch(model, dataset, tokenizer, optimizer, scheduler, args, e=0, curriculum=False):
model.train()
# make iterator
if curriculum:
train_sampler = SequentialSampler(dataset)
else:
train_sampler = RandomSampler(dataset)
model_collate_fn = functools.partial(
make_qa_s2s_batch, tokenizer=tokenizer, max_len=args.max_length, device="cuda:0"
)
data_loader = DataLoader(dataset, batch_size=args.batch_size, sampler=train_sampler, collate_fn=model_collate_fn)
epoch_iterator = tqdm(data_loader, desc="Iteration", disable=True)
# accumulate loss since last print
loc_steps = 0
loc_loss = 0.0
st_time = time()
for step, batch_inputs in enumerate(epoch_iterator):
pre_loss = model(**batch_inputs)[0]
loss = pre_loss.sum() / pre_loss.shape[0]
loss.backward()
# optimizer
if step % args.backward_freq == 0:
optimizer.step()
scheduler.step()
model.zero_grad()
# some printing within the epoch
loc_loss += loss.item()
loc_steps += 1
if step % args.print_freq == 0 or step == 1:
print(
"{:2d} {:5d} of {:5d} \t L: {:.3f} \t -- {:.3f}".format(
e,
step,
len(dataset) // args.batch_size,
loc_loss / loc_steps,
time() - st_time,
)
)
loc_loss = 0
loc_steps = 0
def eval_qa_s2s_epoch(model, dataset, tokenizer, args):
model.eval()
# make iterator
train_sampler = SequentialSampler(dataset)
model_collate_fn = functools.partial(
make_qa_s2s_batch, tokenizer=tokenizer, max_len=args.max_length, device="cuda:0"
)
data_loader = DataLoader(dataset, batch_size=args.batch_size, sampler=train_sampler, collate_fn=model_collate_fn)
epoch_iterator = tqdm(data_loader, desc="Iteration", disable=True)
# accumulate loss since last print
loc_steps = 0
loc_loss = 0.0
st_time = time()
with torch.no_grad():
for step, batch_inputs in enumerate(epoch_iterator):
pre_loss = model(**batch_inputs)[0]
loss = pre_loss.sum() / pre_loss.shape[0]
loc_loss += loss.item()
loc_steps += 1
if step % args.print_freq == 0:
print(
"{:5d} of {:5d} \t L: {:.3f} \t -- {:.3f}".format(
step,
len(dataset) // args.batch_size,
loc_loss / loc_steps,
time() - st_time,
)
)
print(
"Total \t L: {:.3f} \t -- {:.3f}".format(
loc_loss / loc_steps,
time() - st_time,
)
)
def train_qa_s2s(qa_s2s_model, qa_s2s_tokenizer, s2s_train_dset, s2s_valid_dset, s2s_args):
s2s_optimizer = AdamW(qa_s2s_model.parameters(), lr=s2s_args.learning_rate, eps=1e-8)
s2s_scheduler = get_linear_schedule_with_warmup(
s2s_optimizer,
num_warmup_steps=400,
num_training_steps=(s2s_args.num_epochs + 1) * math.ceil(len(s2s_train_dset) / s2s_args.batch_size),
)
for e in range(s2s_args.num_epochs):
train_qa_s2s_epoch(
qa_s2s_model,
s2s_train_dset,
qa_s2s_tokenizer,
s2s_optimizer,
s2s_scheduler,
s2s_args,
e,
curriculum=(e == 0),
)
m_save_dict = {
"model": qa_s2s_model.state_dict(),
"optimizer": s2s_optimizer.state_dict(),
"scheduler": s2s_scheduler.state_dict(),
}
print("Saving model {}".format(s2s_args.model_save_name))
eval_qa_s2s_epoch(qa_s2s_model, s2s_valid_dset, qa_s2s_tokenizer, s2s_args)
torch.save(m_save_dict, "{}_{}.pth".format(s2s_args.model_save_name, e))
# generate answer from input "question: ... context: <p> ..."
def qa_s2s_generate(
question_doc,
qa_s2s_model,
qa_s2s_tokenizer,
num_answers=1,
num_beams=None,
min_len=64,
max_len=256,
do_sample=False,
temp=1.0,
top_p=None,
top_k=None,
max_input_length=512,
device="cuda:0",
):
model_inputs = make_qa_s2s_batch(
[(question_doc, "A")],
qa_s2s_tokenizer,
max_input_length,
device=device,
)
n_beams = num_answers if num_beams is None else max(num_beams, num_answers)
generated_ids = qa_s2s_model.generate(
input_ids=model_inputs["input_ids"],
attention_mask=model_inputs["attention_mask"],
min_length=min_len,
max_length=max_len,
do_sample=do_sample,
early_stopping=True,
num_beams=1 if do_sample else n_beams,
temperature=temp,
top_k=top_k,
top_p=top_p,
eos_token_id=qa_s2s_tokenizer.eos_token_id,
no_repeat_ngram_size=3,
num_return_sequences=num_answers,
decoder_start_token_id=qa_s2s_tokenizer.bos_token_id,
)
return [qa_s2s_tokenizer.decode(ans_ids, skip_special_tokens=True).strip() for ans_ids in generated_ids]
###############
# ELI5-trained retrieval model usage
###############
def embed_passages_for_retrieval(passages, tokenizer, qa_embedder, max_length=128, device="cuda:0"):
a_toks = tokenizer(passages, max_length=max_length, padding="max_length", truncation=True)
a_ids, a_mask = (
torch.LongTensor(a_toks["input_ids"]).to(device),
torch.LongTensor(a_toks["attention_mask"]).to(device),
)
with torch.no_grad():
a_reps = qa_embedder.embed_answers(a_ids, a_mask).cpu().type(torch.float)
return a_reps.numpy()
def embed_questions_for_retrieval(q_ls, tokenizer, qa_embedder, device="cuda:0"):
q_toks = tokenizer(q_ls, max_length=128, padding="max_length", truncation=True)
q_ids, q_mask = (
torch.LongTensor(q_toks["input_ids"]).to(device),
torch.LongTensor(q_toks["attention_mask"]).to(device),
)
with torch.no_grad():
q_reps = qa_embedder.embed_questions(q_ids, q_mask).cpu().type(torch.float)
return q_reps.numpy()
def make_qa_dense_index(
qa_embedder,
tokenizer,
passages_dset,
batch_size=512,
max_length=128,
index_name="kilt_passages_reps.dat",
dtype="float32",
device="cuda:0",
):
st_time = time()
fp = np.memmap(index_name, dtype=dtype, mode="w+", shape=(passages_dset.num_rows, 128))
n_batches = math.ceil(passages_dset.num_rows / batch_size)
for i in range(n_batches):
passages = list(passages_dset[i * batch_size : (i + 1) * batch_size]["passage_text"])
reps = embed_passages_for_retrieval(passages, tokenizer, qa_embedder, max_length, device)
fp[i * batch_size : (i + 1) * batch_size] = reps
if i % 50 == 0:
print(i, time() - st_time)
def evaluate_retriever(qa_list, retriever_func, scoring_func, n_ret=10, verbose=False):
total_retriever_time = 0.0
total_retriever_score = 0.0
st_time = time()
for i, (question, answer) in enumerate(qa_list):
r_time = time()
retrieved_passages = retriever_func(question, n_ret)
total_retriever_time += time() - r_time
total_retriever_score += scoring_func(retrieved_passages, answer)
if verbose and ((i + 1) % 500 == 0 or i <= 1):
print(
"{:03d}: S-{:.4f} T-{:.4f} | {:.2f}".format(
i + 1, total_retriever_score / (i + 1), total_retriever_time / (i + 1), time() - st_time
)
)
return {"idf_recall": total_retriever_score / (i + 1), "retrieval_time": total_retriever_time / (i + 1)}
# build a support document for the question out of Wikipedia snippets
def query_qa_dense_index(
question, qa_embedder, tokenizer, wiki_passages, wiki_index, n_results=10, min_length=20, device="cuda:0"
):
q_rep = embed_questions_for_retrieval([question], tokenizer, qa_embedder, device=device)
D, I = wiki_index.search(q_rep, 2 * n_results)
res_passages = [wiki_passages[int(i)] for i in I[0]]
support_doc = "<P> " + " <P> ".join([p["passage_text"] for p in res_passages])
res_list = [{k: p[k] for k in wiki_passages.column_names} for p in res_passages]
res_list = [res for res in res_list if len(res["passage_text"].split()) > min_length][:n_results]
for r, sc in zip(res_list, D[0]):
r["score"] = float(sc)
return support_doc, res_list
def batch_query_qa_dense_index(questions, qa_embedder, tokenizer, wiki_passages, wiki_index, n_results=10):
q_rep = embed_questions_for_retrieval(questions, tokenizer, qa_embedder)
D, I = wiki_index.search(q_rep, n_results)
res_passages_lst = [[wiki_passages[int(i)] for i in i_lst] for i_lst in I]
support_doc_lst = [
"<P> " + " <P> ".join([p["passage_text"] for p in res_passages]) for res_passages in res_passages_lst
]
all_res_lists = []
for res_passages, dl in zip(res_passages_lst, D):
res_list = [{k: p[k] for k in wiki_passages.column_names} for p in res_passages]
for r, sc in zip(res_list, dl):
r["score"] = float(sc)
all_res_lists += [res_list[:]]
return support_doc_lst, all_res_lists
# find nearest neighbors of an answer or declarative text in Wikipedia snippets
def query_qa_dense_index_nn(passage, qa_embedder, tokenizer, wiki_passages, wiki_index, n_results=10, min_length=20):
a_rep = embed_passages_for_retrieval([passage], tokenizer, qa_embedder)
D, I = wiki_index.search(a_rep, 2 * n_results)
res_passages = [wiki_passages[int(i)] for i in I[0]]
support_doc = "<P> " + " <P> ".join([p["passage_text"] for p in res_passages])
res_list = [{k: p[k] for k in wiki_passages.column_names} for p in res_passages]
res_list = [res for res in res_list if len(res["passage_text"].split()) > min_length][:n_results]
for r, sc, i in zip(res_list, D[0], I[0]):
r["passage_id"] = int(i)
r["score"] = float(sc)
return support_doc, res_list
def batch_query_qa_dense_index_nn(passages, qa_embedder, tokenizer, wiki_passages, wiki_index, n_results=10):
a_reps = embed_passages_for_retrieval(passages, tokenizer, qa_embedder)
D, I = wiki_index.search(a_reps, n_results)
res_passages_lst = [[wiki_passages[int(i)] for i in i_lst] for i_lst in I]
support_doc_lst = [
"<P> " + " <P> ".join([p["passage_text"] for p in res_passages]) for res_passages in res_passages_lst
]
all_res_lists = []
for res_passages, dl, il in zip(res_passages_lst, D, I):
res_list = [{k: p[k] for k in wiki_passages.column_names} for p in res_passages]
for r, sc, i in zip(res_list, dl, il):
r["passage_id"] = int(i)
r["score"] = float(sc)
all_res_lists += [res_list[:]]
return support_doc_lst, all_res_lists
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/longform-qa/eli5_app.py
|
import datasets
import faiss
import numpy as np
import streamlit as st
import torch
from elasticsearch import Elasticsearch
from eli5_utils import (
embed_questions_for_retrieval,
make_qa_s2s_model,
qa_s2s_generate,
query_es_index,
query_qa_dense_index,
)
import transformers
from transformers import AutoModel, AutoModelForSeq2SeqLM, AutoTokenizer
MODEL_TYPE = "bart"
LOAD_DENSE_INDEX = True
@st.cache(allow_output_mutation=True)
def load_models():
if LOAD_DENSE_INDEX:
qar_tokenizer = AutoTokenizer.from_pretrained("yjernite/retribert-base-uncased")
qar_model = AutoModel.from_pretrained("yjernite/retribert-base-uncased").to("cuda:0")
_ = qar_model.eval()
else:
qar_tokenizer, qar_model = (None, None)
if MODEL_TYPE == "bart":
s2s_tokenizer = AutoTokenizer.from_pretrained("yjernite/bart_eli5")
s2s_model = AutoModelForSeq2SeqLM.from_pretrained("yjernite/bart_eli5").to("cuda:0")
save_dict = torch.load("seq2seq_models/eli5_bart_model_blm_2.pth")
s2s_model.load_state_dict(save_dict["model"])
_ = s2s_model.eval()
else:
s2s_tokenizer, s2s_model = make_qa_s2s_model(
model_name="t5-small", from_file="seq2seq_models/eli5_t5_model_1024_4.pth", device="cuda:0"
)
return (qar_tokenizer, qar_model, s2s_tokenizer, s2s_model)
@st.cache(allow_output_mutation=True)
def load_indexes():
if LOAD_DENSE_INDEX:
faiss_res = faiss.StandardGpuResources()
wiki40b_passages = datasets.load_dataset(path="wiki_snippets", name="wiki40b_en_100_0")["train"]
wiki40b_passage_reps = np.memmap(
"wiki40b_passages_reps_32_l-8_h-768_b-512-512.dat",
dtype="float32",
mode="r",
shape=(wiki40b_passages.num_rows, 128),
)
wiki40b_index_flat = faiss.IndexFlatIP(128)
wiki40b_gpu_index_flat = faiss.index_cpu_to_gpu(faiss_res, 1, wiki40b_index_flat)
wiki40b_gpu_index_flat.add(wiki40b_passage_reps) # TODO fix for larger GPU
else:
wiki40b_passages, wiki40b_gpu_index_flat = (None, None)
es_client = Elasticsearch([{"host": "localhost", "port": "9200"}])
return (wiki40b_passages, wiki40b_gpu_index_flat, es_client)
@st.cache(allow_output_mutation=True)
def load_train_data():
eli5 = datasets.load_dataset("eli5", name="LFQA_reddit")
eli5_train = eli5["train_eli5"]
eli5_train_q_reps = np.memmap(
"eli5_questions_reps.dat", dtype="float32", mode="r", shape=(eli5_train.num_rows, 128)
)
eli5_train_q_index = faiss.IndexFlatIP(128)
eli5_train_q_index.add(eli5_train_q_reps)
return (eli5_train, eli5_train_q_index)
passages, gpu_dense_index, es_client = load_indexes()
qar_tokenizer, qar_model, s2s_tokenizer, s2s_model = load_models()
eli5_train, eli5_train_q_index = load_train_data()
def find_nearest_training(question, n_results=10):
q_rep = embed_questions_for_retrieval([question], qar_tokenizer, qar_model)
D, I = eli5_train_q_index.search(q_rep, n_results)
nn_examples = [eli5_train[int(i)] for i in I[0]]
return nn_examples
def make_support(question, source="wiki40b", method="dense", n_results=10):
if source == "none":
support_doc, hit_lst = (" <P> ".join(["" for _ in range(11)]).strip(), [])
else:
if method == "dense":
support_doc, hit_lst = query_qa_dense_index(
question, qar_model, qar_tokenizer, passages, gpu_dense_index, n_results
)
else:
support_doc, hit_lst = query_es_index(
question,
es_client,
index_name="english_wiki40b_snippets_100w",
n_results=n_results,
)
support_list = [
(res["article_title"], res["section_title"].strip(), res["score"], res["passage_text"]) for res in hit_lst
]
question_doc = "question: {} context: {}".format(question, support_doc)
return question_doc, support_list
@st.cache(
hash_funcs={
torch.Tensor: (lambda _: None),
transformers.models.bart.tokenization_bart.BartTokenizer: (lambda _: None),
}
)
def answer_question(
question_doc, s2s_model, s2s_tokenizer, min_len=64, max_len=256, sampling=False, n_beams=2, top_p=0.95, temp=0.8
):
with torch.no_grad():
answer = qa_s2s_generate(
question_doc,
s2s_model,
s2s_tokenizer,
num_answers=1,
num_beams=n_beams,
min_len=min_len,
max_len=max_len,
do_sample=sampling,
temp=temp,
top_p=top_p,
top_k=None,
max_input_length=1024,
device="cuda:0",
)[0]
return (answer, support_list)
st.title("Long Form Question Answering with ELI5")
# Start sidebar
header_html = "<img src='https://huggingface.co/front/assets/huggingface_logo.svg'>"
header_full = """
<html>
<head>
<style>
.img-container {
padding-left: 90px;
padding-right: 90px;
padding-top: 50px;
padding-bottom: 50px;
background-color: #f0f3f9;
}
</style>
</head>
<body>
<span class="img-container"> <!-- Inline parent element -->
%s
</span>
</body>
</html>
""" % (header_html,)
st.sidebar.markdown(
header_full,
unsafe_allow_html=True,
)
# Long Form QA with ELI5 and Wikipedia
description = """
This demo presents a model trained to [provide long-form answers to open-domain questions](https://yjernite.github.io/lfqa.html).
First, a document retriever fetches a set of relevant Wikipedia passages given the question from the [Wiki40b](https://research.google/pubs/pub49029/) dataset,
a pre-processed fixed snapshot of Wikipedia.
"""
st.sidebar.markdown(description, unsafe_allow_html=True)
action_list = [
"Answer the question",
"View the retrieved document only",
"View the most similar ELI5 question and answer",
"Show me everything, please!",
]
demo_options = st.sidebar.checkbox("Demo options")
if demo_options:
action_st = st.sidebar.selectbox(
"",
action_list,
index=3,
)
action = action_list.index(action_st)
show_type = st.sidebar.selectbox(
"",
["Show full text of passages", "Show passage section titles"],
index=0,
)
show_passages = show_type == "Show full text of passages"
else:
action = 3
show_passages = True
retrieval_options = st.sidebar.checkbox("Retrieval options")
if retrieval_options:
retriever_info = """
### Information retriever options
The **sparse** retriever uses ElasticSearch, while the **dense** retriever uses max-inner-product search between a question and passage embedding
trained using the [ELI5](https://arxiv.org/abs/1907.09190) questions-answer pairs.
The answer is then generated by sequence to sequence model which takes the question and retrieved document as input.
"""
st.sidebar.markdown(retriever_info)
wiki_source = st.sidebar.selectbox("Which Wikipedia format should the model use?", ["wiki40b", "none"])
index_type = st.sidebar.selectbox("Which Wikipedia indexer should the model use?", ["dense", "sparse", "mixed"])
else:
wiki_source = "wiki40b"
index_type = "dense"
sampled = "beam"
n_beams = 2
min_len = 64
max_len = 256
top_p = None
temp = None
generate_options = st.sidebar.checkbox("Generation options")
if generate_options:
generate_info = """
### Answer generation options
The sequence-to-sequence model was initialized with [BART](https://huggingface.co/facebook/bart-large)
weights and fine-tuned on the ELI5 QA pairs and retrieved documents. You can use the model for greedy decoding with
**beam** search, or **sample** from the decoder's output probabilities.
"""
st.sidebar.markdown(generate_info)
sampled = st.sidebar.selectbox("Would you like to use beam search or sample an answer?", ["beam", "sampled"])
min_len = st.sidebar.slider(
"Minimum generation length", min_value=8, max_value=256, value=64, step=8, format=None, key=None
)
max_len = st.sidebar.slider(
"Maximum generation length", min_value=64, max_value=512, value=256, step=16, format=None, key=None
)
if sampled == "beam":
n_beams = st.sidebar.slider("Beam size", min_value=1, max_value=8, value=2, step=None, format=None, key=None)
else:
top_p = st.sidebar.slider(
"Nucleus sampling p", min_value=0.1, max_value=1.0, value=0.95, step=0.01, format=None, key=None
)
temp = st.sidebar.slider(
"Temperature", min_value=0.1, max_value=1.0, value=0.7, step=0.01, format=None, key=None
)
n_beams = None
# start main text
questions_list = [
"<MY QUESTION>",
"How do people make chocolate?",
"Why do we get a fever when we are sick?",
"How can different animals perceive different colors?",
"What is natural language processing?",
"What's the best way to treat a sunburn?",
"What exactly are vitamins ?",
"How does nuclear energy provide electricity?",
"What's the difference between viruses and bacteria?",
"Why are flutes classified as woodwinds when most of them are made out of metal ?",
"Why do people like drinking coffee even though it tastes so bad?",
"What happens when wine ages? How does it make the wine taste better?",
"If an animal is an herbivore, where does it get the protein that it needs to survive if it only eats grass?",
"How can we set a date to the beginning or end of an artistic period? Doesn't the change happen gradually?",
"How does New Zealand have so many large bird predators?",
]
question_s = st.selectbox(
"What would you like to ask? ---- select <MY QUESTION> to enter a new query",
questions_list,
index=1,
)
if question_s == "<MY QUESTION>":
question = st.text_input("Enter your question here:", "")
else:
question = question_s
if st.button("Show me!"):
if action in [0, 1, 3]:
if index_type == "mixed":
_, support_list_dense = make_support(question, source=wiki_source, method="dense", n_results=10)
_, support_list_sparse = make_support(question, source=wiki_source, method="sparse", n_results=10)
support_list = []
for res_d, res_s in zip(support_list_dense, support_list_sparse):
if tuple(res_d) not in support_list:
support_list += [tuple(res_d)]
if tuple(res_s) not in support_list:
support_list += [tuple(res_s)]
support_list = support_list[:10]
question_doc = "<P> " + " <P> ".join([res[-1] for res in support_list])
else:
question_doc, support_list = make_support(question, source=wiki_source, method=index_type, n_results=10)
if action in [0, 3]:
answer, support_list = answer_question(
question_doc,
s2s_model,
s2s_tokenizer,
min_len=min_len,
max_len=int(max_len),
sampling=(sampled == "sampled"),
n_beams=n_beams,
top_p=top_p,
temp=temp,
)
st.markdown("### The model generated answer is:")
st.write(answer)
if action in [0, 1, 3] and wiki_source != "none":
st.markdown("--- \n ### The model is drawing information from the following Wikipedia passages:")
for i, res in enumerate(support_list):
wiki_url = "https://en.wikipedia.org/wiki/{}".format(res[0].replace(" ", "_"))
sec_titles = res[1].strip()
if sec_titles == "":
sections = "[{}]({})".format(res[0], wiki_url)
else:
sec_list = sec_titles.split(" & ")
sections = " & ".join(
["[{}]({}#{})".format(sec.strip(), wiki_url, sec.strip().replace(" ", "_")) for sec in sec_list]
)
st.markdown(
"{0:02d} - **Article**: {1:<18} <br> _Section_: {2}".format(i + 1, res[0], sections),
unsafe_allow_html=True,
)
if show_passages:
st.write(
'> <span style="font-family:arial; font-size:10pt;">' + res[-1] + "</span>", unsafe_allow_html=True
)
if action in [2, 3]:
nn_train_list = find_nearest_training(question)
train_exple = nn_train_list[0]
st.markdown(
"--- \n ### The most similar question in the ELI5 training set was: \n\n {}".format(train_exple["title"])
)
answers_st = [
"{}. {}".format(i + 1, " \n".join([line.strip() for line in ans.split("\n") if line.strip() != ""]))
for i, (ans, sc) in enumerate(zip(train_exple["answers"]["text"], train_exple["answers"]["score"]))
if i == 0 or sc > 2
]
st.markdown("##### Its answers were: \n\n {}".format("\n".join(answers_st)))
disclaimer = """
---
**Disclaimer**
*The intent of this app is to provide some (hopefully entertaining) insights into the behavior of a current LFQA system.
Evaluating biases of such a model and ensuring factual generations are still very much open research problems.
Therefore, until some significant progress is achieved, we caution against using the generated answers for practical purposes.*
"""
st.sidebar.markdown(disclaimer, unsafe_allow_html=True)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/longform-qa/requirements.txt
|
datasets >= 1.1.3
faiss-cpu
streamlit
elasticsearch
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/robust-speech-event/README.md
|
# Robust Speech Challenge 🤗
Welcome to the robust speech recognition challenge 🎙️ !
The goal of this event is to build **robust**, **real-world** speech recognition (ASR) systems in as many languages as possible 🌏🌍🌎.
If necessary and available, free access to a V100S 32 GB GPU will kindly be provided by the [OVHcloud team]( https://www.ovhcloud.com/) 🚀.
This document summarizes all the relevant information required for the speech community event 📋.
To sign-up, please see [this forum post](https://discuss.huggingface.co/t/open-to-the-community-robust-speech-recognition-challenge/13614) 🤗. Please make sure to:
- Read it in detail
- Fill the google form
- Join our Discord server in the #join-sprint channel.
## Table of Contents
- [TLDR;](#tldr)
- [Important dates](#important-dates)
- [How to install pytorch, transformers, datasets](#how-to-install-relevant-libraries)
- [Data and Preprocessing](#data-and-preprocessing)
- [How to fine-tune an acoustic model](#how-to-finetune-an-acoustic-model)
- [How to fine-tune with OVH could](#how-to-finetune-with-ovh-cloud)
- [How to combine n-gram language models with acoustic model](#how-to-combine-n-gram-with-acoustic-model)
- [Evaluation](#evaluation)
- [Prizes](#prizes)
- [Communication and Problems](#communication-and-problems)
- [Talks](#talks)
- [General Tips & Tricks](#general-tips-and-tricks)
## TLDR
Participants are encouraged to leverage pre-trained speech recognition checkpoints,
preferably [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53),
to train a speech recognition system in a language of their choice.
Speech recognition systems should be trained using **PyTorch**, **🤗 Transformers**, and, **🤗 Datasets**.
For more information on how to install the above libraries, please read through
[How to install pytorch, transformers, datasets](#how-to-install-relevant-libraries).
Participants can make use of whatever data they think is useful to build a
speech recognition system for **real-world** audio data -
**except** the Common Voice `"test"` split of their chosen language.
The section [Data and preprocessing](#data-and-preprocessing) explains
in more detail what audio data can be used, how to find suitable audio data, and
how the audio data can be processed.
For training, it is recommended to use the [official training script](https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py) or a modification thereof. A step-by-step guide on how to fine-tune
an acoustic model for a speech recognition system can be found under [How to fine-tune an acoustic model](#how-to-finetune-an-acoustic-model).
If possible it is encouraged to fine-tune the acoustic models on local GPU machines, but
if those are not available, the OVH could team kindly provides a limited
number of GPUs for the event. Simply fill out [this google form](https://forms.gle/GFZkMkKLiufi75g28) to get access to a GPU.
For more information on how to train an acoustic model on one of OVH's GPU - see [How to fine-tune a speech recognition model with OVHcould](#how-to-fine-tune-with-ovh-cloud).
The performance of speech recognition system can often significantly be improved by adding a
language model for decoding. For more information on how to add a language model, please
take a look at [How to combine n-gram language models with speech recognition models](#how-to-combine-n-gram-with-model).
During the event, the speech recognition system will be evaluated on both the Common Voice `"test"` split
of the participants' chosen language as well as the *real-world* `"dev"` data provided by
the Hugging Face team.
At the end of the robust speech recognition challenge, the speech recognition system will also be evaluated on the
*real-world* `"test"` data provided by the Hugging Face team. Each participant should add an
`eval.py` script to her/his model repository in a specific format that lets one easily
evaluate the speech recognition system on both Common Voice's `"test"` data as well as the *real-world* audio
data. Please read through the [Evaluation](#evaluation) section to make sure your evaluation script is in the correct format. Speech recognition systems
with evaluation scripts in an incorrect format can sadly not be considered for the Challenge.
At the end of the event, the best performing speech recognition system
will receive a prize 🏆 - more information regarding the prizes can be found under [Prizes](#prizes).
We believe that framing the event as a competition is more fun, but at the core, the event is about
creating speech recognition systems in as many languages as possible as a community.
This can be achieved by working together, helping each other to solve bugs, share important findings, etc...🤗
**Note**:
Please, read through the section on [Communication & Problems](#communication-and-problems) to make sure you
know how to ask for help, etc...
All important announcements will be made on discord. Please make sure that
you've joined [this discord channel](https://discord.gg/SHr5wC7m)
Also, please make sure that you have been added to the [Speech Event Organization](https://huggingface.co/speech-recognition-community-v2).
You should have received an invite by email. If you didn't receive an invite, please contact the organizers, *e.g.* Anton, Patrick, or Omar directly on discord.
## Important dates
## Data and preprocessing
In this section, we will quickly go over how to find suitable training data and
how to preprocess it.
To begin with, **all data except Common Voice's `"test"` data can be used as training data.**
The exception includes all Common Voice versions as the test data split of later Common Voice versions often
overlaps with the one of previous versions, *e.g.* the test data of Common Voice 7 in English is
to a big part identical to the test data of Common Voice 6 in English:
```python
load_dataset("mozilla-foundation/common_voice_7_0", "en", split="test")
```
includes more or less the same data as
```python
load_dataset("mozilla-foundation/common_voice_6_1", "en", split="test")
```
However, we strongly encourage participants to make use of Common Voice's other splits, *e.g.* `"train"` and `"validation"`.
For most languages, the Common Voice dataset offers already a decent amount of training data. It is usually
always advantageous to collect additional data. To do so, the participants are in a first step encouraged to search the
Hugging Face Hub for additional audio data, for example by selecting the category
["speech-processing"](https://huggingface.co/datasets?task_categories=task_categories:speech-processing&sort=downloads).
All datasets that are available on the Hub can be downloaded via the 🤗 Datasets library in the same way Common Voice is downloaded.
If one wants to combine multiple datasets for training, it might make sense to take a look at
the [`interleave_datasets`](https://huggingface.co/docs/datasets/package_reference/main_classes?highlight=interleave#datasets.interleave_datasets) function.
In addition, participants can also make use of their audio data. Here, please make sure that you **are allowed to use the audio data**. E.g., if audio data
is taken from media platforms, such as YouTube, it should be verified that the media platform and the owner of the data have given her/his approval to use the audio
data in the context of machine learning research. If you are not sure whether the data you want to use has the appropriate licensing, please contact the Hugging Face
team on discord.
Next, let's talk about preprocessing. Audio data and transcriptions have to be brought into the correct format when
training the acoustic model (example shown in [How to fine-tune an acoustic model](#how-to-finetune-an-acoustic-model)).
It is recommended that this is done by using 🤗 Datasets `.map()` function as shown
[here](https://github.com/huggingface/transformers/blob/9a2dabae7002258e41419491c73dd43ad61b5de7/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py#L444). As can be
see we can pass some characters that will be removed from the transcriptions, *e.g.*: `--chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \`
on the official ["Single GPU Example"](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition#single-gpu-ctc).
The participants are free to modify this preprocessing by removing more characters or even replacing characters as
it is done in the [official blog post](https://github.com/huggingface/transformers/blob/9a2dabae7002258e41419491c73dd43ad61b5de7/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py#L444).
**However**, there are some rules regarding what characters are allowed to be removed/replaced and which are not.
These rules are not this straightforward and therefore often have to be evaluated case-by-case.
It is allowed (and recommended) to normalize the data to only have lower-case characters. It is also allowed (and recommended) to remove typographical
symbols and punctuation marks. A list of such symbols can *e.g.* be found [here](https://en.wikipedia.org/wiki/List_of_typographical_symbols_and_punctuation_marks) - however here we already must be careful. We should **not** remove a symbol that would change the meaning of the words, *e.g.* in English,
we should not remove the single quotation mark `'` since it would change the meaning of the word `"it's"` to `"its"` which would then be incorrect.
So the golden rule here is to not remove any characters that could change the meaning of a word into another word. This is not always obvious and should
be given some consideration. As another example, it is fine to remove the "Hyphen-minus" sign "`-`" since it doesn't change the
meaning of a word to another one. *E.g.* "`fine-tuning`" would be changed to "`finetuning`" which has still the same meaning.
Since those choices are not always obvious when in doubt feel free to ask on Discord or even better post your question on the forum, as was
done, *e.g.* [here](https://discuss.huggingface.co/t/spanish-asr-fine-tuning-wav2vec2/4586).
## How to install relevant libraries
The following libraries are required to fine-tune a speech model with 🤗 Transformers and 🤗 Datasets in PyTorch.
- [PyTorch](https://pytorch.org/)
- [Transformers](https://github.com/huggingface/transformers)
- [Datasets](https://github.com/huggingface/datasets)
We recommend installing the above libraries in a [virtual environment](https://docs.python.org/3/library/venv.html).
If you're unfamiliar with Python virtual environments, check out the [user guide](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/). Create a virtual environment with the version of Python you're going
to use and activate it.
You should be able to run the command:
```bash
python3 -m venv <your-venv-name>
```
You can activate your venv by running
```bash
source ~/<your-venv-name>/bin/activate
```
To begin with please make sure you have PyTorch and CUDA correctly installed.
The following command should return ``True``:
```bash
python -c "import torch; print(torch.cuda.is_available())"
```
If the above command doesn't print ``True``, in the first step, please follow the
instructions [here](https://pytorch.org/) to install PyTorch with CUDA.
We strongly recommend making use of the provided PyTorch examples scripts in [transformers/examples/pytorch/speech-recognition](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition) to train your speech recognition
system.
In all likelihood, you will adjust one of the example scripts, so we recommend forking and cloning the 🤗 Transformers repository as follows.
1. Fork the [repository](https://github.com/huggingface/transformers) by
clicking on the 'Fork' button on the repository's page. This creates a copy of the code
under your GitHub user account.
2. Clone your fork to your local disk, and add the base repository as a remote:
```bash
$ git clone https://github.com/<your Github handle>/transformers.git
$ cd transformers
$ git remote add upstream https://github.com/huggingface/transformers.git
```
3. Create a new branch to hold your development changes. This is especially useful to share code changes with your team:
```bash
$ git checkout -b a-descriptive-name-for-my-project
```
4. Set up a PyTorch environment by running the following command your virtual environment:
```bash
$ pip install -e ".[torch-speech]"
```
(If transformers was already installed in the virtual environment, remove
it with `pip uninstall transformers` before reinstalling it in editable
mode with the `-e` flag.)
If you have already cloned that repo, you might need to `git pull` to get the most recent changes in the `transformers`
library.
Running this command will automatically install `torch` and the most relevant
libraries required for fine-tuning a speech recognition system.
Next, you should also install the 🤗 Datasets library. We strongly recommend installing the
library from source to profit from the most current additions during the community week.
Simply run the following steps:
```
$ cd ~/
$ git clone https://github.com/huggingface/datasets.git
$ cd datasets
$ pip install -e ".[streaming]"
```
If you plan on contributing a specific dataset during
the community week, please fork the datasets repository and follow the instructions
[here](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-create-a-pull-request).
To verify that all libraries are correctly installed, you can run the following command in a Python shell.
It verifies that both `transformers` and `datasets` have been correclty installed.
```python
from transformers import AutoModelForCTC, AutoProcessor
from datasets import load_dataset
dummy_dataset = load_dataset("common_voice", "ab", split="test")
model = AutoModelForCTC.from_pretrained("hf-internal-testing/tiny-random-wav2vec2")
model.to("cuda")
processor = AutoProcessor.from_pretrained("hf-internal-testing/tiny-random-wav2vec2")
input_values = processor(dummy_dataset[0]["audio"]["array"], return_tensors="pt", sampling_rate=16_000).input_values
input_values = input_values.to("cuda")
logits = model(input_values).logits
assert logits.shape[-1] == 32
```
## How to finetune an acoustic model
In this section, we show you how to fine-tune a pre-trained [XLS-R Model](https://huggingface.co/docs/transformers/model_doc/xls_r) on the [Common Voice 7 dataset](https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0).
We recommend fine-tuning one of the following pre-trained XLS-R checkpoints:
- [300M parameters version](https://huggingface.co/facebook/wav2vec2-xls-r-300m)
- [1B parameters version](https://huggingface.co/facebook/wav2vec2-xls-r-1b)
- [2B parameters version](https://huggingface.co/facebook/wav2vec2-xls-r-2b)
To begin with, please note that to use the Common Voice dataset, you
have to accept that **your email address** and **username** are shared with the
mozilla-foundation. To get access to the dataset please click on "*Access repository*" [here](https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0).
Next, we recommended that you get familiar with the XLS-R model and its capabilities.
In collaboration with [Fairseq's Wav2Vec2 team](https://github.com/pytorch/fairseq/tree/main/examples/wav2vec),
we've written ["Fine-tuning XLS-R for Multi-Lingual ASR with 🤗 Transformers"](https://huggingface.co/blog/fine-tune-xlsr-wav2vec2) which gives an in-detail explanation of how XLS-R functions and how it can be fine-tuned.
The blog can also be opened and directly fine-tuned in a google colab notebook.
In this section, we will explain how to fine-tune the model on a local machine.
1. **Log in**
To begin with, you should check that you are correctly logged in and that you have `git-lfs` installed so that your fine-tuned model can automatically be uploaded.
Run:
```bash
huggingface-cli login
```
to login. It is recommended to login with your access token that can be found under your hugging face profile (icon in the top right corner on [hf.co](http://hf.co/), then Settings -> Access Tokens -> User Access Tokens -> New Token (if haven't generated one already)
You can then copy-paste this token to log in locally.
2. **Create your model repository**
First, let's make sure that `git-lfs` is correctly installed. To so, simply run:
```bash
git-lfs -v
```
The output should show something like `git-lfs/2.13.2 (GitHub; linux amd64; go 1.15.4)`. If your console states that the `git-lfs` command was not found, please make
sure to install it [here](https://git-lfs.github.com/) or simply via:
```bash
sudo apt-get install git-lfs
```
Now you can create your model repository which will contain all relevant files to
reproduce your training. You can either directly create the model repository on the
Hub (Settings -> New Model) or via the CLI. Here we choose to use the CLI instead.
Assuming that we want to call our model repository *xls-r-ab-test*, we can run the
following command:
```bash
huggingface-cli repo create xls-r-ab-test
```
You can now see the model on the Hub, *e.g.* under https://huggingface.co/hf-test/xls-r-ab-test .
Let's clone the repository so that we can define our training script inside.
```bash
git lfs install
git clone https://huggingface.co/hf-test/xls-r-ab-test
```
3. **Add your training script and `run`-command to the repository**
We encourage participants to add all relevant files for training directly to the
directory so that everything is fully reproducible.
Let's first copy-paste the official training script from our clone
of `transformers` to our just created directory:
```bash
cp ~/transformers/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py ./
```
Next, we'll create a bash file to define the hyper-parameters and configurations
for training. More detailed information on different settings (single-GPU vs. multi-GPU) can be found [here](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition#connectionist-temporal-classification).
For demonstration purposes, we will use a dummy XLS-R model `model_name_or_path="hf-test/xls-r-dummy"` on the very low-resource language of "Abkhaz" of [Common Voice 7](https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0): `dataset_config_name="ab"` for just a single epoch.
Before starting to train, let's make sure we have installed all the required libraries. You might want to run:
```bash
pip install -r ~/transformers/examples/pytorch/speech-recognition/requirements.txt
```
Alright, finally we can define the training script. We'll simply use some
dummy hyper-parameters and configurations for demonstration purposes.
Note that we add the flag `--use_auth_token` so that datasets requiring access,
such as [Common Voice 7](https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0) can be downloaded. In addition, we add the `--push_to_hub` flag to make use of the
[Trainers `push_to-hub` functionality](https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Trainer.push_to_hub) so that your model will be automatically uploaded to the Hub.
Let's copy the following code snippet in a file called `run.sh`
```bash
echo '''python run_speech_recognition_ctc.py \
--dataset_name="mozilla-foundation/common_voice_7_0" \
--model_name_or_path="hf-test/xls-r-dummy" \
--dataset_config_name="ab" \
--output_dir="./" \
--overwrite_output_dir \
--max_steps="10" \
--per_device_train_batch_size="2" \
--learning_rate="3e-4" \
--save_total_limit="1" \
--evaluation_strategy="steps" \
--text_column_name="sentence" \
--length_column_name="input_length" \
--save_steps="5" \
--layerdrop="0.0" \
--freeze_feature_encoder \
--gradient_checkpointing \
--fp16 \
--group_by_length \
--push_to_hub \
--use_auth_token \
--do_train --do_eval''' > run.sh
```
4. **Start training**
Now all that is left to do is to start training the model by executing the
run file.
```bash
bash run.sh
```
The training should not take more than a couple of minutes.
During the training intermediate saved checkpoints are automatically uploaded to
your model repository as can be seen [on this commit](https://huggingface.co/hf-test/xls-r-ab-test/commit/0eb19a0fca4d7d163997b59663d98cd856022aa6) .
At the end of the training, the [Trainer](https://huggingface.co/docs/transformers/main/en/main_classes/trainer) automatically creates a nice model card and all
relevant files are uploaded.
5. **Tips for real model training**
The above steps illustrate how a model can technically be fine-tuned.
However as you can see on the model card [hf-test/xls-r-ab-test](https://huggingface.co/hf-test/xls-r-ab-test), our demonstration has a very poor performance which is
not surprising given that we trained for just 10 steps on a randomly initialized
model.
For real model training, it is recommended to use one of the actual pre-trained XLS-R models:
- [300M parameters version](https://huggingface.co/facebook/wav2vec2-xls-r-300m)
- [1B parameters version](https://huggingface.co/facebook/wav2vec2-xls-r-1b)
- [2B parameters version](https://huggingface.co/facebook/wav2vec2-xls-r-2b)
Also, the hyper-parameters should be carefully chosen depending on the dataset.
As an example, we will fine-tune the 300M parameters model on Swedish on a single
TITAN RTX 24GB GPU.
The model will be called `"xls-r-300m-sv"`.
Following the above steps we first create the model:
```bash
huggingface-cli repo create xls-r-300m-sv
```
, clone it locally (assuming the `<username>` is `hf-test`)
```bash
git clone hf-test/xls-r-300m-sv
```
, and, define the following hyperparameters for training
```bash
echo '''python run_speech_recognition_ctc.py \
--dataset_name="mozilla-foundation/common_voice_7_0" \
--model_name_or_path="facebook/wav2vec2-xls-r-300m" \
--dataset_config_name="sv-SE" \
--output_dir="./" \
--overwrite_output_dir \
--num_train_epochs="50" \
--per_device_train_batch_size="8" \
--per_device_eval_batch_size="8" \
--gradient_accumulation_steps="4" \
--learning_rate="7.5e-5" \
--warmup_steps="2000" \
--length_column_name="input_length" \
--evaluation_strategy="steps" \
--text_column_name="sentence" \
--chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … – \
--save_steps="500" \
--eval_steps="500" \
--logging_steps="100" \
--layerdrop="0.0" \
--activation_dropout="0.1" \
--save_total_limit="3" \
--freeze_feature_encoder \
--feat_proj_dropout="0.0" \
--mask_time_prob="0.75" \
--mask_time_length="10" \
--mask_feature_prob="0.25" \
--mask_feature_length="64" \
--gradient_checkpointing \
--use_auth_token \
--fp16 \
--group_by_length \
--do_train --do_eval \
--push_to_hub''' > run.sh
```
The training takes *ca.* 7 hours and yields a reasonable test word
error rate of 27% as can be seen on the automatically generated [model card](https://huggingface.co/hf-test/xls-r-300m-sv).
The above-chosen hyperparameters probably work quite well on a range of different
datasets and languages but are by no means optimal. It is up to you to find a good set of
hyperparameters.
## How to finetune with OVH cloud
[](https://youtu.be/XkMnYocAEO0) For a more detailed guide on setting up OVHcloud please watch this video: https://youtu.be/XkMnYocAEO0
### Creating an OVHCloud account
*TIP*: If you haven't created a project on OVHcloud yet, make sure you've received your GPU voucher code *beforehand*,
so that you can skip entering the credit card information.
1. If you're a US citizen, create an account via [OVHcloud.CA](https://ovhcloud.ca/).
If you're from anywhere else in the world, create an account via [OVHcloud.COM](https://ovhcloud.com/).
2. Once logged in, click `Public Cloud` from the top menu and then click `Create your first OVH Public Cloud project`.
Then enter a project name (e.g. "huggingface"), enter your voucher code, and click `Continue` -> `Create my project`.
*Note: if you see a request for credit card details during the last step, and you can't skip it, then your voucher code
is invalid. Please report it to the [#ovh-support](https://discord.gg/p4qqDV3M) channel on Discord.*
### Setting up an AI notebook
1. Go to the `Public Cloud` page and select `Project Management` -> `Users & Roles` from the menu on the left.
2. Click `+ Add user`. Write a user description (e.g. `AI Trainer`), and select an `AI Training Operator` user role.
Click `Confirm`.
3. Write down the *username* and *password* (at the top of the screen) somewhere. They will be needed during step 7.
4. Select `AI & Machine Learning` -> `AI Training` from the menu on the left.
Click `+ Launch a new job` on the AI Training page.
5. On the `Launch a new job` page:
* In `1. Choose a region` select a region closest to you.
* In `2. Enter the Docker image` select `Custom image` -> `baaastijn/ovh_huggingface`.
* You can skip steps `3.` and `4.` if you will be using the Hugging Face Hub to store the models after training.
* In `5. Configure your job` select **1** `GPU`.
* Validate the info and Create the job.
6. On the `AI Training Jobs` screen wait until the job's status changes from `Pending` to `Running`.
7. Click `HTTP Access` from the Job's details page and log in with the AI training user you've created earlier.
Once logged in, you can close the page and click `HTTP Access` to launch a JupyterLab notebook.
8. Awesome, now you have a free GPU-enabled Jupyter instance!
**Note**: If you're an experienced Docker user, feel free to create a custom docker image with all of the needed packages
like the one in step 5. The Dockerfile for it is available here:
[baaastijn/Dockerimages](https://github.com/baaastijn/Dockerimages/tree/main/Hugginface_challenge_speech).
Once you've built your image, push it to https://hub.docker.com/ and select it during the OVHcloud job creation.
For more quick tutorials about OVHcloud AI products, check out the showcase https://vimeo.com/showcase/8903300
## How to combine n-gram with acoustic model
Having trained a speech recognition model with CTC as shown in the section above,
one can further improve the model's performance by adding an **n-gram language model**
to the decoding process of the model. By doing so, we are replacing the naive greedy decoding
with **n-gram-boosted** beam search decoding.
N-gram language models can be built on CPU in just a few minutes. *N-gram-boosted* beam search decoding noticeably slows down the
inference time, but also yields significant word error rates improvements - usually between 10-40 %.
You can find an in-detail blog post on how to build an *n-gram* [here](https://huggingface.co/blog/wav2vec2-with-ngram).
The blog post can be opened in a google colab and by adapting three lines of the example for your use case, one can directly
create an *n-gram* in the google colab.
The blog post gives in-detail instructions on how to build an n-gram and how to add it to your trained speech recognition model.
- why one should add an *n-gram* to her/his speech recognition system,
- how to build an *n-gram*, and,
- how to add the built *n-gram* the speech recognition system for seamless decoding
Our previously trained model - [xls-r-300m-sv](https://huggingface.co/hf-test/xls-r-300m-sv) - enjoys a 30% word error rate reduction after
having added an n-gram. As shown in the example of the blog post, we strongly advise participants to upload all files required for combining
the *n-gram* with a trained speech recognition model directly into the same model repository.
## Evaluation
Finally, we have arrived at the most fun part of the challenge - sitting back and
watching the model transcribe audio. If possible, every participant should evaluate
the speech recognition system on the test set of Common Voice 7 and
ideally also on the real-world audio data (if available).
For languages that have neither a Common Voice evaluation dataset nor a real world
evaluation dataset, please contact the organizers on Discord so that we can work
together to find some evaluation data.
As a first step, one should copy the official `eval.py` script to her/his model
repository. Let's use our previously trained [xls-r-300m-sv](https://huggingface.co/hf-test/xls-r-300m-sv) again as an example.
Assuming that we have a clone of the model's repo under `~/xls-r-300m-sv`, we can
copy the `eval.py` script to the repo.
```bash
cp ~/transformers/examples/research_projects/robust-speech-event/eval.py ~/xls-r-300m-sv
```
Next, we should adapt `eval.py` so that it fits our evaluation data. Here it is
important to keep the `eval.py` file in the following format:
- 1. The following input arguments should not be changed and keep their original functionality/meaning (being to load the model and dataset): `"--model_id"`, `"--dataset"`, `"--config"`, `"--split"`. We recommend to not change any of the code written under `if __name__ == "__main__":`.
- 2. The function `def log_results(result: Dataset, args: Dict[str, str])` should also not be changed. The function expects the above names attached to the `args` object as well as a `datasets.Dataset` object, called `result` which includes all predictions and target transcriptions under the names `"predictions"` and `"targets"` respectively.
- 3. All other code can be changed and adapted. Participants are especially invited to change the `def normalize_text(text: str) -> str:` function as this might be a very language and model-training specific function.
- 4. **Important**: It is not allowed to "cheat" in any way when in comes to pre-and postprocessing. In short, "cheating" refers to any of the following:
- a. Somehow giving the model access to the target transcriptions to improve performance. The model is not allowed to use the target transcriptions to generate its predictions.
- b. Pre-processing the target transcriptions in a way that makes the target transcriptions lose their original meaning. This corresponds to what has already been said in [Data and Preprocessing](#data-and-preprocessing) and is somewhat of a grey zone. It means that one should not remove characters that would make a word to lose its meaning. E.g., it is not allowed to replace all `e` in English with `i` and simply make the model learn that `e` and `i` are the same letter for a better word error rate. This would destroy the meaning of words such as `fell -> fill`. However, it is totally fine to normalize (*e.g.* lowercase) all letters, remove punctuation. There can be a lot of language-specific exceptions and in case you are not sure whether your target transcription pre-processing is allowed, please ask on the Discord channel.
Uff, that was a lot of text describing how to make sure your `eval.py` script
is in the correct format. If you have any questions, please ask openly in Discord.
Great, now that we have adapted the `eval.py` script, we can lean back and run the
evaluation.
First, one should evaluate the model on Common Voice 7's test data. This might
already have been done for your acoustic model during training but in case you
added an *n-gram* language model after having fine-tuned the acoustic model, you
should now see a nice improvement.
The command to evaluate our test model [xls-r-300m-sv](https://huggingface.co/hf-test/xls-r-300m-sv) on Common Voice 7's test data is the following:
```bash
cd xls-r-300m-sv
./eval.py --model_id ./ --dataset mozilla-foundation/common_voice_7_0 --config sv-SE --split test --log_outputs
```
To log each of the model's predictions with the target transcriptions, you can just
add the `--log_outputs` flag.
Running this command should automatically create the file:
`mozilla-foundation_common_voice_7_0_sv-SE_test_eval_results.txt` that contains
both the word- and character error rate.
In a few days, we will give everybody access to some real-world audio data for as many languages as possible.
If your language has real-world audio data, it will most likely have audio input
of multiple minutes. 🤗Transformer's [ASR pipeline](https://huggingface.co/docs/transformers/main/en/main_classes/pipelines#transformers.AutomaticSpeechRecognitionPipeline) supports audio chunking out-of-the-box. You only need to specify
how song each audio chunk should be (`chunk_length_s`) and how much audio stride
(`stride_length_s`) each chunk should use.
For more information on the chunking works, please have a look at [this nice blog post](TODO: ).
In the case of `xls-r-300m-sv`, the following command can be run:
```bash
cd xls-r-300m-sv
./eval.py --model_id hf-test/xls-r-300m-sv --dataset <to-be-announced> --config sv --split validation --chunk_length_s 5.0 --stride_length_s 1.0 --log_outputs
```
Great, now you should have successfully evaluated your model. Finally, there is one
**important** thing you should do so that your model is taken into account
for the final evaluation. You should add two tags to your model, one being `robust-speech-event`, one being the ISO code of your chosen language, *e.g.* `"sv"` for the
exemplary model we used above. You can find a list of all available languages and
their ISO code [here](https://huggingface.co/languages).
To add the tags, simply edit the README.md of your model repository and add
```
- "sv"
- "robust-speech-event"
```
under `tags:` as done [here](https://huggingface.co/hf-test/xls-r-300m-sv/commit/a495fd70c96bb7d019729be9273a265c2557345e).
To verify that you've added the tags correctly make sure that your model
appears when clicking on [this link](https://huggingface.co/models?other=robust-speech-event).
Great that's it! This should give you all the necessary information to evaluate
your model. For the final evaluation, we will verify each evaluation result to
determine the final score and thereby the winning models for each language.
The final score is calculated as follows:
```bash
FINAL_SCORE = 1/3 * WER_Common_Voice_7_test + 1/3 * WER_REAL_AUDIO_DEV + 1/3 * WER_REAL_AUDIO_TEST
```
The dataset `WER_REAL_AUDIO_TEST` is hidden and will only be published
at the end of the robust speech challenge.
If there is no real audio data for your language the final score will be
computed solely based on the Common Voice 7 test dataset. If there is also
no Common Voice 7 test dataset for your language, we will see together how to
score your model - if this is the case, please don't be discouraged. We are
especially excited about speech recognition systems of such low-resource
languages and will make sure that we'll decide on a good approach to evaluating
your model.
## Prizes
TODO(Patrick, Omar, ...)
## Communication and Problems
If you encounter any problems or have any questions, you should use one of the following platforms
depending on your type of problem. Hugging Face is an "open-source-first" organization meaning
that we'll try to solve all problems in the most public and most transparent way possible so that everybody
in the community profits.
The following table summarizes what platform to use for which problem.
- Problem/question/bug with the 🤗 Datasets library that you think is a general problem that also impacts other people, please open an [Issues on Datasets](https://github.com/huggingface/datasets/issues/new?assignees=&labels=bug&template=bug-report.md&title=) and ping @anton-l and @patrickvonplaten.
- Problem/question/bug with the 🤗 Transformers library that you think is a general problem that also impacts other people, please open an [Issues on Transformers](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title=) and ping @anton-l and @patrickvonplaten.
- Problem/question with a modified, customized training script that is less likely to impact other people, please post your problem/question [on the forum](https://discuss.huggingface.co/) and ping @anton-l and @patrickvonplaten.
- Questions regarding access to the OVHcloud GPU, please ask in the Discord channel **#ovh-support**.
- Other questions regarding the event, rules of the event, or if you are not sure where to post your question, please ask in the Discord channel **#sprint-discussions**.
## Talks
We are very excited to be hosting 2 days of talks from Kensho-Technologies, Mozilla's Common Voice, Meta AI Research and Hugging Face.
### Thursday, January 20th
Speaker | Topic | Time | Video |
|-------------|---------------------------------|------------------------|------------------------|
| Patrick von Platen, Hugging Face | Introduction to Robust Speech Challenge | 4h30pm - 5h00pm UTC | [](https://www.youtube.com/watch?v=X9e5Tto-Iuk)
| Raymond Grossman and Jeremy Lopez, Kensho-Technologies | Pyctcdecode & Speech2text decoding | 5h30pm - 6h00pm UTC | [](https://www.youtube.com/watch?v=mp7fHMTnK9A)
### Friday, January 21th
Speaker | Topic | Time | Video |
|-------------|---------------------------------|------------------------|------------------------|
| Gabriel Habayeb, Mozilla Common Voice | Unlocking global speech with Mozilla Common Voice | 4h30pm - 5h00pm UTC | [](https://www.youtube.com/watch?v=Vvn984QmAVg)
| Changhan Wang, Meta AI Research | XLS-R: Large-Scale Cross-lingual Speech Representation Learning on 128 Languages | 5h30pm - 6h00pm UTC | [](https://www.youtube.com/watch?v=ic_J7ZCROBM)
### Talks & Speakers
#### Patrick von Platen, Research Engineer, Hugging Face
- Talk: Introduction to Robust Speech Challenge
- Abstract: In this talk, Patrick outlines the Robust Speech Challenge and gives tips and tricks on how to train and evaluate speech recognition systems with 🤗 Transformers and 🤗 Datasets, and PyTorch.
- Speaker info: Patrick von Platen is a research engineer at Hugging Face and one of the core maintainers of the popular Transformers library. He specializes in speech recognition, encoder-decoder models, and long-range sequence modeling. Before joining Hugging Face, Patrick researched speech recognition at Uber AI, Cambridge University, and RWTH Aachen University.
#### Raymond Grossman, Jeremy Lopez, Machine Learning Engineer, Kensho Technologies
- Talk: PyCTCDecode & Speech2text decoding
- Abstract: PyCTCDecode is a fast and feature-rich CTC beam search decoder for speech recognition written in Python, providing n-gram (kenlm) language model support similar to PaddlePaddle's decoder, but incorporating many new features such as byte pair encoding and real-time decoding to support models like Nvidia's Conformer-CTC or Facebook's Wav2Vec2.
- Speaker info :
- Raymond works as a machine learning engineer at Kensho Technologies, specializing in speech and natural language domains. Before coming to Kensho, he studied mathematics at Princeton and was an avid Kaggler under the moniker @ToTrainThemIsMyCause.
- Jeremy is a machine learning engineer at Kensho Technologies and has worked on a variety of different topics including search and speech recognition. Before working at Kensho, he earned a PhD in experimental particle physics at MIT and continued doing physics research as a postdoc at the University of Colorado Boulder.
#### Gabriel Habayeb, Data Engineer, Common Voice @ Mozilla
- Talk: Unlocking global speech with Mozilla Common Voice
- Abstract: Hear from Common Voice Data Engineer Gabriel Habayeb (Mozilla Foundation) as he talks about how Common Voice makes it easy to crowdsource voice data in global languages, as well as getting key insights into the dataset itself, how we maintain quality, use metadata - and our plans for the future!
- Speaker info: Gabriel is a software developer with the Common Voice team at the Mozilla Foundation with a focus on data engineering. Before joining the Foundation, he spent the last six years working across different industries, including education, enterprise and not-for-profit organizations.
#### Changhan Wang, Main author of XLS-R and Research Engineer, Meta AI Research
- Talk: XLS-R: Large-Scale Cross-lingual Speech Representation Learning on 128 Languages
- Abstract: In this talk, Changhan will present XLS-R, a large-scale model for cross-lingual speech representation learning based on wav2vec 2.0. XLS-R has up to 2B parameters and was trained on nearly half a million hours of publicly available speech audio in 128 languages, an order of magnitude more public data than the largest known prior work. On the CoVoST-2 speech translation benchmark, XLS-R improves the previous state of the art by an average of 7.4 BLEU over 21 translation directions into English. For speech recognition, XLS-R improves over the best known prior work on BABEL, MLS, CommonVoice as well as VoxPopuli, lowering error rates by 14-34% relative on average. XLS-R also sets a new state of the art on VoxLingua107 language identification. The XLS-R team hopes to work together with the open-source community to improve speech processing tasks for many more languages of the world.
## General Tips and Tricks
- Memory efficient training:
In case, you are getting out-of-memory errors on your GPU, we recommend to use
[bitsandbytes](https://github.com/TimDettmers/bitsandbytes) to replace the
native memory-intensive Adam optimizer with the one of `bitsandbytes`. You
can simply run the script `./run_speech_recognition_ctc_bnb.py` provided in this
folder that makes use of `bitsandbytes` instead of the official one.
- Dataset streaming
TODO(Patrick)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/robust-speech-event/run_speech_recognition_ctc_bnb.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2021 The HuggingFace Inc. team. 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
""" Fine-tuning a 🤗 Transformers CTC model for automatic speech recognition"""
import functools
import json
import logging
import os
import re
import sys
import warnings
from dataclasses import dataclass, field
from typing import Dict, List, Optional, Union
import bitsandbytes as bnb
import datasets
import numpy as np
import torch
from datasets import DatasetDict, load_dataset, load_metric
import transformers
from transformers import (
AutoConfig,
AutoFeatureExtractor,
AutoModelForCTC,
AutoProcessor,
AutoTokenizer,
HfArgumentParser,
Trainer,
TrainingArguments,
Wav2Vec2Processor,
set_seed,
)
from transformers.trainer_pt_utils import get_parameter_names
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.16.0.dev0")
require_version("datasets>=1.13.3", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
logger = logging.getLogger(__name__)
def list_field(default=None, metadata=None):
return field(default_factory=lambda: default, metadata=metadata)
@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"}
)
tokenizer_name_or_path: Optional[str] = field(
default=None,
metadata={"help": "Path to pretrained tokenizer or tokenizer identifier from huggingface.co/models"},
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
)
freeze_feature_encoder: bool = field(
default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
)
attention_dropout: float = field(
default=0.0, metadata={"help": "The dropout ratio for the attention probabilities."}
)
activation_dropout: float = field(
default=0.0, metadata={"help": "The dropout ratio for activations inside the fully connected layer."}
)
feat_proj_dropout: float = field(default=0.0, metadata={"help": "The dropout ratio for the projected features."})
hidden_dropout: float = field(
default=0.0,
metadata={
"help": "The dropout probability for all fully connected layers in the embeddings, encoder, and pooler."
},
)
final_dropout: float = field(
default=0.0,
metadata={"help": "The dropout probability for the final projection layer."},
)
mask_time_prob: float = field(
default=0.05,
metadata={
"help": (
"Probability of each feature vector along the time axis to be chosen as the start of the vector "
"span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
"vectors will be masked along the time axis."
)
},
)
mask_time_length: int = field(
default=10,
metadata={"help": "Length of vector span to mask along the time axis."},
)
mask_feature_prob: float = field(
default=0.0,
metadata={
"help": (
"Probability of each feature vector along the feature axis to be chosen as the start of the vectorspan"
" to be masked. Approximately ``mask_feature_prob * sequence_length // mask_feature_length`` feature"
" bins will be masked along the time axis."
)
},
)
mask_feature_length: int = field(
default=10,
metadata={"help": "Length of vector span to mask along the feature axis."},
)
layerdrop: float = field(default=0.0, metadata={"help": "The LayerDrop probability."})
ctc_loss_reduction: Optional[str] = field(
default="mean", metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."}
)
@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.
"""
dataset_name: str = field(
metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
dataset_config_name: str = field(
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
train_split_name: str = field(
default="train+validation",
metadata={
"help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
},
)
eval_split_name: str = field(
default="test",
metadata={
"help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
},
)
audio_column_name: str = field(
default="audio",
metadata={"help": "The name of the dataset column containing the audio data. Defaults to 'audio'"},
)
text_column_name: str = field(
default="text",
metadata={"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
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 validation examples to this "
"value if set."
)
},
)
chars_to_ignore: Optional[List[str]] = list_field(
default=None,
metadata={"help": "A list of characters to remove from the transcripts."},
)
eval_metrics: List[str] = list_field(
default=["wer"],
metadata={"help": "A list of metrics the model should be evaluated on. E.g. `'wer cer'`"},
)
max_duration_in_seconds: float = field(
default=20.0,
metadata={
"help": (
"Filter audio files that are longer than `max_duration_in_seconds` seconds to"
" 'max_duration_in_seconds`"
)
},
)
min_duration_in_seconds: float = field(
default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
)
preprocessing_only: bool = field(
default=False,
metadata={
"help": (
"Whether to only do data preprocessing and skip training. This is especially useful when data"
" preprocessing errors out in distributed training due to timeout. In this case, one should run the"
" preprocessing in a non-distributed setup with `preprocessing_only=True` so that the cached datasets"
" can consequently be loaded in distributed training"
)
},
)
use_auth_token: bool = field(
default=False,
metadata={
"help": (
"If :obj:`True`, will use the token generated when running"
":obj:`huggingface-cli login` as HTTP bearer authorization for remote files."
)
},
)
unk_token: str = field(
default="[UNK]",
metadata={"help": "The unk token for the tokenizer"},
)
pad_token: str = field(
default="[PAD]",
metadata={"help": "The padding token for the tokenizer"},
)
word_delimiter_token: str = field(
default="|",
metadata={"help": "The word delimiter token for the tokenizer"},
)
phoneme_language: Optional[str] = field(
default=None,
metadata={
"help": (
"The target language that should be used be"
" passed to the tokenizer for tokenization. Note that"
" this is only relevant if the model classifies the"
" input audio to a sequence of phoneme sequences."
)
},
)
@dataclass
class DataCollatorCTCWithPadding:
"""
Data collator that will dynamically pad the inputs received.
Args:
processor (:class:`~transformers.AutoProcessor`)
The processor used for proccessing the data.
padding (:obj:`bool`, :obj:`str` or :class:`~transformers.tokenization_utils_base.PaddingStrategy`, `optional`, defaults to :obj:`True`):
Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
among:
* :obj:`True` or :obj:`'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
sequence if provided).
* :obj:`'max_length'`: Pad to a maximum length specified with the argument :obj:`max_length` or to the
maximum acceptable input length for the model if that argument is not provided.
* :obj:`False` or :obj:`'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of
different lengths).
max_length (:obj:`int`, `optional`):
Maximum length of the ``input_values`` of the returned list and optionally padding length (see above).
max_length_labels (:obj:`int`, `optional`):
Maximum length of the ``labels`` returned list and optionally padding length (see above).
pad_to_multiple_of (:obj:`int`, `optional`):
If set will pad the sequence to a multiple of the provided value.
This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
7.5 (Volta).
"""
processor: AutoProcessor
padding: Union[bool, str] = "longest"
pad_to_multiple_of: Optional[int] = None
pad_to_multiple_of_labels: Optional[int] = None
def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
# split inputs and labels since they have to be of different lengths and need
# different padding methods
input_features = [{"input_values": feature["input_values"]} for feature in features]
label_features = [{"input_ids": feature["labels"]} for feature in features]
batch = self.processor.pad(
input_features,
padding=self.padding,
pad_to_multiple_of=self.pad_to_multiple_of,
return_tensors="pt",
)
labels_batch = self.processor.pad(
labels=label_features,
padding=self.padding,
pad_to_multiple_of=self.pad_to_multiple_of_labels,
return_tensors="pt",
)
# replace padding with -100 to ignore loss correctly
labels = labels_batch["input_ids"].masked_fill(labels_batch.attention_mask.ne(1), -100)
batch["labels"] = labels
return batch
def create_vocabulary_from_data(
datasets: DatasetDict,
word_delimiter_token: Optional[str] = None,
unk_token: Optional[str] = None,
pad_token: Optional[str] = None,
):
# Given training and test labels create vocabulary
def extract_all_chars(batch):
all_text = " ".join(batch["target_text"])
vocab = list(set(all_text))
return {"vocab": [vocab], "all_text": [all_text]}
vocabs = datasets.map(
extract_all_chars,
batched=True,
batch_size=-1,
keep_in_memory=True,
remove_columns=datasets["train"].column_names,
)
# take union of all unique characters in each dataset
vocab_set = functools.reduce(
lambda vocab_1, vocab_2: set(vocab_1["vocab"][0]) | set(vocab_2["vocab"][0]), vocabs.values()
)
vocab_dict = {v: k for k, v in enumerate(sorted(vocab_set))}
# replace white space with delimiter token
if word_delimiter_token is not None:
vocab_dict[word_delimiter_token] = vocab_dict[" "]
del vocab_dict[" "]
# add unk and pad token
if unk_token is not None:
vocab_dict[unk_token] = len(vocab_dict)
if pad_token is not None:
vocab_dict[pad_token] = len(vocab_dict)
return vocab_dict
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()
# 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:
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."
)
# 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)],
)
logger.setLevel(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()
logger.info("Training/evaluation parameters %s", training_args)
# Set seed before initializing model.
set_seed(training_args.seed)
# 1. First, let's load the dataset
raw_datasets = DatasetDict()
if training_args.do_train:
raw_datasets["train"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=data_args.train_split_name,
token=data_args.use_auth_token,
)
if data_args.audio_column_name not in raw_datasets["train"].column_names:
raise ValueError(
f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'."
" Make sure to set `--audio_column_name` to the correct audio column - one of"
f" {', '.join(raw_datasets['train'].column_names)}."
)
if data_args.text_column_name not in raw_datasets["train"].column_names:
raise ValueError(
f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
"Make sure to set `--text_column_name` to the correct text column - one of "
f"{', '.join(raw_datasets['train'].column_names)}."
)
if data_args.max_train_samples is not None:
raw_datasets["train"] = raw_datasets["train"].select(range(data_args.max_train_samples))
if training_args.do_eval:
raw_datasets["eval"] = load_dataset(
data_args.dataset_name,
data_args.dataset_config_name,
split=data_args.eval_split_name,
token=data_args.use_auth_token,
)
if data_args.max_eval_samples is not None:
raw_datasets["eval"] = raw_datasets["eval"].select(range(data_args.max_eval_samples))
# 2. We remove some special characters from the datasets
# that make training complicated and do not help in transcribing the speech
# E.g. characters, such as `,` and `.` do not really have an acoustic characteristic
# that could be easily picked up by the model
chars_to_ignore_regex = (
f'[{"".join(data_args.chars_to_ignore)}]' if data_args.chars_to_ignore is not None else None
)
text_column_name = data_args.text_column_name
def remove_special_characters(batch):
if chars_to_ignore_regex is not None:
batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[text_column_name]).lower() + " "
else:
batch["target_text"] = batch[text_column_name].lower() + " "
return batch
with training_args.main_process_first(desc="dataset map special characters removal"):
raw_datasets = raw_datasets.map(
remove_special_characters,
remove_columns=[text_column_name],
desc="remove special characters from datasets",
)
# save special tokens for tokenizer
word_delimiter_token = data_args.word_delimiter_token
unk_token = data_args.unk_token
pad_token = data_args.pad_token
# 3. Next, let's load the config as we might need it to create
# the tokenizer
# load config
config = AutoConfig.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, token=data_args.use_auth_token
)
# 4. Next, if no tokenizer file is defined,
# we create the vocabulary of the model by extracting all unique characters from
# the training and evaluation datasets
# We need to make sure that only first rank saves vocabulary
# make sure all processes wait until vocab is created
tokenizer_name_or_path = model_args.tokenizer_name_or_path
tokenizer_kwargs = {}
if tokenizer_name_or_path is None:
# save vocab in training output dir
tokenizer_name_or_path = training_args.output_dir
vocab_file = os.path.join(tokenizer_name_or_path, "vocab.json")
with training_args.main_process_first():
if training_args.overwrite_output_dir and os.path.isfile(vocab_file):
os.remove(vocab_file)
with training_args.main_process_first(desc="dataset map vocabulary creation"):
if not os.path.isfile(vocab_file):
os.makedirs(tokenizer_name_or_path, exist_ok=True)
vocab_dict = create_vocabulary_from_data(
raw_datasets,
word_delimiter_token=word_delimiter_token,
unk_token=unk_token,
pad_token=pad_token,
)
# save vocab dict to be loaded into tokenizer
with open(vocab_file, "w") as file:
json.dump(vocab_dict, file)
# if tokenizer has just been created
# it is defined by `tokenizer_class` if present in config else by `model_type`
tokenizer_kwargs = {
"config": config if config.tokenizer_class is not None else None,
"tokenizer_type": config.model_type if config.tokenizer_class is None else None,
"unk_token": unk_token,
"pad_token": pad_token,
"word_delimiter_token": word_delimiter_token,
}
# 5. Now we can instantiate the feature extractor, tokenizer and model
# Note for distributed training, the .from_pretrained methods guarantee that only
# one local process can concurrently download model & vocab.
# load feature_extractor and tokenizer
tokenizer = AutoTokenizer.from_pretrained(
tokenizer_name_or_path,
token=data_args.use_auth_token,
**tokenizer_kwargs,
)
feature_extractor = AutoFeatureExtractor.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, token=data_args.use_auth_token
)
# adapt config
config.update(
{
"feat_proj_dropout": model_args.feat_proj_dropout,
"attention_dropout": model_args.attention_dropout,
"hidden_dropout": model_args.hidden_dropout,
"final_dropout": model_args.final_dropout,
"mask_time_prob": model_args.mask_time_prob,
"mask_time_length": model_args.mask_time_length,
"mask_feature_prob": model_args.mask_feature_prob,
"mask_feature_length": model_args.mask_feature_length,
"gradient_checkpointing": training_args.gradient_checkpointing,
"layerdrop": model_args.layerdrop,
"ctc_loss_reduction": model_args.ctc_loss_reduction,
"pad_token_id": tokenizer.pad_token_id,
"vocab_size": len(tokenizer),
"activation_dropout": model_args.activation_dropout,
}
)
# create model
model = AutoModelForCTC.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
config=config,
token=data_args.use_auth_token,
)
# freeze encoder
if model_args.freeze_feature_encoder:
model.freeze_feature_encoder()
# 6. Now we preprocess the datasets including loading the audio, resampling and normalization
# Thankfully, `datasets` takes care of automatically loading and resampling the audio,
# so that we just need to set the correct target sampling rate and normalize the input
# via the `feature_extractor`
# make sure that dataset decodes audio with correct sampling rate
dataset_sampling_rate = next(iter(raw_datasets.values())).features[data_args.audio_column_name].sampling_rate
if dataset_sampling_rate != feature_extractor.sampling_rate:
raw_datasets = raw_datasets.cast_column(
data_args.audio_column_name, datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
)
# derive max & min input length for sample rate & max duration
max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
audio_column_name = data_args.audio_column_name
num_workers = data_args.preprocessing_num_workers
# `phoneme_language` is only relevant if the model is fine-tuned on phoneme classification
phoneme_language = data_args.phoneme_language
# Preprocessing the datasets.
# We need to read the audio files as arrays and tokenize the targets.
def prepare_dataset(batch):
# load audio
sample = batch[audio_column_name]
inputs = feature_extractor(sample["array"], sampling_rate=sample["sampling_rate"])
batch["input_values"] = inputs.input_values[0]
batch["input_length"] = len(batch["input_values"])
# encode targets
additional_kwargs = {}
if phoneme_language is not None:
additional_kwargs["phonemizer_lang"] = phoneme_language
batch["labels"] = tokenizer(batch["target_text"], **additional_kwargs).input_ids
return batch
with training_args.main_process_first(desc="dataset map preprocessing"):
vectorized_datasets = raw_datasets.map(
prepare_dataset,
remove_columns=next(iter(raw_datasets.values())).column_names,
num_proc=num_workers,
desc="preprocess datasets",
)
def is_audio_in_length_range(length):
return length > min_input_length and length < max_input_length
# filter data that is shorter than min_input_length
vectorized_datasets = vectorized_datasets.filter(
is_audio_in_length_range,
num_proc=num_workers,
input_columns=["input_length"],
)
# 7. Next, we can prepare the training.
# Let's use word error rate (WER) as our evaluation metric,
# instantiate a data collator and the trainer
# Define evaluation metrics during training, *i.e.* word error rate, character error rate
eval_metrics = {metric: load_metric(metric) for metric in data_args.eval_metrics}
# for large datasets it is advised to run the preprocessing on a
# single machine first with ``args.preprocessing_only`` since there will mostly likely
# be a timeout when running the script in distributed mode.
# In a second step ``args.preprocessing_only`` can then be set to `False` to load the
# cached dataset
if data_args.preprocessing_only:
logger.info(f"Data preprocessing finished. Files cached at {vectorized_datasets.cache_files}")
return
def compute_metrics(pred):
pred_logits = pred.predictions
pred_ids = np.argmax(pred_logits, axis=-1)
pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
pred_str = tokenizer.batch_decode(pred_ids)
# we do not want to group tokens when computing the metrics
label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)
metrics = {k: v.compute(predictions=pred_str, references=label_str) for k, v in eval_metrics.items()}
return metrics
# Now save everything to be able to create a single processor later
if is_main_process(training_args.local_rank):
# save feature extractor, tokenizer and config
feature_extractor.save_pretrained(training_args.output_dir)
tokenizer.save_pretrained(training_args.output_dir)
config.save_pretrained(training_args.output_dir)
try:
processor = AutoProcessor.from_pretrained(training_args.output_dir)
except (OSError, KeyError):
warnings.warn(
"Loading a processor from a feature extractor config that does not"
" include a `processor_class` attribute is deprecated and will be removed in v5. Please add the following "
" attribute to your `preprocessor_config.json` file to suppress this warning: "
" `'processor_class': 'Wav2Vec2Processor'`",
FutureWarning,
)
processor = Wav2Vec2Processor.from_pretrained(training_args.output_dir)
# Instantiate custom data collator
data_collator = DataCollatorCTCWithPadding(processor=processor)
decay_parameters = get_parameter_names(model, [torch.nn.LayerNorm])
decay_parameters = [name for name in decay_parameters if "bias" not in name]
optimizer_grouped_parameters = [
{
"params": [p for n, p in model.named_parameters() if n in decay_parameters],
"weight_decay": training_args.weight_decay,
},
{
"params": [p for n, p in model.named_parameters() if n not in decay_parameters],
"weight_decay": 0.0,
},
]
optimizer = bnb.optim.Adam8bit(
params=optimizer_grouped_parameters,
lr=training_args.learning_rate,
betas=(training_args.adam_beta1, training_args.adam_beta2),
eps=training_args.adam_epsilon,
)
optimizers = (optimizer, None)
# Initialize Trainer
trainer = Trainer(
model=model,
data_collator=data_collator,
args=training_args,
compute_metrics=compute_metrics,
train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
tokenizer=feature_extractor,
optimizers=optimizers,
)
# 8. Finally, we can start training
# Training
if training_args.do_train:
# use last checkpoint if exist
if last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model()
metrics = train_result.metrics
max_train_samples = (
data_args.max_train_samples
if data_args.max_train_samples is not None
else len(vectorized_datasets["train"])
)
metrics["train_samples"] = min(max_train_samples, len(vectorized_datasets["train"]))
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
metrics = trainer.evaluate()
max_eval_samples = (
data_args.max_eval_samples if data_args.max_eval_samples is not None else len(vectorized_datasets["eval"])
)
metrics["eval_samples"] = min(max_eval_samples, len(vectorized_datasets["eval"]))
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
# Write model card and (optionally) push to hub
config_name = data_args.dataset_config_name if data_args.dataset_config_name is not None else "na"
kwargs = {
"finetuned_from": model_args.model_name_or_path,
"tasks": "automatic-speech-recognition",
"tags": ["automatic-speech-recognition", data_args.dataset_name],
"dataset_args": (
f"Config: {config_name}, Training split: {data_args.train_split_name}, Eval split:"
f" {data_args.eval_split_name}"
),
"dataset": f"{data_args.dataset_name.upper()} - {config_name.upper()}",
}
if "common_voice" in data_args.dataset_name:
kwargs["language"] = config_name
if training_args.push_to_hub:
trainer.push_to_hub(**kwargs)
else:
trainer.create_model_card(**kwargs)
return results
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/robust-speech-event/eval.py
|
#!/usr/bin/env python3
import argparse
import re
from typing import Dict
import torch
from datasets import Audio, Dataset, load_dataset, load_metric
from transformers import AutoFeatureExtractor, pipeline
def log_results(result: Dataset, args: Dict[str, str]):
"""DO NOT CHANGE. This function computes and logs the result metrics."""
log_outputs = args.log_outputs
dataset_id = "_".join(args.dataset.split("/") + [args.config, args.split])
# load metric
wer = load_metric("wer")
cer = load_metric("cer")
# compute metrics
wer_result = wer.compute(references=result["target"], predictions=result["prediction"])
cer_result = cer.compute(references=result["target"], predictions=result["prediction"])
# print & log results
result_str = f"WER: {wer_result}\nCER: {cer_result}"
print(result_str)
with open(f"{dataset_id}_eval_results.txt", "w") as f:
f.write(result_str)
# log all results in text file. Possibly interesting for analysis
if log_outputs is not None:
pred_file = f"log_{dataset_id}_predictions.txt"
target_file = f"log_{dataset_id}_targets.txt"
with open(pred_file, "w") as p, open(target_file, "w") as t:
# mapping function to write output
def write_to_file(batch, i):
p.write(f"{i}" + "\n")
p.write(batch["prediction"] + "\n")
t.write(f"{i}" + "\n")
t.write(batch["target"] + "\n")
result.map(write_to_file, with_indices=True)
def normalize_text(text: str) -> str:
"""DO ADAPT FOR YOUR USE CASE. this function normalizes the target text."""
chars_to_ignore_regex = '[,?.!\-\;\:"“%‘”�—’…–]' # noqa: W605 IMPORTANT: this should correspond to the chars that were ignored during training
text = re.sub(chars_to_ignore_regex, "", text.lower())
# In addition, we can normalize the target text, e.g. removing new lines characters etc...
# note that order is important here!
token_sequences_to_ignore = ["\n\n", "\n", " ", " "]
for t in token_sequences_to_ignore:
text = " ".join(text.split(t))
return text
def main(args):
# load dataset
dataset = load_dataset(args.dataset, args.config, split=args.split, token=True)
# for testing: only process the first two examples as a test
# dataset = dataset.select(range(10))
# load processor
feature_extractor = AutoFeatureExtractor.from_pretrained(args.model_id)
sampling_rate = feature_extractor.sampling_rate
# resample audio
dataset = dataset.cast_column("audio", Audio(sampling_rate=sampling_rate))
# load eval pipeline
if args.device is None:
args.device = 0 if torch.cuda.is_available() else -1
asr = pipeline("automatic-speech-recognition", model=args.model_id, device=args.device)
# map function to decode audio
def map_to_pred(batch):
prediction = asr(
batch["audio"]["array"], chunk_length_s=args.chunk_length_s, stride_length_s=args.stride_length_s
)
batch["prediction"] = prediction["text"]
batch["target"] = normalize_text(batch["sentence"])
return batch
# run inference on all examples
result = dataset.map(map_to_pred, remove_columns=dataset.column_names)
# compute and log_results
# do not change function below
log_results(result, args)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--model_id", type=str, required=True, help="Model identifier. Should be loadable with 🤗 Transformers"
)
parser.add_argument(
"--dataset",
type=str,
required=True,
help="Dataset name to evaluate the `model_id`. Should be loadable with 🤗 Datasets",
)
parser.add_argument(
"--config", type=str, required=True, help="Config of the dataset. *E.g.* `'en'` for Common Voice"
)
parser.add_argument("--split", type=str, required=True, help="Split of the dataset. *E.g.* `'test'`")
parser.add_argument(
"--chunk_length_s", type=float, default=None, help="Chunk length in seconds. Defaults to 5 seconds."
)
parser.add_argument(
"--stride_length_s", type=float, default=None, help="Stride of the audio chunks. Defaults to 1 second."
)
parser.add_argument(
"--log_outputs", action="store_true", help="If defined, write outputs to log file for analysis."
)
parser.add_argument(
"--device",
type=int,
default=None,
help="The device to run the pipeline on. -1 for CPU (default), 0 for the first GPU and so on.",
)
args = parser.parse_args()
main(args)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/robust-speech-event/run_speech_recognition_ctc_streaming.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2022 The HuggingFace Inc. team. 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
""" Fine-tuning a 🤗 Transformers CTC model for automatic speech recognition in streaming mode"""
import logging
import os
import re
import sys
import warnings
from dataclasses import dataclass, field
from typing import Dict, List, Optional, Union
import datasets
import numpy as np
import torch
from datasets import IterableDatasetDict, interleave_datasets, load_dataset, load_metric
from torch.utils.data import IterableDataset
import transformers
from transformers import (
AutoConfig,
AutoFeatureExtractor,
AutoModelForCTC,
AutoProcessor,
AutoTokenizer,
HfArgumentParser,
Trainer,
TrainerCallback,
TrainingArguments,
Wav2Vec2Processor,
set_seed,
)
from transformers.trainer_pt_utils import IterableDatasetShard
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risk.
check_min_version("4.17.0.dev0")
require_version("datasets>=1.18.2", "To fix: pip install 'datasets>=1.18.2'")
logger = logging.getLogger(__name__)
def list_field(default=None, metadata=None):
return field(default_factory=lambda: default, metadata=metadata)
@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"}
)
tokenizer_name_or_path: Optional[str] = field(
default=None,
metadata={"help": "Path to pretrained tokenizer or tokenizer identifier from huggingface.co/models"},
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
)
freeze_feature_encoder: bool = field(
default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
)
attention_dropout: float = field(
default=0.0, metadata={"help": "The dropout ratio for the attention probabilities."}
)
activation_dropout: float = field(
default=0.0, metadata={"help": "The dropout ratio for activations inside the fully connected layer."}
)
feat_proj_dropout: float = field(default=0.0, metadata={"help": "The dropout ratio for the projected features."})
hidden_dropout: float = field(
default=0.0,
metadata={
"help": "The dropout probability for all fully connected layers in the embeddings, encoder, and pooler."
},
)
final_dropout: float = field(
default=0.0,
metadata={"help": "The dropout probability for the final projection layer."},
)
mask_time_prob: float = field(
default=0.05,
metadata={
"help": (
"Probability of each feature vector along the time axis to be chosen as the start of the vector "
"span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature "
"vectors will be masked along the time axis."
)
},
)
mask_time_length: int = field(
default=10,
metadata={"help": "Length of vector span to mask along the time axis."},
)
mask_feature_prob: float = field(
default=0.0,
metadata={
"help": (
"Probability of each feature vector along the feature axis to be chosen as the start of the vectorspan"
" to be masked. Approximately ``mask_feature_prob * sequence_length // mask_feature_length`` feature"
" bins will be masked along the time axis."
)
},
)
mask_feature_length: int = field(
default=10,
metadata={"help": "Length of vector span to mask along the feature axis."},
)
layerdrop: float = field(default=0.0, metadata={"help": "The LayerDrop probability."})
ctc_loss_reduction: Optional[str] = field(
default="mean", metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."}
)
@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.
"""
dataset_name: str = field(
metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
dataset_config_name: str = field(
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
train_split_name: str = field(
default="train+validation",
metadata={
"help": (
"The name of the training data set split to use (via the datasets library). Defaults to "
"'train+validation'"
)
},
)
eval_split_name: str = field(
default="test",
metadata={
"help": "The name of the training data set split to use (via the datasets library). Defaults to 'test'"
},
)
audio_column_name: str = field(
default="audio",
metadata={"help": "The name of the dataset column containing the audio data. Defaults to 'audio'"},
)
text_column_name: str = field(
default="text",
metadata={"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
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 validation examples to this "
"value if set."
)
},
)
shuffle_buffer_size: Optional[int] = field(
default=500,
metadata={
"help": (
"The number of streamed examples to download before shuffling them. The large the buffer, "
"the closer it is to real offline shuffling."
)
},
)
chars_to_ignore: Optional[List[str]] = list_field(
default=None,
metadata={"help": "A list of characters to remove from the transcripts."},
)
eval_metrics: List[str] = list_field(
default=["wer"],
metadata={"help": "A list of metrics the model should be evaluated on. E.g. `'wer cer'`"},
)
max_duration_in_seconds: float = field(
default=20.0,
metadata={"help": "Filter audio files that are longer than `max_duration_in_seconds` seconds."},
)
preprocessing_only: bool = field(
default=False,
metadata={
"help": (
"Whether to only do data preprocessing and skip training. This is especially useful when data"
" preprocessing errors out in distributed training due to timeout. In this case, one should run the"
" preprocessing in a non-distributed setup with `preprocessing_only=True` so that the cached datasets"
" can consequently be loaded in distributed training"
)
},
)
use_auth_token: bool = field(
default=False,
metadata={
"help": (
"If :obj:`True`, will use the token generated when running"
":obj:`huggingface-cli login` as HTTP bearer authorization for remote files."
)
},
)
phoneme_language: Optional[str] = field(
default=None,
metadata={
"help": (
"The target language that should be used be"
" passed to the tokenizer for tokenization. Note that"
" this is only relevant if the model classifies the"
" input audio to a sequence of phoneme sequences."
)
},
)
@dataclass
class DataCollatorCTCWithPadding:
"""
Data collator that will dynamically pad the inputs received.
Args:
processor (:class:`~transformers.AutoProcessor`)
The processor used for proccessing the data.
padding (:obj:`bool`, :obj:`str` or :class:`~transformers.tokenization_utils_base.PaddingStrategy`, `optional`, defaults to :obj:`True`):
Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
among:
* :obj:`True` or :obj:`'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
sequence if provided).
* :obj:`'max_length'`: Pad to a maximum length specified with the argument :obj:`max_length` or to the
maximum acceptable input length for the model if that argument is not provided.
* :obj:`False` or :obj:`'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of
different lengths).
max_length (:obj:`int`, `optional`):
Maximum length of the ``input_values`` of the returned list and optionally padding length (see above).
max_length_labels (:obj:`int`, `optional`):
Maximum length of the ``labels`` returned list and optionally padding length (see above).
pad_to_multiple_of (:obj:`int`, `optional`):
If set will pad the sequence to a multiple of the provided value.
This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
7.5 (Volta).
"""
processor: AutoProcessor
padding: Union[bool, str] = "longest"
max_length: Optional[int] = None
pad_to_multiple_of: Optional[int] = None
pad_to_multiple_of_labels: Optional[int] = None
def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
# split inputs and labels since they have to be of different lengths and need
# different padding methods
input_features = []
label_features = []
for feature in features:
if self.max_length and feature["input_values"].shape[-1] > self.max_length:
continue
input_features.append({"input_values": feature["input_values"]})
label_features.append({"input_ids": feature["labels"]})
batch = self.processor.pad(
input_features,
padding=self.padding,
pad_to_multiple_of=self.pad_to_multiple_of,
return_tensors="pt",
)
labels_batch = self.processor.pad(
labels=label_features,
padding=self.padding,
pad_to_multiple_of=self.pad_to_multiple_of_labels,
return_tensors="pt",
)
# replace padding with -100 to ignore loss correctly
labels = labels_batch["input_ids"].masked_fill(labels_batch.attention_mask.ne(1), -100)
batch["labels"] = labels
return batch
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()
# 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:
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."
)
# 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)],
)
logger.setLevel(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()
logger.info("Training/evaluation parameters %s", training_args)
# Set seed before initializing model.
set_seed(training_args.seed)
# 1. First, let's load the dataset
raw_datasets = IterableDatasetDict()
raw_column_names = {}
def load_streaming_dataset(split, sampling_rate, **kwargs):
if "+" in split:
dataset_splits = [load_dataset(split=split_name, **kwargs) for split_name in split.split("+")]
# `features` and `cast_column` won't be available after interleaving, so we'll use them here
features = dataset_splits[0].features
# make sure that the dataset decodes audio with a correct sampling rate
dataset_splits = [
dataset.cast_column(data_args.audio_column_name, datasets.features.Audio(sampling_rate=sampling_rate))
for dataset in dataset_splits
]
interleaved_dataset = interleave_datasets(dataset_splits)
return interleaved_dataset, features
else:
dataset = load_dataset(split=split, **kwargs)
features = dataset.features
# make sure that the dataset decodes audio with a correct sampling rate
dataset = dataset.cast_column(
data_args.audio_column_name, datasets.features.Audio(sampling_rate=sampling_rate)
)
return dataset, features
# `datasets` takes care of automatically loading and resampling the audio,
# so we just need to set the correct target sampling rate and normalize the input
# via the `feature_extractor`
feature_extractor = AutoFeatureExtractor.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, token=data_args.use_auth_token
)
if training_args.do_train:
raw_datasets["train"], train_features = load_streaming_dataset(
path=data_args.dataset_name,
name=data_args.dataset_config_name,
split=data_args.train_split_name,
token=data_args.use_auth_token,
streaming=True,
sampling_rate=feature_extractor.sampling_rate,
)
raw_column_names["train"] = list(train_features.keys())
if data_args.audio_column_name not in raw_column_names["train"]:
raise ValueError(
f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_name}'."
" Make sure to set `--audio_column_name` to the correct audio column - one of"
f" {', '.join(raw_column_names['train'])}."
)
if data_args.text_column_name not in raw_column_names["train"]:
raise ValueError(
f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
"Make sure to set `--text_column_name` to the correct text column - one of "
f"{', '.join(raw_column_names['train'])}."
)
if data_args.max_train_samples is not None:
raw_datasets["train"] = raw_datasets["train"].take(range(data_args.max_train_samples))
if training_args.do_eval:
raw_datasets["eval"], eval_features = load_streaming_dataset(
path=data_args.dataset_name,
name=data_args.dataset_config_name,
split=data_args.eval_split_name,
token=data_args.use_auth_token,
streaming=True,
sampling_rate=feature_extractor.sampling_rate,
)
raw_column_names["eval"] = list(eval_features.keys())
if data_args.max_eval_samples is not None:
raw_datasets["eval"] = raw_datasets["eval"].take(range(data_args.max_eval_samples))
# 2. We remove some special characters from the datasets
# that make training complicated and do not help in transcribing the speech
# E.g. characters, such as `,` and `.` do not really have an acoustic characteristic
# that could be easily picked up by the model
chars_to_ignore_regex = (
f'[{"".join(data_args.chars_to_ignore)}]' if data_args.chars_to_ignore is not None else None
)
text_column_name = data_args.text_column_name
def remove_special_characters(batch):
if chars_to_ignore_regex is not None:
batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[text_column_name]).lower() + " "
else:
batch["target_text"] = batch[text_column_name].lower() + " "
return batch
with training_args.main_process_first(desc="dataset map special characters removal"):
for split, dataset in raw_datasets.items():
raw_datasets[split] = dataset.map(
remove_special_characters,
).remove_columns([text_column_name])
# 3. Next, let's load the config as we might need it to create
# the tokenizer
config = AutoConfig.from_pretrained(
model_args.model_name_or_path, cache_dir=model_args.cache_dir, token=data_args.use_auth_token
)
# 4. Now we can instantiate the tokenizer and model
# Note for distributed training, the .from_pretrained methods guarantee that only
# one local process can concurrently download model & vocab.
tokenizer_name_or_path = model_args.tokenizer_name_or_path
if tokenizer_name_or_path is None:
raise ValueError(
"Tokenizer has to be created before training in streaming mode. Please specify --tokenizer_name_or_path"
)
# load feature_extractor and tokenizer
tokenizer = AutoTokenizer.from_pretrained(
tokenizer_name_or_path,
config=config,
token=data_args.use_auth_token,
)
# adapt config
config.update(
{
"feat_proj_dropout": model_args.feat_proj_dropout,
"attention_dropout": model_args.attention_dropout,
"hidden_dropout": model_args.hidden_dropout,
"final_dropout": model_args.final_dropout,
"mask_time_prob": model_args.mask_time_prob,
"mask_time_length": model_args.mask_time_length,
"mask_feature_prob": model_args.mask_feature_prob,
"mask_feature_length": model_args.mask_feature_length,
"gradient_checkpointing": training_args.gradient_checkpointing,
"layerdrop": model_args.layerdrop,
"ctc_loss_reduction": model_args.ctc_loss_reduction,
"pad_token_id": tokenizer.pad_token_id,
"vocab_size": len(tokenizer),
"activation_dropout": model_args.activation_dropout,
}
)
# create model
model = AutoModelForCTC.from_pretrained(
model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
config=config,
token=data_args.use_auth_token,
)
# freeze encoder
if model_args.freeze_feature_encoder:
model.freeze_feature_encoder()
# 5. Now we preprocess the datasets including loading the audio, resampling and normalization
audio_column_name = data_args.audio_column_name
# `phoneme_language` is only relevant if the model is fine-tuned on phoneme classification
phoneme_language = data_args.phoneme_language
# Preprocessing the datasets.
# We need to read the audio files as arrays and tokenize the targets.
def prepare_dataset(batch):
# load audio
sample = batch[audio_column_name]
inputs = feature_extractor(sample["array"], sampling_rate=sample["sampling_rate"])
batch["input_values"] = inputs.input_values[0]
batch["input_length"] = len(batch["input_values"])
# encode targets
additional_kwargs = {}
if phoneme_language is not None:
additional_kwargs["phonemizer_lang"] = phoneme_language
batch["labels"] = tokenizer(batch["target_text"], **additional_kwargs).input_ids
return batch
vectorized_datasets = IterableDatasetDict()
with training_args.main_process_first(desc="dataset map preprocessing"):
for split, dataset in raw_datasets.items():
vectorized_datasets[split] = (
dataset.map(prepare_dataset)
.remove_columns(raw_column_names[split] + ["target_text"])
.with_format("torch")
)
if split == "train":
vectorized_datasets[split] = vectorized_datasets[split].shuffle(
buffer_size=data_args.shuffle_buffer_size,
seed=training_args.seed,
)
# 6. Next, we can prepare the training.
# Let's use word error rate (WER) as our evaluation metric,
# instantiate a data collator and the trainer
# Define evaluation metrics during training, *i.e.* word error rate, character error rate
eval_metrics = {metric: load_metric(metric) for metric in data_args.eval_metrics}
def compute_metrics(pred):
pred_logits = pred.predictions
pred_ids = np.argmax(pred_logits, axis=-1)
pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
pred_str = tokenizer.batch_decode(pred_ids)
# we do not want to group tokens when computing the metrics
label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)
metrics = {k: v.compute(predictions=pred_str, references=label_str) for k, v in eval_metrics.items()}
return metrics
# Now save everything to be able to create a single processor later
if is_main_process(training_args.local_rank):
# save feature extractor, tokenizer and config
feature_extractor.save_pretrained(training_args.output_dir)
tokenizer.save_pretrained(training_args.output_dir)
config.save_pretrained(training_args.output_dir)
try:
processor = AutoProcessor.from_pretrained(training_args.output_dir)
except (OSError, KeyError):
warnings.warn(
"Loading a processor from a feature extractor config that does not"
" include a `processor_class` attribute is deprecated and will be removed in v5. Please add the following "
" attribute to your `preprocessor_config.json` file to suppress this warning: "
" `'processor_class': 'Wav2Vec2Processor'`",
FutureWarning,
)
processor = Wav2Vec2Processor.from_pretrained(training_args.output_dir)
# Instantiate custom data collator
max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
data_collator = DataCollatorCTCWithPadding(processor=processor, max_length=max_input_length)
# trainer callback to reinitialize and reshuffle the streamable datasets at the beginning of each epoch
class ShuffleCallback(TrainerCallback):
def on_epoch_begin(self, args, state, control, train_dataloader, **kwargs):
if isinstance(train_dataloader.dataset, IterableDatasetShard):
pass # set_epoch() is handled by the Trainer
elif isinstance(train_dataloader.dataset, IterableDataset):
train_dataloader.dataset.set_epoch(train_dataloader.dataset._epoch + 1)
# Initialize Trainer
trainer = Trainer(
model=model,
data_collator=data_collator,
args=training_args,
compute_metrics=compute_metrics,
train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
tokenizer=processor,
callbacks=[ShuffleCallback()],
)
# 7. Finally, we can start training
# Training
if training_args.do_train:
# use last checkpoint if exist
if last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model()
metrics = train_result.metrics
if data_args.max_train_samples:
metrics["train_samples"] = data_args.max_train_samples
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
metrics = trainer.evaluate()
if data_args.max_eval_samples:
metrics["eval_samples"] = data_args.max_eval_samples
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
# Write model card and (optionally) push to hub
config_name = data_args.dataset_config_name if data_args.dataset_config_name is not None else "na"
kwargs = {
"finetuned_from": model_args.model_name_or_path,
"tasks": "automatic-speech-recognition",
"tags": ["automatic-speech-recognition", data_args.dataset_name],
"dataset_args": (
f"Config: {config_name}, Training split: {data_args.train_split_name}, Eval split:"
f" {data_args.eval_split_name}"
),
"dataset": f"{data_args.dataset_name.upper()} - {config_name.upper()}",
}
if "common_voice" in data_args.dataset_name:
kwargs["language"] = config_name
if training_args.push_to_hub:
trainer.push_to_hub(**kwargs)
else:
trainer.create_model_card(**kwargs)
return results
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/luke/README.md
|
# Token classification
## PyTorch version, no Trainer
Fine-tuning (m)LUKE for token classification task such as Named Entity Recognition (NER), Parts-of-speech
tagging (POS) or phrase extraction (CHUNKS). You can easily
customize it to your needs if you need extra processing on your datasets.
It will either run on a datasets hosted on our [hub](https://huggingface.co/datasets) or with your own text files for
training and validation, you might just need to add some tweaks in the data preprocessing.
The script can be run in a distributed setup, on TPU and supports mixed precision by
the mean of the [🤗 `Accelerate`](https://github.com/huggingface/accelerate) library. You can use the script normally
after installing it:
```bash
pip install git+https://github.com/huggingface/accelerate
```
then to train English LUKE on CoNLL2003:
```bash
export TASK_NAME=ner
python run_luke_ner_no_trainer.py \
--model_name_or_path studio-ousia/luke-base \
--dataset_name conll2003 \
--task_name $TASK_NAME \
--max_length 128 \
--per_device_train_batch_size 32 \
--learning_rate 2e-5 \
--num_train_epochs 3 \
--output_dir /tmp/$TASK_NAME/
```
You can then use your usual launchers to run in it in a distributed environment, but the easiest way is to run
```bash
accelerate config
```
and reply to the questions asked. Then
```bash
accelerate test
```
that will check everything is ready for training. Finally, you can launch training with
```bash
export TASK_NAME=ner
accelerate launch run_ner_no_trainer.py \
--model_name_or_path studio-ousia/luke-base \
--dataset_name conll2003 \
--task_name $TASK_NAME \
--max_length 128 \
--per_device_train_batch_size 32 \
--learning_rate 2e-5 \
--num_train_epochs 3 \
--output_dir /tmp/$TASK_NAME/
```
This command is the same and will work for:
- a CPU-only setup
- a setup with one GPU
- a distributed training with several GPUs (single or multi node)
- a training on TPUs
Note that this library is in alpha release so your feedback is more than welcome if you encounter any problem using it.
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/luke/run_luke_ner_no_trainer.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2022 The HuggingFace Inc. team. 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.
"""
Fine-tuning (m)LUKE model on token classification tasks (NER, POS, CHUNKS) relying on the accelerate library 🤗
without using a Trainer.
"""
import argparse
import logging
import math
import os
import random
from pathlib import Path
import datasets
import torch
from accelerate import Accelerator, DistributedDataParallelKwargs
from datasets import ClassLabel, load_dataset, load_metric
from huggingface_hub import Repository, create_repo
from luke_utils import DataCollatorForLukeTokenClassification, is_punctuation, padding_tensor
from torch.utils.data import DataLoader
from tqdm.auto import tqdm
import transformers
from transformers import (
AdamW,
LukeConfig,
LukeForEntitySpanClassification,
LukeTokenizer,
SchedulerType,
default_data_collator,
get_scheduler,
set_seed,
)
from transformers.utils.versions import require_version
logger = logging.getLogger(__name__)
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/token-classification/requirements.txt")
def parse_args():
parser = argparse.ArgumentParser(
description="Finetune (m)LUKE on a token classification task (such as NER) with the accelerate library"
)
parser.add_argument(
"--dataset_name",
type=str,
default=None,
help="The name of the dataset to use (via the datasets library).",
)
parser.add_argument(
"--dataset_config_name",
type=str,
default=None,
help="The configuration name of the dataset to use (via the datasets library).",
)
parser.add_argument(
"--train_file", type=str, default=None, help="A csv or a json file containing the training data."
)
parser.add_argument(
"--validation_file", type=str, default=None, help="A csv or a json file containing the validation data."
)
parser.add_argument(
"--text_column_name",
type=str,
default=None,
help="The column name of text to input in the file (a csv or JSON file).",
)
parser.add_argument(
"--label_column_name",
type=str,
default=None,
help="The column name of label to input in the file (a csv or JSON file).",
)
parser.add_argument(
"--max_length",
type=int,
default=128,
help=(
"The maximum total input sequence length after tokenization. Sequences longer than this will be truncated,"
" sequences shorter will be padded if `--pad_to_max_length` is passed."
),
)
parser.add_argument(
"--max_entity_length",
type=int,
default=32,
help=(
"The maximum total input entity length after tokenization (Used only for (M)Luke models). Sequences longer"
" than this will be truncated, sequences shorter will be padded if `--pad_to_max_length` is passed."
),
)
parser.add_argument(
"--max_mention_length",
type=int,
default=30,
help=(
"The maximum total input mention length after tokenization (Used only for (M)Luke models). Sequences"
" longer than this will be truncated, sequences shorter will be padded if `--pad_to_max_length` is passed."
),
)
parser.add_argument(
"--pad_to_max_length",
action="store_true",
help="If passed, pad all samples to `max_length`. Otherwise, dynamic padding is used.",
)
parser.add_argument(
"--model_name_or_path",
type=str,
help="Path to pretrained model or model identifier from huggingface.co/models.",
required=True,
)
parser.add_argument(
"--config_name",
type=str,
default=None,
help="Pretrained config name or path if not the same as model_name",
)
parser.add_argument(
"--tokenizer_name",
type=str,
default=None,
help="Pretrained tokenizer name or path if not the same as model_name",
)
parser.add_argument(
"--per_device_train_batch_size",
type=int,
default=8,
help="Batch size (per device) for the training dataloader.",
)
parser.add_argument(
"--per_device_eval_batch_size",
type=int,
default=8,
help="Batch size (per device) for the evaluation dataloader.",
)
parser.add_argument(
"--learning_rate",
type=float,
default=5e-5,
help="Initial learning rate (after the potential warmup period) to use.",
)
parser.add_argument("--weight_decay", type=float, default=0.0, help="Weight decay to use.")
parser.add_argument("--num_train_epochs", type=int, default=3, help="Total number of training epochs to perform.")
parser.add_argument(
"--max_train_steps",
type=int,
default=None,
help="Total number of training steps to perform. If provided, overrides num_train_epochs.",
)
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument(
"--lr_scheduler_type",
type=SchedulerType,
default="linear",
help="The scheduler type to use.",
choices=["linear", "cosine", "cosine_with_restarts", "polynomial", "constant", "constant_with_warmup"],
)
parser.add_argument(
"--num_warmup_steps", type=int, default=0, help="Number of steps for the warmup in the lr scheduler."
)
parser.add_argument("--output_dir", type=str, default=None, help="Where to store the final model.")
parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
parser.add_argument(
"--label_all_tokens",
action="store_true",
help="Setting labels of all special tokens to -100 and thus PyTorch will ignore them.",
)
parser.add_argument(
"--return_entity_level_metrics",
action="store_true",
help="Indication whether entity level metrics are to be returner.",
)
parser.add_argument(
"--task_name",
type=str,
default="ner",
choices=["ner", "pos", "chunk"],
help="The name of the task.",
)
parser.add_argument(
"--debug",
action="store_true",
help="Activate debug mode and run training only with a subset of data.",
)
parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
parser.add_argument(
"--hub_model_id", type=str, help="The name of the repository to keep in sync with the local `output_dir`."
)
parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
args = parser.parse_args()
# Sanity checks
if args.task_name is None and args.train_file is None and args.validation_file is None:
raise ValueError("Need either a task name or a training/validation file.")
else:
if args.train_file is not None:
extension = args.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if args.validation_file is not None:
extension = args.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
if args.push_to_hub:
assert args.output_dir is not None, "Need an `output_dir` to create a repo when `--push_to_hub` is passed."
return args
def main():
args = parse_args()
# Initialize the accelerator. We will let the accelerator handle device placement for us in this example.
handler = DistributedDataParallelKwargs(find_unused_parameters=True)
accelerator = Accelerator(kwargs_handlers=[handler])
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
logger.info(accelerator.state)
# Setup logging, we only want one process per machine to log things on the screen.
# accelerator.is_local_main_process is only True for one process per machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed)
# Handle the repository creation
if accelerator.is_main_process:
if args.push_to_hub:
# Retrieve of infer repo_name
repo_name = args.hub_model_id
if repo_name is None:
repo_name = Path(args.output_dir).absolute().name
# Create repo and retrieve repo_id
repo_id = create_repo(repo_name, exist_ok=True, token=args.hub_token).repo_id
# Clone repo locally
repo = Repository(args.output_dir, clone_from=repo_id, token=args.hub_token)
elif args.output_dir is not None:
os.makedirs(args.output_dir, exist_ok=True)
accelerator.wait_for_everyone()
# 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 for token classification task 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 'tokens' or the first column if no column called
# 'tokens' is found. 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 args.dataset_name is not None:
# Downloading and loading a dataset from the hub.
raw_datasets = load_dataset(args.dataset_name, args.dataset_config_name)
else:
data_files = {}
if args.train_file is not None:
data_files["train"] = args.train_file
if args.validation_file is not None:
data_files["validation"] = args.validation_file
extension = args.train_file.split(".")[-1]
raw_datasets = load_dataset(extension, data_files=data_files)
# Trim a number of training examples
if args.debug:
for split in raw_datasets.keys():
raw_datasets[split] = raw_datasets[split].select(range(100))
# 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.
if raw_datasets["train"] is not None:
column_names = raw_datasets["train"].column_names
features = raw_datasets["train"].features
else:
column_names = raw_datasets["validation"].column_names
features = raw_datasets["validation"].features
if args.text_column_name is not None:
text_column_name = args.text_column_name
elif "tokens" in column_names:
text_column_name = "tokens"
else:
text_column_name = column_names[0]
if args.label_column_name is not None:
label_column_name = args.label_column_name
elif f"{args.task_name}_tags" in column_names:
label_column_name = f"{args.task_name}_tags"
else:
label_column_name = column_names[1]
# In the event the labels are not a `Sequence[ClassLabel]`, we will need to go through the dataset to get the
# unique labels.
def get_label_list(labels):
unique_labels = set()
for label in labels:
unique_labels = unique_labels | set(label)
label_list = list(unique_labels)
label_list.sort()
return label_list
if isinstance(features[label_column_name].feature, ClassLabel):
label_list = features[label_column_name].feature.names
# No need to convert the labels since they are already ints.
else:
label_list = get_label_list(raw_datasets["train"][label_column_name])
num_labels = len(label_list)
# Map that sends B-Xxx label to its I-Xxx counterpart
b_to_i_label = []
for idx, label in enumerate(label_list):
if label.startswith("B-") and label.replace("B-", "I-") in label_list:
b_to_i_label.append(label_list.index(label.replace("B-", "I-")))
else:
b_to_i_label.append(idx)
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
if args.config_name:
config = LukeConfig.from_pretrained(args.config_name, num_labels=num_labels)
elif args.model_name_or_path:
config = LukeConfig.from_pretrained(args.model_name_or_path, num_labels=num_labels)
else:
logger.warning("You are instantiating a new config instance from scratch.")
tokenizer_name_or_path = args.tokenizer_name if args.tokenizer_name else args.model_name_or_path
if not tokenizer_name_or_path:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_name."
)
tokenizer = LukeTokenizer.from_pretrained(
tokenizer_name_or_path,
use_fast=False,
task="entity_span_classification",
max_entity_length=args.max_entity_length,
max_mention_length=args.max_mention_length,
)
if args.model_name_or_path:
model = LukeForEntitySpanClassification.from_pretrained(
args.model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
)
else:
logger.info("Training new model from scratch")
model = LukeForEntitySpanClassification.from_config(config)
model.resize_token_embeddings(len(tokenizer))
# Preprocessing the datasets.
# First we tokenize all the texts.
padding = "max_length" if args.pad_to_max_length else False
def compute_sentence_boundaries_for_luke(examples):
sentence_boundaries = []
for tokens in examples[text_column_name]:
sentence_boundaries.append([0, len(tokens)])
examples["sentence_boundaries"] = sentence_boundaries
return examples
def compute_entity_spans_for_luke(examples):
all_entity_spans = []
texts = []
all_labels_entity_spans = []
all_original_entity_spans = []
for labels, tokens, sentence_boundaries in zip(
examples[label_column_name], examples[text_column_name], examples["sentence_boundaries"]
):
subword_lengths = [len(tokenizer.tokenize(token)) for token in tokens]
total_subword_length = sum(subword_lengths)
_, context_end = sentence_boundaries
if total_subword_length > args.max_length - 2:
cur_length = sum(subword_lengths[:context_end])
idx = context_end - 1
while cur_length > args.max_length - 2:
cur_length -= subword_lengths[idx]
context_end -= 1
idx -= 1
text = ""
sentence_words = tokens[:context_end]
sentence_subword_lengths = subword_lengths[:context_end]
word_start_char_positions = []
word_end_char_positions = []
labels_positions = {}
for word, label in zip(sentence_words, labels):
if word[0] == "'" or (len(word) == 1 and is_punctuation(word)):
text = text.rstrip()
word_start_char_positions.append(len(text))
text += word
word_end_char_positions.append(len(text))
text += " "
labels_positions[(word_start_char_positions[-1], word_end_char_positions[-1])] = label
text = text.rstrip()
texts.append(text)
entity_spans = []
labels_entity_spans = []
original_entity_spans = []
for word_start in range(len(sentence_words)):
for word_end in range(word_start, len(sentence_words)):
if (
sum(sentence_subword_lengths[word_start:word_end]) <= tokenizer.max_mention_length
and len(entity_spans) < tokenizer.max_entity_length
):
entity_spans.append((word_start_char_positions[word_start], word_end_char_positions[word_end]))
original_entity_spans.append((word_start, word_end + 1))
if (
word_start_char_positions[word_start],
word_end_char_positions[word_end],
) in labels_positions:
labels_entity_spans.append(
labels_positions[
(word_start_char_positions[word_start], word_end_char_positions[word_end])
]
)
else:
labels_entity_spans.append(0)
all_entity_spans.append(entity_spans)
all_labels_entity_spans.append(labels_entity_spans)
all_original_entity_spans.append(original_entity_spans)
examples["entity_spans"] = all_entity_spans
examples["text"] = texts
examples["labels_entity_spans"] = all_labels_entity_spans
examples["original_entity_spans"] = all_original_entity_spans
return examples
def tokenize_and_align_labels(examples):
entity_spans = []
for v in examples["entity_spans"]:
entity_spans.append(list(map(tuple, v)))
tokenized_inputs = tokenizer(
examples["text"],
entity_spans=entity_spans,
max_length=args.max_length,
padding=padding,
truncation=True,
)
if padding == "max_length":
tokenized_inputs["labels"] = padding_tensor(
examples["labels_entity_spans"], -100, tokenizer.padding_side, tokenizer.max_entity_length
)
tokenized_inputs["original_entity_spans"] = padding_tensor(
examples["original_entity_spans"], (-1, -1), tokenizer.padding_side, tokenizer.max_entity_length
)
tokenized_inputs[label_column_name] = padding_tensor(
examples[label_column_name], -1, tokenizer.padding_side, tokenizer.max_entity_length
)
else:
tokenized_inputs["labels"] = [ex[: tokenizer.max_entity_length] for ex in examples["labels_entity_spans"]]
tokenized_inputs["original_entity_spans"] = [
ex[: tokenizer.max_entity_length] for ex in examples["original_entity_spans"]
]
tokenized_inputs[label_column_name] = [
ex[: tokenizer.max_entity_length] for ex in examples[label_column_name]
]
return tokenized_inputs
with accelerator.main_process_first():
raw_datasets = raw_datasets.map(
compute_sentence_boundaries_for_luke,
batched=True,
desc="Adding sentence boundaries",
)
raw_datasets = raw_datasets.map(
compute_entity_spans_for_luke,
batched=True,
desc="Adding sentence spans",
)
processed_raw_datasets = raw_datasets.map(
tokenize_and_align_labels,
batched=True,
remove_columns=raw_datasets["train"].column_names,
desc="Running tokenizer on dataset",
)
train_dataset = processed_raw_datasets["train"]
eval_dataset = processed_raw_datasets["validation"]
# Log a few random samples from the training set:
for index in random.sample(range(len(train_dataset)), 3):
logger.info(f"Sample {index} of the training set: {train_dataset[index]}.")
# DataLoaders creation:
if args.pad_to_max_length:
# If padding was already done ot max length, we use the default data collator that will just convert everything
# to tensors.
data_collator = default_data_collator
else:
# Otherwise, `DataCollatorForTokenClassification` will apply dynamic padding for us (by padding to the maximum length of
# the samples passed). When using mixed precision, we add `pad_to_multiple_of=8` to pad all tensors to multiple
# of 8s, which will enable the use of Tensor Cores on NVIDIA hardware with compute capability >= 7.5 (Volta).
data_collator = DataCollatorForLukeTokenClassification(
tokenizer, pad_to_multiple_of=(8 if accelerator.use_fp16 else None)
)
train_dataloader = DataLoader(
train_dataset, shuffle=True, collate_fn=data_collator, batch_size=args.per_device_train_batch_size
)
eval_dataloader = DataLoader(eval_dataset, collate_fn=data_collator, batch_size=args.per_device_eval_batch_size)
# Optimizer
# Split weights in two groups, one with weight decay and the other not.
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{
"params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{
"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)],
"weight_decay": 0.0,
},
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
# Use the device given by the `accelerator` object.
device = accelerator.device
model.to(device)
# Prepare everything with our `accelerator`.
model, optimizer, train_dataloader, eval_dataloader = accelerator.prepare(
model, optimizer, train_dataloader, eval_dataloader
)
# Note -> the training dataloader needs to be prepared before we grab his length below (cause its length will be
# shorter in multiprocess)
# Scheduler and math around the number of training steps.
num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
if args.max_train_steps is None:
args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
else:
args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
lr_scheduler = get_scheduler(
name=args.lr_scheduler_type,
optimizer=optimizer,
num_warmup_steps=args.num_warmup_steps,
num_training_steps=args.max_train_steps,
)
# Metrics
metric = load_metric("seqeval")
def get_luke_labels(outputs, ner_tags, original_entity_spans):
true_predictions = []
true_labels = []
for output, original_spans, tags in zip(outputs.logits, original_entity_spans, ner_tags):
true_tags = [val for val in tags if val != -1]
true_original_spans = [val for val in original_spans if val != (-1, -1)]
max_indices = torch.argmax(output, axis=1)
max_logits = torch.max(output, axis=1).values
predictions = []
for logit, index, span in zip(max_logits, max_indices, true_original_spans):
if index != 0:
predictions.append((logit, span, label_list[index]))
predicted_sequence = [label_list[0]] * len(true_tags)
for _, span, label in sorted(predictions, key=lambda o: o[0], reverse=True):
if all(o == label_list[0] for o in predicted_sequence[span[0] : span[1]]):
predicted_sequence[span[0]] = label
if span[1] - span[0] > 1:
predicted_sequence[span[0] + 1 : span[1]] = [label] * (span[1] - span[0] - 1)
true_predictions.append(predicted_sequence)
true_labels.append([label_list[tag_id] for tag_id in true_tags])
return true_predictions, true_labels
def compute_metrics():
results = metric.compute()
if args.return_entity_level_metrics:
# Unpack nested dictionaries
final_results = {}
for key, value in results.items():
if isinstance(value, dict):
for n, v in value.items():
final_results[f"{key}_{n}"] = v
else:
final_results[key] = value
return final_results
else:
return {
"precision": results["overall_precision"],
"recall": results["overall_recall"],
"f1": results["overall_f1"],
"accuracy": results["overall_accuracy"],
}
# Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {args.num_train_epochs}")
logger.info(f" Instantaneous batch size per device = {args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
logger.info(f" Gradient Accumulation steps = {args.gradient_accumulation_steps}")
logger.info(f" Total optimization steps = {args.max_train_steps}")
# Only show the progress bar once on each machine.
progress_bar = tqdm(range(args.max_train_steps), disable=not accelerator.is_local_main_process)
completed_steps = 0
for epoch in range(args.num_train_epochs):
model.train()
for step, batch in enumerate(train_dataloader):
_ = batch.pop("original_entity_spans")
outputs = model(**batch)
loss = outputs.loss
loss = loss / args.gradient_accumulation_steps
accelerator.backward(loss)
if step % args.gradient_accumulation_steps == 0 or step == len(train_dataloader) - 1:
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
progress_bar.update(1)
completed_steps += 1
if completed_steps >= args.max_train_steps:
break
model.eval()
for step, batch in enumerate(eval_dataloader):
original_entity_spans = batch.pop("original_entity_spans")
with torch.no_grad():
outputs = model(**batch)
preds, refs = get_luke_labels(outputs, batch[label_column_name], original_entity_spans)
metric.add_batch(
predictions=preds,
references=refs,
) # predictions and preferences are expected to be a nested list of labels, not label_ids
eval_metric = compute_metrics()
accelerator.print(f"epoch {epoch}:", eval_metric)
if args.push_to_hub and epoch < args.num_train_epochs - 1:
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
if accelerator.is_main_process:
tokenizer.save_pretrained(args.output_dir)
repo.push_to_hub(
commit_message=f"Training in progress epoch {epoch}", blocking=False, auto_lfs_prune=True
)
if args.output_dir is not None:
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(args.output_dir, save_function=accelerator.save)
if accelerator.is_main_process:
tokenizer.save_pretrained(args.output_dir)
if args.push_to_hub:
repo.push_to_hub(commit_message="End of training", auto_lfs_prune=True)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/luke/luke_utils.py
|
import unicodedata
from dataclasses import dataclass
from typing import Optional, Union
import numpy as np
from transformers.data.data_collator import DataCollatorMixin
from transformers.file_utils import PaddingStrategy
from transformers.tokenization_utils_base import PreTrainedTokenizerBase
def padding_tensor(sequences, padding_value, padding_side, sequence_length):
if isinstance(padding_value, tuple):
out_tensor = np.full((len(sequences), sequence_length, 2), padding_value)
else:
out_tensor = np.full((len(sequences), sequence_length), padding_value)
for i, tensor in enumerate(sequences):
if padding_side == "right":
if isinstance(padding_value, tuple):
out_tensor[i, : len(tensor[:sequence_length]), :2] = tensor[:sequence_length]
else:
out_tensor[i, : len(tensor[:sequence_length])] = tensor[:sequence_length]
else:
if isinstance(padding_value, tuple):
out_tensor[i, len(tensor[:sequence_length]) - 1 :, :2] = tensor[:sequence_length]
else:
out_tensor[i, len(tensor[:sequence_length]) - 1 :] = tensor[:sequence_length]
return out_tensor.tolist()
def is_punctuation(char):
cp = ord(char)
if (cp >= 33 and cp <= 47) or (cp >= 58 and cp <= 64) or (cp >= 91 and cp <= 96) or (cp >= 123 and cp <= 126):
return True
cat = unicodedata.category(char)
if cat.startswith("P"):
return True
return False
@dataclass
class DataCollatorForLukeTokenClassification(DataCollatorMixin):
"""
Data collator that will dynamically pad the inputs received, as well as the labels.
Args:
tokenizer ([`PreTrainedTokenizer`] or [`PreTrainedTokenizerFast`]):
The tokenizer used for encoding the data.
padding (`bool`, `str` or [`~file_utils.PaddingStrategy`], *optional*, defaults to `True`):
Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
among:
- `True` or `'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
sequence if provided).
- `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the
maximum acceptable input length for the model if that argument is not provided.
- `False` or `'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of
different lengths).
max_length (`int`, *optional*):
Maximum length of the returned list and optionally padding length (see above).
pad_to_multiple_of (`int`, *optional*):
If set will pad the sequence to a multiple of the provided value.
This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
7.5 (Volta).
label_pad_token_id (`int`, *optional*, defaults to -100):
The id to use when padding the labels (-100 will be automatically ignore by PyTorch loss functions).
return_tensors (`str`):
The type of Tensor to return. Allowable values are "np", "pt" and "tf".
"""
tokenizer: PreTrainedTokenizerBase
padding: Union[bool, str, PaddingStrategy] = True
max_length: Optional[int] = None
pad_to_multiple_of: Optional[int] = None
label_pad_token_id: int = -100
return_tensors: str = "pt"
def torch_call(self, features):
import torch
label_name = "label" if "label" in features[0].keys() else "labels"
labels = [feature[label_name] for feature in features] if label_name in features[0].keys() else None
batch = self.tokenizer.pad(
features,
padding=self.padding,
max_length=self.max_length,
pad_to_multiple_of=self.pad_to_multiple_of,
# Conversion to tensors will fail if we have labels as they are not of the same length yet.
return_tensors="pt" if labels is None else None,
)
if labels is None:
return batch
sequence_length = torch.tensor(batch["entity_ids"]).shape[1]
padding_side = self.tokenizer.padding_side
if padding_side == "right":
batch[label_name] = [
list(label) + [self.label_pad_token_id] * (sequence_length - len(label)) for label in labels
]
else:
batch[label_name] = [
[self.label_pad_token_id] * (sequence_length - len(label)) + list(label) for label in labels
]
ner_tags = [feature["ner_tags"] for feature in features]
batch["ner_tags"] = padding_tensor(ner_tags, -1, padding_side, sequence_length)
original_entity_spans = [feature["original_entity_spans"] for feature in features]
batch["original_entity_spans"] = padding_tensor(original_entity_spans, (-1, -1), padding_side, sequence_length)
batch = {k: torch.tensor(v, dtype=torch.int64) for k, v in batch.items()}
return batch
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/tapex/README.md
|
<!---
Copyright 2022 The Microsoft Inc. and The HuggingFace Inc. Team. 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.
-->
# Run Table Tasks with TAPEX
TAPEX is a table pre-training approach for table-related tasks. By learning a neural SQL executor over a synthetic corpus based on generative language models (e.g., BART), it achieves state-of-the-art performance on several table-based question answering benchmarks and table-based fact verification benchmark. More details can be found in the original paper [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/pdf/2107.07653.pdf).
> If you are also familiar with [fairseq](https://github.com/pytorch/fairseq), you may also find [the official implementation](https://github.com/microsoft/Table-Pretraining) useful, which leverages the framework.
## Table Question Answering Tasks
### What is Table Question Answering
The task of Table Question Answering (TableQA) is to empower machines to answer users' questions over a given table. The resulting answer(s) can be a region in the table, or a number calculated by applying aggregation operators to a specific region.
### What Questions Can be Answered
Benefiting from the powerfulness of generative models, TAPEX can deal with almost all kinds of questions over tables (if there is training data). Below are some typical question and their answers taken from [WikiTableQuestion](https://nlp.stanford.edu/blog/wikitablequestions-a-complex-real-world-question-understanding-dataset).
| Question | Answer |
| :---: | :---: |
| What is the years won for each team? | 2004, 2008, 2012 |
| How long did Taiki Tsuchiya last? | 4:27 |
| What is the total amount of matches drawn? | 1 |
| Besides Tiger Woods, what other player won between 2007 and 2009? | Camilo Villegas |
| What was the last Baekje Temple? | Uija |
| What is the difference between White voters and Black voters in 1948? | 0 |
| What is the average number of sailors for each country during the worlds qualification tournament? | 2 |
### How to Fine-tune TAPEX on TableQA
We provide a fine-tuning script of tapex for TableQA on the WikiSQL benchmark: [WikiSQL](https://github.com/salesforce/WikiSQL).
This script is customized for tapex models, and can be easily adapted to other benchmarks such as WikiTableQuestion
(only some tweaks in the function `preprocess_tableqa_function`).
#### TAPEX-Base on WikiSQL
Here is how to run the script on the WikiSQL with `tapex-base`:
> The default hyper-parameter may allow you to reproduce our reported tapex-base results within the memory budget of 16GB and 1 GPU card. If you have more GPU cards, you could reduce `gradient_accumulation_steps` accordingly.
```bash
export EXP_NAME=wikisql_tapex_base
python run_wikisql_with_tapex.py \
--do_train \
--do_eval \
--output_dir $EXP_NAME \
--model_name_or_path microsoft/tapex-base \
--overwrite_output_dir \
--per_device_train_batch_size 4 \
--gradient_accumulation_steps 8 \
--per_device_eval_batch_size 4 \
--learning_rate 3e-5 \
--logging_steps 10 \
--eval_steps 1000 \
--save_steps 1000 \
--warmup_steps 1000 \
--evaluation_strategy steps \
--predict_with_generate \
--num_beams 5 \
--weight_decay 1e-2 \
--label_smoothing_factor 0.1 \
--max_steps 20000
```
#### TAPEX-Large on WikiSQL
Here is how to run the script on the WikiSQL with `tapex-large`:
> The default hyper-parameter may allow you to reproduce our reported tapex-large results within the memory budget of 16GB and 1 GPU card with fp16. If you have more GPU cards, you could reduce `gradient_accumulation_steps` accordingly. If you do not install apex or other mixed-precision-training libs, you could disable the `predict_with_generate` option to save GPU memory and manually evaluate the model once the fine-tuning finished. Or just pick up the last checkpoint, which usually performs good enough on the dataset.
```bash
export EXP_NAME=wikisql_tapex_large
python run_wikisql_with_tapex.py \
--do_train \
--do_eval \
--output_dir $EXP_NAME \
--model_name_or_path microsoft/tapex-large \
--overwrite_output_dir \
--per_device_train_batch_size 1 \
--gradient_accumulation_steps 32 \
--per_device_eval_batch_size 4 \
--learning_rate 3e-5 \
--logging_steps 10 \
--eval_steps 1000 \
--save_steps 1000 \
--warmup_steps 1000 \
--evaluation_strategy steps \
--predict_with_generate \
--num_beams 5 \
--weight_decay 1e-2 \
--label_smoothing_factor 0.1 \
--max_steps 20000 \
--fp16
```
#### TAPEX-Base on WikiTableQuestions
Here is how to run the script on the WikiTableQuestions with `tapex-base`:
> The default hyper-parameter may allow you to reproduce our reported tapex-base results within the memory budget of 16GB and 1 GPU card. If you have more GPU cards, you could reduce `gradient_accumulation_steps` accordingly.
```bash
export EXP_NAME=wikitablequestions_tapex_base
python run_wikitablequestions_with_tapex.py \
--do_train \
--do_eval \
--output_dir $EXP_NAME \
--model_name_or_path microsoft/tapex-base \
--overwrite_output_dir \
--per_device_train_batch_size 4 \
--gradient_accumulation_steps 8 \
--per_device_eval_batch_size 4 \
--learning_rate 3e-5 \
--logging_steps 10 \
--eval_steps 1000 \
--save_steps 1000 \
--warmup_steps 1000 \
--evaluation_strategy steps \
--predict_with_generate \
--num_beams 5 \
--weight_decay 1e-2 \
--label_smoothing_factor 0.1 \
--max_steps 20000
```
#### TAPEX-Large on WikiTableQuestions
Here is how to run the script on the WikiTableQuestions with `tapex-large`:
> The default hyper-parameter may allow you to reproduce our reported tapex-large results within the memory budget of 16GB and 1 GPU card with fp16. If you have more GPU cards, you could reduce `gradient_accumulation_steps` accordingly. If you do not install apex or other mixed-precision-training libs, you could reduce the `per_device_train_batch_size` and `per_device_eval_batch_size` and have another try. Or you could disable the `predict_with_generate` option to save GPU memory and manually evaluate the model once the fine-tuning finished. Or just pick up the last checkpoint, which usually performs good enough on the dataset.
```bash
export EXP_NAME=wikitablequestions_tapex_large
python run_wikitablequestions_with_tapex.py \
--do_train \
--do_eval \
--output_dir $EXP_NAME \
--model_name_or_path microsoft/tapex-large \
--overwrite_output_dir \
--per_device_train_batch_size 2 \
--gradient_accumulation_steps 12 \
--per_device_eval_batch_size 4 \
--learning_rate 3e-5 \
--logging_steps 10 \
--eval_steps 1000 \
--save_steps 1000 \
--warmup_steps 1000 \
--evaluation_strategy steps \
--predict_with_generate \
--num_beams 5 \
--weight_decay 1e-2 \
--label_smoothing_factor 0.1 \
--max_steps 20000 \
--fp16
```
### How to Evaluate TAPEX Fine-tuned Models on TableQA
We provide fine-tuned model weights to reproduce our results. You can evaluate them using the following command:
> You can also replace `microsoft/tapex-base-finetuned-wikisql` with your local directory to evaluate your fine-tuned models. Notice that if the model has a larger size, you should reduce `per_device_eval_batch_size` to fit the memory requirement.
```bash
export EXP_NAME=wikisql_tapex_base_eval
python run_wikisql_with_tapex.py \
--do_eval \
--model_name_or_path microsoft/tapex-base-finetuned-wikisql \
--output_dir $EXP_NAME \
--per_device_eval_batch_size 4 \
--predict_with_generate \
--num_beams 5
```
## Table Fact Verification Tasks
### What is Table Fact Verification
The task of Table Fact Verification (TableFV) is to empower machines to justify if a statement follows facts in a given table. The result is a binary classification belonging to `1` (entailed) or `0` (refused).
### How to Fine-tune TAPEX on TableFV
#### TAPEX-Base on TabFact
We provide a fine-tuning script of tapex for TableFV on the TabFact benchmark: [TabFact](https://github.com/wenhuchen/Table-Fact-Checking).
Here is how to run the script on the TabFact:
> The default hyper-parameter may allow you to reproduce our reported tapex-base results within the memory budget of 16GB and 1 GPU card. If you have more GPU cards, you could reduce `gradient_accumulation_steps` accordingly. Note that the `eval_accumulation_steps` is necessary, otherwise GPU memory leaks will occur during the evaluation.
```bash
export EXP_NAME=tabfact_tapex_base
python run_tabfact_with_tapex.py \
--do_train \
--do_eval \
--output_dir $EXP_NAME \
--model_name_or_path microsoft/tapex-base \
--overwrite_output_dir \
--per_device_train_batch_size 3 \
--gradient_accumulation_steps 16 \
--per_device_eval_batch_size 12 \
--eval_accumulation_steps 6 \
--warm_steps 1000 \
--logging_steps 10 \
--learning_rate 3e-5 \
--eval_steps 1000 \
--save_steps 1000 \
--evaluation_strategy steps \
--weight_decay 1e-2 \
--max_steps 30000 \
--max_grad_norm 0.1
```
#### TAPEX-Large on TabFact
Here is how to run the script on the TabFact:
> The default hyper-parameter may allow you to reproduce our reported tapex-base results within the memory budget of 24GB and 1 GPU card. Sorry we cannot reduce the memory consumption since the model input in TabFact usually contains nearly ~1000 tokens. If you have more GPU cards, you could reduce `gradient_accumulation_steps` accordingly. Note that the `eval_accumulation_steps` is necessary, otherwise GPU memory leaks will occur during the evaluation.
```bash
export EXP_NAME=tabfact_tapex_large
python run_tabfact_with_tapex.py \
--do_train \
--do_eval \
--output_dir $EXP_NAME \
--model_name_or_path microsoft/tapex-large \
--overwrite_output_dir \
--per_device_train_batch_size 2 \
--gradient_accumulation_steps 18 \
--per_device_eval_batch_size 4 \
--eval_accumulation_steps 12 \
--warm_steps 1000 \
--logging_steps 10 \
--learning_rate 3e-5 \
--eval_steps 1000 \
--save_steps 1000 \
--evaluation_strategy steps \
--weight_decay 1e-2 \
--max_steps 30000 \
--max_grad_norm 0.1
```
### How to Evaluate TAPEX Fine-tuned Models on TableFV
We provide fine-tuned model weights to reproduce our results. You can evaluate them using the following command:
> You can also replace `microsoft/tapex-base-finetuned-tabfact` with your local directory to evaluate your fine-tuned models. Notice that if the model has a larger size, you should reduce `per_device_eval_batch_size` to fit the memory requirement.
```bash
export EXP_NAME=tabfact_tapex_base_eval
python run_tabfact_with_tapex.py \
--do_eval \
--model_name_or_path microsoft/tapex-base-finetuned-tabfact \
--output_dir $EXP_NAME \
--per_device_eval_batch_size 12 \
--eval_accumulation_steps 6
```
## Reproduced Results
We get the following results on the dev set of the benchmark with the previous commands:
| Task | Model Size | Metric | Result |
|:---:|:---:|:---:|:---:|
| WikiSQL (Weak) | Base | Denotation Accuracy | 88.1 |
| WikiSQL (Weak) | Large | Denotation Accuracy | 89.5 |
| WikiTableQuestion | Base | Denotation Accuracy | 47.1 |
| WikiTableQuestion | Large | Denotation Accuracy | 57.2 |
| TabFact | Base | Accuracy | 78.7 |
| TabFact | Large | Accuracy | 83.6 |
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/tapex/run_wikitablequestions_with_tapex.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2022 The Microsoft and The HuggingFace Inc. team. 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.
"""
Fine-tuning the library models for tapex on table-based question answering tasks.
Adapted from script: https://github.com/huggingface/transformers/blob/master/examples/pytorch/summarization/run_summarization.py
"""
import logging
import os
import sys
from collections import defaultdict
from dataclasses import dataclass, field
from functools import partial
from typing import List, Optional
import nltk # Here to have a nice missing dependency error message early on
import numpy as np
import pandas as pd
from datasets import load_dataset
from filelock import FileLock
import transformers
from transformers import (
AutoConfig,
BartForConditionalGeneration,
DataCollatorForSeq2Seq,
HfArgumentParser,
Seq2SeqTrainer,
Seq2SeqTrainingArguments,
TapexTokenizer,
set_seed,
)
from transformers.file_utils import is_offline_mode
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
logger = logging.getLogger(__name__)
try:
nltk.data.find("tokenizers/punkt")
except (LookupError, OSError):
if is_offline_mode():
raise LookupError(
"Offline mode: run this script without TRANSFORMERS_OFFLINE first to download nltk data files"
)
with FileLock(".lock") as lock:
nltk.download("punkt", quiet=True)
@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. "
"By default we use BART-large tokenizer for TAPEX-large."
)
},
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where 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 `huggingface-cli login` (necessary to use this script "
"with private models)."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
dataset_name: Optional[str] = field(
default="wikitablequestions", 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)."}
)
train_file: Optional[str] = field(
default=None, metadata={"help": "The input training data file (a jsonlines or csv file)."}
)
validation_file: Optional[str] = field(
default=None,
metadata={
"help": (
"An optional input evaluation data file to evaluate the metrics (rouge) on (a jsonlines or csv file)."
)
},
)
test_file: Optional[str] = field(
default=None,
metadata={
"help": "An optional input test data file to evaluate the metrics (rouge) on (a jsonlines or csv file)."
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
max_source_length: Optional[int] = field(
default=1024,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
max_target_length: Optional[int] = field(
default=128,
metadata={
"help": (
"The maximum total sequence length for target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
val_max_target_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total sequence length for validation target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
"This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
"during ``evaluate`` and ``predict``."
)
},
)
pad_to_max_length: bool = field(
default=False,
metadata={
"help": (
"Whether to pad all samples to model maximum sentence length. "
"If False, will pad the samples dynamically when batching to the maximum length in the batch. More "
"efficient on GPU but very bad for TPU."
)
},
)
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."
)
},
)
num_beams: Optional[int] = field(
default=None,
metadata={
"help": (
"Number of beams to use for evaluation. This argument will be passed to ``model.generate``, "
"which is used during ``evaluate`` and ``predict``."
)
},
)
ignore_pad_token_for_loss: bool = field(
default=True,
metadata={
"help": "Whether to ignore the tokens corresponding to padded labels in the loss computation or not."
},
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
if self.val_max_target_length is None:
self.val_max_target_length = self.max_target_length
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, Seq2SeqTrainingArguments))
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()
# 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."
)
# 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)],
)
logger.setLevel(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()
logger.info(f"Training/evaluation parameters {training_args}")
# Set seed before initializing model.
set_seed(training_args.seed)
# Get the datasets: you can either provide your own CSV/JSON 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 JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# 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, cache_dir=model_args.cache_dir)
else:
data_files = {}
if data_args.train_file is not None:
data_files["train"] = data_args.train_file
extension = data_args.train_file.split(".")[-1]
if data_args.validation_file is not None:
data_files["validation"] = data_args.validation_file
extension = data_args.validation_file.split(".")[-1]
if data_args.test_file is not None:
data_files["test"] = data_args.test_file
extension = data_args.test_file.split(".")[-1]
datasets = load_dataset(extension, data_files=data_files, cache_dir=model_args.cache_dir)
# 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.
# 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,
cache_dir=model_args.cache_dir,
revision=model_args.model_revision,
token=True if model_args.use_auth_token else None,
)
# IMPORTANT: the initial BART model's decoding is penalized by no_repeat_ngram_size, and thus
# we should disable it here to avoid problematic generation
config.no_repeat_ngram_size = 0
config.max_length = 1024
config.early_stopping = False
# load tapex tokenizer
tokenizer = TapexTokenizer.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,
revision=model_args.model_revision,
token=True if model_args.use_auth_token else None,
add_prefix_space=True,
)
# load Bart based Tapex model (default tapex-large)
model = BartForConditionalGeneration.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,
token=True if model_args.use_auth_token else None,
)
if model.config.decoder_start_token_id is None:
raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")
# Preprocessing the datasets.
# We need to tokenize inputs and targets.
if training_args.do_train:
column_names = datasets["train"].column_names
elif training_args.do_eval:
column_names = datasets["validation"].column_names
elif training_args.do_predict:
column_names = datasets["test"].column_names
else:
logger.info("There is nothing to do. Please pass `do_train`, `do_eval` and/or `do_predict`.")
return
# Temporarily set max_target_length for training.
max_target_length = data_args.max_target_length
padding = "max_length" if data_args.pad_to_max_length else False
if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
logger.warning(
"label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
)
def preprocess_tableqa_function(examples, is_training=False):
"""
The is_training FLAG is used to identify if we could use the supervision
to truncate the table content if it is required.
"""
questions = [question.lower() for question in examples["question"]]
example_tables = examples["table"]
tables = [
pd.DataFrame.from_records(example_table["rows"], columns=example_table["header"])
for example_table in example_tables
]
# using wikitablequestion's answer set
answers = examples["answers"]
# IMPORTANT: we cannot pass by answers during evaluation, answers passed during training are used to
# truncate large tables in the train set!
if is_training:
model_inputs = tokenizer(
table=tables,
query=questions,
answer=answers,
max_length=data_args.max_source_length,
padding=padding,
truncation=True,
)
else:
model_inputs = tokenizer(
table=tables, query=questions, max_length=data_args.max_source_length, padding=padding, truncation=True
)
labels = tokenizer(
answer=[", ".join(answer) for answer in answers],
max_length=max_target_length,
padding=padding,
truncation=True,
)
# If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore
# padding in the loss.
if padding == "max_length" and data_args.ignore_pad_token_for_loss:
labels["input_ids"] = [
[(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
]
model_inputs["labels"] = labels["input_ids"]
return model_inputs
# in training, we can use the answer as extra information to truncate large tables
preprocess_tableqa_function_training = partial(preprocess_tableqa_function, is_training=True)
if training_args.do_train:
if "train" not in datasets:
raise ValueError("--do_train requires a train dataset")
train_dataset = datasets["train"]
if data_args.max_train_samples is not None:
train_dataset = train_dataset.select(range(data_args.max_train_samples))
train_dataset = train_dataset.map(
preprocess_tableqa_function_training,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
if training_args.do_eval:
max_target_length = data_args.val_max_target_length
if "validation" not in datasets:
raise ValueError("--do_eval requires a validation dataset")
eval_dataset = datasets["validation"]
if data_args.max_eval_samples is not None:
eval_dataset = eval_dataset.select(range(data_args.max_eval_samples))
eval_dataset = eval_dataset.map(
preprocess_tableqa_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
if training_args.do_predict:
max_target_length = data_args.val_max_target_length
if "test" not in datasets:
raise ValueError("--do_predict requires a test dataset")
predict_dataset = datasets["test"]
if data_args.max_predict_samples is not None:
predict_dataset = predict_dataset.select(range(data_args.max_predict_samples))
predict_dataset = predict_dataset.map(
preprocess_tableqa_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
# Data collator
label_pad_token_id = -100 if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id
data_collator = DataCollatorForSeq2Seq(
tokenizer,
model=model,
label_pad_token_id=label_pad_token_id,
pad_to_multiple_of=8 if training_args.fp16 else None,
)
def postprocess_text(preds, labels):
preds = [pred.strip() for pred in preds]
labels = [label.strip() for label in labels]
return preds, labels
def compute_metrics(eval_preds):
preds, labels = eval_preds
if isinstance(preds, tuple):
preds = preds[0]
decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True)
if data_args.ignore_pad_token_for_loss:
# Replace -100 in the labels as we can't decode them.
labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)
# Some simple post-processing
decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels)
delimiter = ", "
# define example evaluation
def evaluate_example(predict_str: str, ground_str: str):
predict_spans = predict_str.split(delimiter)
ground_spans = ground_str.split(delimiter)
predict_values = defaultdict(lambda: 0)
ground_values = defaultdict(lambda: 0)
for span in predict_spans:
try:
predict_values[float(span)] += 1
except ValueError:
predict_values[span.strip()] += 1
for span in ground_spans:
try:
ground_values[float(span)] += 1
except ValueError:
ground_values[span.strip()] += 1
_is_correct = predict_values == ground_values
return _is_correct
def get_denotation_accuracy(predictions: List[str], references: List[str]):
assert len(predictions) == len(references)
correct_num = 0
for predict_str, ground_str in zip(predictions, references):
is_correct = evaluate_example(predict_str.lower(), ground_str.lower())
if is_correct:
correct_num += 1
return correct_num / len(predictions)
accuracy = get_denotation_accuracy(decoded_preds, decoded_labels)
result = {"denotation_accuracy": accuracy}
return result
# Initialize our Trainer
trainer = Seq2SeqTrainer(
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,
tokenizer=tokenizer,
data_collator=data_collator,
compute_metrics=compute_metrics if training_args.predict_with_generate else None,
)
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)
trainer.save_model() # Saves the tokenizer too for easy upload
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.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
metrics = trainer.evaluate(
max_length=data_args.val_max_target_length, num_beams=data_args.num_beams, metric_key_prefix="eval"
)
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)
if training_args.do_predict:
logger.info("*** Predict ***")
predict_results = trainer.predict(
predict_dataset,
metric_key_prefix="predict",
max_length=data_args.val_max_target_length,
num_beams=data_args.num_beams,
)
metrics = predict_results.metrics
max_predict_samples = (
data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset)
)
metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset))
trainer.log_metrics("predict", metrics)
trainer.save_metrics("predict", metrics)
if trainer.is_world_process_zero():
if training_args.predict_with_generate:
predictions = tokenizer.batch_decode(
predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True
)
predictions = [pred.strip() for pred in predictions]
output_prediction_file = os.path.join(training_args.output_dir, "tapex_predictions.txt")
with open(output_prediction_file, "w") as writer:
writer.write("\n".join(predictions))
return results
def _mp_fn(index):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/tapex/wikisql_utils.py
|
# coding=utf-8
# Copyright 2022 The Microsoft, The Google and The HuggingFace Inc. team. 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.
import dataclasses
import enum
import functools
import math
import re
# The following script is adapted from the script of TaPas.
# Original: https://github.com/google-research/tapas/master/wikisql_utils.py
from typing import Any, List, Text
EMPTY_ANSWER = "none"
EMPTY_ANSWER_AGG = "none"
def _split_thousands(delimiter, value):
split = value.split(delimiter)
return len(split) > 1 and any((len(x) == 3 for x in split))
def convert_to_float(value):
"""Converts value to a float using a series of increasingly complex heuristics.
Args:
value: object that needs to be converted. Allowed types include
float/int/strings.
Returns:
A float interpretation of value.
Raises:
ValueError if the float conversion of value fails.
"""
if isinstance(value, float):
return value
if isinstance(value, int):
return float(value)
if not isinstance(value, str):
raise ValueError("Argument value is not a string. Can't parse it as float")
sanitized = value
try:
# Example: 1,000.7
if "." in sanitized and "," in sanitized:
return float(sanitized.replace(",", ""))
# 1,000
if "," in sanitized and _split_thousands(",", sanitized):
return float(sanitized.replace(",", ""))
# 5,5556
if "," in sanitized and sanitized.count(",") == 1 and not _split_thousands(",", sanitized):
return float(sanitized.replace(",", "."))
# 0.0.0.1
if sanitized.count(".") > 1:
return float(sanitized.replace(".", ""))
# 0,0,0,1
if sanitized.count(",") > 1:
return float(sanitized.replace(",", ""))
return float(sanitized)
except ValueError:
# Avoid adding the sanitized value in the error message.
raise ValueError("Unable to convert value to float")
def _normalize_float(answer):
if answer is None:
return None
try:
value = convert_to_float(answer)
if isinstance(value, float) and math.isnan(value):
return None
return value
except ValueError:
return answer.lower()
_TYPE_CONVERTER = {
"text": lambda x: x,
"real": convert_to_float,
}
class _Aggregation(enum.Enum):
"""Aggregations as defined by WikiSQL. Indexes match the data."""
NONE = 0
MAX = 1
MIN = 2
COUNT = 3
SUM = 4
AVERAGE = 5
class _Operator(enum.Enum):
"""The boolean operators used by WikiSQL. Indexes match the data."""
EQUALS = 0
GREATER = 1
LESSER = 2
@dataclasses.dataclass
class _Condition:
"""Represents an SQL where clauses (e.g A = "a" or B > 5)."""
column: Text
operator: _Operator
cmp_value: Any
_TOKENIZER = re.compile(r"\w+|[^\w\s]+", re.UNICODE | re.MULTILINE | re.DOTALL)
def _normalize_for_match(x):
return list(_TOKENIZER.findall(x.lower()))
def _compare(operator, src, tgt):
if operator == _Operator.EQUALS:
return src == tgt
elif operator == _Operator.GREATER:
return src > tgt
elif operator == _Operator.LESSER:
return src < tgt
raise ValueError(f"Unknown operator: {operator}")
def _parse_value(table, column, cell_value):
"""Convert numeric values to floats and keeps everything else as string."""
types = table["types"]
return _TYPE_CONVERTER[types[column]](cell_value)
def _is_string(x):
return isinstance(x, str)
def _respect_conditions(table, row, conditions):
"""True if 'row' satisfies all 'conditions'."""
for cond in conditions:
table_value = row[cond.column]
cmp_value = _parse_value(table, cond.column, cond.cmp_value)
if _is_string(table_value) and _is_string(cmp_value):
table_value = _normalize_for_match(table_value)
cmp_value = _normalize_for_match(cmp_value)
if not isinstance(table_value, type(cmp_value)):
raise ValueError("Type difference {} != {}".format(type(table_value), type(cmp_value)))
if not _compare(cond.operator, table_value, cmp_value):
return False
return True
def _get_float_answer(table, answer_coordinates, aggregation_op):
"""Applies operation to produce reference float answer."""
if not answer_coordinates:
if aggregation_op == _Aggregation.COUNT:
return 0.0
else:
return EMPTY_ANSWER_AGG
# Count can support non numeric answers.
if aggregation_op == _Aggregation.COUNT:
return float(len(answer_coordinates))
# If we have just one answer, if float returns it or try a conversion.
values = [table["rows"][i][j] for (i, j) in answer_coordinates]
if len(answer_coordinates) == 1:
try:
return convert_to_float(values[0])
except ValueError as e:
if aggregation_op != _Aggregation.NONE:
raise e
if aggregation_op == _Aggregation.NONE:
return None
# Other aggregation only support numeric values. Bail out if we have strings.
if not all((isinstance(v, (int, float)) for v in values)):
return None
if aggregation_op == _Aggregation.SUM:
return float(sum(values))
elif aggregation_op == _Aggregation.AVERAGE:
return sum(values) / len(answer_coordinates)
else:
raise ValueError(f"Unknown aggregation: {aggregation_op}")
def _get_answer_coordinates(table, sql_query):
"""Retrieves references coordinates by executing SQL."""
# MAX and MIN are automatically supported by the model.
aggregation_op_index = sql_query["agg"]
if aggregation_op_index >= 3:
aggregation_op = _Aggregation(aggregation_op_index)
else:
aggregation_op = _Aggregation.NONE
target_column = sql_query["sel"]
conditions = [
_Condition(column, _Operator(operator), cmp_value)
for column, operator, cmp_value in zip(
sql_query["conds"]["column_index"], sql_query["conds"]["operator_index"], sql_query["conds"]["condition"]
)
]
indices = []
for row in range(len(table["rows"])):
if _respect_conditions(table, table["rows"][row], conditions):
indices.append((row, target_column))
if not indices:
return [], aggregation_op
if len(indices) == 1:
return indices, aggregation_op
# Parsing of MIN/MAX.
if aggregation_op_index in (1, 2):
operators = {2: min, 1: max}
values = [(table["rows"][i][j], index) for index, (i, j) in enumerate(indices)]
reduced = functools.reduce(operators[sql_query["agg"]], values)
ret = [indices[reduced[1]]]
return ret, _Aggregation.NONE
return indices, aggregation_op
def _get_answer_text(table, answer_coordinates, float_answer):
if float_answer is not None:
return [str(float_answer)]
return [str(table["real_rows"][r][c]) for r, c in answer_coordinates]
def retrieve_wikisql_query_answer_tapas(table, example) -> List:
answer_coordinates, aggregation_op = _get_answer_coordinates(table, example)
float_answer = _get_float_answer(table, answer_coordinates, aggregation_op)
answer_text = _get_answer_text(table, answer_coordinates, float_answer)
# keep the original data the same with TaPas
if len(answer_text) == 0:
answer_text = [EMPTY_ANSWER]
return answer_text
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/tapex/run_wikisql_with_tapex.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2022 The Microsoft and The HuggingFace Inc. team. 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.
"""
Fine-tuning the library models for tapex on table-based question answering tasks.
Adapted from script: https://github.com/huggingface/transformers/blob/master/examples/pytorch/summarization/run_summarization.py
"""
import logging
import os
import sys
from collections import defaultdict
from copy import deepcopy
from dataclasses import dataclass, field
from functools import partial
from typing import List, Optional
import nltk # Here to have a nice missing dependency error message early on
import numpy as np
import pandas as pd
from datasets import load_dataset
from filelock import FileLock
from wikisql_utils import _TYPE_CONVERTER, retrieve_wikisql_query_answer_tapas
import transformers
from transformers import (
AutoConfig,
BartForConditionalGeneration,
DataCollatorForSeq2Seq,
HfArgumentParser,
Seq2SeqTrainer,
Seq2SeqTrainingArguments,
TapexTokenizer,
set_seed,
)
from transformers.file_utils import is_offline_mode
from transformers.trainer_utils import get_last_checkpoint, is_main_process
from transformers.utils import check_min_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
logger = logging.getLogger(__name__)
try:
nltk.data.find("tokenizers/punkt")
except (LookupError, OSError):
if is_offline_mode():
raise LookupError(
"Offline mode: run this script without TRANSFORMERS_OFFLINE first to download nltk data files"
)
with FileLock(".lock") as lock:
nltk.download("punkt", quiet=True)
@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. "
"By default we use BART-large tokenizer for TAPEX-large."
)
},
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where 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 `huggingface-cli login` (necessary to use this script "
"with private models)."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
dataset_name: Optional[str] = field(
default="wikisql", 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)."}
)
train_file: Optional[str] = field(
default=None, metadata={"help": "The input training data file (a jsonlines or csv file)."}
)
validation_file: Optional[str] = field(
default=None,
metadata={
"help": (
"An optional input evaluation data file to evaluate the metrics (rouge) on (a jsonlines or csv file)."
)
},
)
test_file: Optional[str] = field(
default=None,
metadata={
"help": "An optional input test data file to evaluate the metrics (rouge) on (a jsonlines or csv file)."
},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
max_source_length: Optional[int] = field(
default=1024,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
max_target_length: Optional[int] = field(
default=128,
metadata={
"help": (
"The maximum total sequence length for target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
val_max_target_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total sequence length for validation target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`. "
"This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
"during ``evaluate`` and ``predict``."
)
},
)
pad_to_max_length: bool = field(
default=False,
metadata={
"help": (
"Whether to pad all samples to model maximum sentence length. "
"If False, will pad the samples dynamically when batching to the maximum length in the batch. More "
"efficient on GPU but very bad for TPU."
)
},
)
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."
)
},
)
num_beams: Optional[int] = field(
default=None,
metadata={
"help": (
"Number of beams to use for evaluation. This argument will be passed to ``model.generate``, "
"which is used during ``evaluate`` and ``predict``."
)
},
)
ignore_pad_token_for_loss: bool = field(
default=True,
metadata={
"help": "Whether to ignore the tokens corresponding to padded labels in the loss computation or not."
},
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
if self.val_max_target_length is None:
self.val_max_target_length = self.max_target_length
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, Seq2SeqTrainingArguments))
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()
# 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."
)
# 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)],
)
logger.setLevel(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()
logger.info(f"Training/evaluation parameters {training_args}")
# Set seed before initializing model.
set_seed(training_args.seed)
# Get the datasets: you can either provide your own CSV/JSON 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 JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# 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, cache_dir=model_args.cache_dir)
else:
data_files = {}
if data_args.train_file is not None:
data_files["train"] = data_args.train_file
extension = data_args.train_file.split(".")[-1]
if data_args.validation_file is not None:
data_files["validation"] = data_args.validation_file
extension = data_args.validation_file.split(".")[-1]
if data_args.test_file is not None:
data_files["test"] = data_args.test_file
extension = data_args.test_file.split(".")[-1]
datasets = load_dataset(extension, data_files=data_files, cache_dir=model_args.cache_dir)
# 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.
# 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,
cache_dir=model_args.cache_dir,
revision=model_args.model_revision,
token=True if model_args.use_auth_token else None,
)
# IMPORTANT: the initial BART model's decoding is penalized by no_repeat_ngram_size, and thus
# we should disable it here to avoid problematic generation
config.no_repeat_ngram_size = 0
config.max_length = 1024
config.early_stopping = False
# load tapex tokenizer
tokenizer = TapexTokenizer.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,
revision=model_args.model_revision,
token=True if model_args.use_auth_token else None,
add_prefix_space=True,
)
# load Bart based Tapex model (default tapex-large)
model = BartForConditionalGeneration.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,
token=True if model_args.use_auth_token else None,
)
if model.config.decoder_start_token_id is None:
raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined")
# Preprocessing the datasets.
# We need to tokenize inputs and targets.
if training_args.do_train:
column_names = datasets["train"].column_names
elif training_args.do_eval:
column_names = datasets["validation"].column_names
elif training_args.do_predict:
column_names = datasets["test"].column_names
else:
logger.info("There is nothing to do. Please pass `do_train`, `do_eval` and/or `do_predict`.")
return
# Temporarily set max_target_length for training.
max_target_length = data_args.max_target_length
padding = "max_length" if data_args.pad_to_max_length else False
if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
logger.warning(
"label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for "
f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
)
def preprocess_tableqa_function(examples, is_training=False):
"""
The is_training FLAG is used to identify if we could use the supervision
to truncate the table content if it is required.
"""
# this function is specific for WikiSQL since the util function need the data structure
# to retrieve the WikiSQL answer for each question
def _convert_table_types(_table):
"""Runs the type converter over the table cells."""
ret_table = deepcopy(_table)
types = ret_table["types"]
ret_table["real_rows"] = ret_table["rows"]
typed_rows = []
for row in ret_table["rows"]:
typed_row = []
for column, cell_value in enumerate(row):
typed_row.append(_TYPE_CONVERTER[types[column]](cell_value))
typed_rows.append(typed_row)
ret_table["rows"] = typed_rows
return ret_table
questions = [question.lower() for question in examples["question"]]
example_tables = examples["table"]
example_sqls = examples["sql"]
tables = [
pd.DataFrame.from_records(example_table["rows"], columns=example_table["header"])
for example_table in example_tables
]
# using tapas utils to obtain wikisql answer
answers = []
for example_sql, example_table in zip(example_sqls, example_tables):
tapas_table = _convert_table_types(example_table)
answer_list: List[str] = retrieve_wikisql_query_answer_tapas(tapas_table, example_sql)
# you can choose other delimiters to split each answer
answers.append(answer_list)
# IMPORTANT: we cannot pass by answers during evaluation, answers passed during training are used to
# truncate large tables in the train set!
if is_training:
model_inputs = tokenizer(
table=tables,
query=questions,
answer=answers,
max_length=data_args.max_source_length,
padding=padding,
truncation=True,
)
else:
model_inputs = tokenizer(
table=tables, query=questions, max_length=data_args.max_source_length, padding=padding, truncation=True
)
labels = tokenizer(
answer=[", ".join(answer) for answer in answers],
max_length=max_target_length,
padding=padding,
truncation=True,
)
# If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore
# padding in the loss.
if padding == "max_length" and data_args.ignore_pad_token_for_loss:
labels["input_ids"] = [
[(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
]
model_inputs["labels"] = labels["input_ids"]
return model_inputs
# in training, we can use the answer as extra information to truncate large tables
preprocess_tableqa_function_training = partial(preprocess_tableqa_function, is_training=True)
if training_args.do_train:
if "train" not in datasets:
raise ValueError("--do_train requires a train dataset")
train_dataset = datasets["train"]
if data_args.max_train_samples is not None:
train_dataset = train_dataset.select(range(data_args.max_train_samples))
train_dataset = train_dataset.map(
preprocess_tableqa_function_training,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
if training_args.do_eval:
max_target_length = data_args.val_max_target_length
if "validation" not in datasets:
raise ValueError("--do_eval requires a validation dataset")
eval_dataset = datasets["validation"]
if data_args.max_eval_samples is not None:
eval_dataset = eval_dataset.select(range(data_args.max_eval_samples))
eval_dataset = eval_dataset.map(
preprocess_tableqa_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
if training_args.do_predict:
max_target_length = data_args.val_max_target_length
if "test" not in datasets:
raise ValueError("--do_predict requires a test dataset")
predict_dataset = datasets["test"]
if data_args.max_predict_samples is not None:
predict_dataset = predict_dataset.select(range(data_args.max_predict_samples))
predict_dataset = predict_dataset.map(
preprocess_tableqa_function,
batched=True,
num_proc=data_args.preprocessing_num_workers,
remove_columns=column_names,
load_from_cache_file=not data_args.overwrite_cache,
)
# Data collator
label_pad_token_id = -100 if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id
data_collator = DataCollatorForSeq2Seq(
tokenizer,
model=model,
label_pad_token_id=label_pad_token_id,
pad_to_multiple_of=8 if training_args.fp16 else None,
)
def postprocess_text(preds, labels):
preds = [pred.strip() for pred in preds]
labels = [label.strip() for label in labels]
return preds, labels
def compute_metrics(eval_preds):
preds, labels = eval_preds
if isinstance(preds, tuple):
preds = preds[0]
decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True)
if data_args.ignore_pad_token_for_loss:
# Replace -100 in the labels as we can't decode them.
labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)
# Some simple post-processing
decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels)
delimiter = ", "
# define example evaluation
def evaluate_example(predict_str: str, ground_str: str):
predict_spans = predict_str.split(delimiter)
ground_spans = ground_str.split(delimiter)
predict_values = defaultdict(lambda: 0)
ground_values = defaultdict(lambda: 0)
for span in predict_spans:
try:
predict_values[float(span)] += 1
except ValueError:
predict_values[span.strip()] += 1
for span in ground_spans:
try:
ground_values[float(span)] += 1
except ValueError:
ground_values[span.strip()] += 1
is_correct = predict_values == ground_values
return is_correct
def get_denotation_accuracy(predictions: List[str], references: List[str]):
assert len(predictions) == len(references)
correct_num = 0
for predict_str, ground_str in zip(predictions, references):
is_correct = evaluate_example(predict_str.lower(), ground_str.lower())
if is_correct:
correct_num += 1
return correct_num / len(predictions)
accuracy = get_denotation_accuracy(decoded_preds, decoded_labels)
result = {"denotation_accuracy": accuracy}
return result
# Initialize our Trainer
trainer = Seq2SeqTrainer(
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,
tokenizer=tokenizer,
data_collator=data_collator,
compute_metrics=compute_metrics if training_args.predict_with_generate else None,
)
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)
trainer.save_model() # Saves the tokenizer too for easy upload
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.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()
# Evaluation
results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
metrics = trainer.evaluate(
max_length=data_args.val_max_target_length, num_beams=data_args.num_beams, metric_key_prefix="eval"
)
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)
if training_args.do_predict:
logger.info("*** Predict ***")
predict_results = trainer.predict(
predict_dataset,
metric_key_prefix="predict",
max_length=data_args.val_max_target_length,
num_beams=data_args.num_beams,
)
metrics = predict_results.metrics
max_predict_samples = (
data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset)
)
metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset))
trainer.log_metrics("predict", metrics)
trainer.save_metrics("predict", metrics)
if trainer.is_world_process_zero():
if training_args.predict_with_generate:
predictions = tokenizer.batch_decode(
predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True
)
predictions = [pred.strip() for pred in predictions]
output_prediction_file = os.path.join(training_args.output_dir, "tapex_predictions.txt")
with open(output_prediction_file, "w") as writer:
writer.write("\n".join(predictions))
return results
def _mp_fn(index):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/tapex/run_tabfact_with_tapex.py
|
#!/usr/bin/env python
# coding=utf-8
# Copyright 2022 The Microsoft and The HuggingFace Inc. team. 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.
"""
Fine-tuning the library models for tapex on table-based fact verification tasks.
Adapted from script: https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py
"""
import logging
import os
import random
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import numpy as np
import pandas as pd
from datasets import load_dataset
import transformers
from transformers import (
AutoConfig,
BartForSequenceClassification,
DataCollatorWithPadding,
EvalPrediction,
HfArgumentParser,
TapexTokenizer,
Trainer,
TrainingArguments,
default_data_collator,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint
from transformers.utils import check_min_version
from transformers.utils.versions import require_version
# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
check_min_version("4.17.0.dev0")
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
logger = logging.getLogger(__name__)
@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.
"""
dataset_name: Optional[str] = field(
default="tab_fact", metadata={"help": "The name of the dataset to use (via the datasets library)."}
)
dataset_config_name: Optional[str] = field(
default="tab_fact",
metadata={"help": "The configuration name of the dataset to use (via the datasets library)."},
)
max_seq_length: int = field(
default=1024,
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=False,
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.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"], "`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`."
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
model_name_or_path: str = field(
default=None, 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 `huggingface-cli login` (necessary to use this script "
"with private models)."
)
},
)
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()
# 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)
# Get the datasets: you can either provide your own CSV/JSON training and evaluation files (see below)
# or specify a GLUE benchmark task (the dataset will be downloaded automatically from the datasets Hub).
#
# For JSON files, this script will use the `question` column for the input question and `table` column for the corresponding table.
#
# If the CSVs/JSONs contain only one non-label column, the script does single sentence classification on this
# single 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.
raw_datasets = load_dataset(
data_args.dataset_name, data_args.dataset_config_name, cache_dir=model_args.cache_dir
)
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}
# Get the test dataset: you can provide your own CSV/JSON test file (see below)
# when you use `do_predict` without specifying a GLUE benchmark task.
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)
else:
# Loading a dataset from local json files
raw_datasets = load_dataset("json", data_files=data_files, cache_dir=model_args.cache_dir)
# See more about loading any type of standard or custom dataset at
# https://huggingface.co/docs/datasets/loading_datasets.
# Labels
label_list = raw_datasets["train"].features["label"].names
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,
token=True if model_args.use_auth_token else None,
)
# load tapex tokenizer
tokenizer = TapexTokenizer.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,
revision=model_args.model_revision,
token=True if model_args.use_auth_token else None,
add_prefix_space=True,
)
model = BartForSequenceClassification.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,
token=True if model_args.use_auth_token else None,
)
# 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
# Some models have set the order of the labels to use, so let's make sure we do use it.
model.config.label2id = {"Refused": 0, "Entailed": 1}
model.config.id2label = {0: "Refused", 1: "Entailed"}
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_tabfact_function(examples):
# Tokenize the texts
def _convert_table_text_to_pandas(_table_text):
"""Runs the structured pandas table object for _table_text.
An example _table_text can be: round#clubs remaining\nfirst round#156\n
"""
_table_content = [_table_row.split("#") for _table_row in _table_text.strip("\n").split("\n")]
_table_pd = pd.DataFrame.from_records(_table_content[1:], columns=_table_content[0])
return _table_pd
questions = examples["statement"]
tables = list(map(_convert_table_text_to_pandas, examples["table_text"]))
result = tokenizer(tables, questions, padding=padding, max_length=max_seq_length, truncation=True)
result["label"] = examples["label"]
return result
with training_args.main_process_first(desc="dataset map pre-processing"):
raw_datasets = raw_datasets.map(
preprocess_tabfact_function,
batched=True,
load_from_cache_file=not data_args.overwrite_cache,
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"]
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.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"]
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]}.")
# 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.argmax(preds, axis=1)
return {"accuracy": (preds == p.label_ids).astype(np.float32).mean().item()}
# 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
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,
)
# 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()
# Evaluation
if training_args.do_eval:
logger.info("*** Evaluate ***")
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)
if training_args.do_predict:
logger.info("*** Predict ***")
# Removing the `label` columns because it contains -1 and Trainer won't like that.
predict_dataset = predict_dataset.remove_columns("label")
predictions = trainer.predict(predict_dataset, metric_key_prefix="predict").predictions
predictions = np.argmax(predictions, axis=1)
output_predict_file = os.path.join(training_args.output_dir, "predict_results_tabfact.txt")
if trainer.is_world_process_zero():
with open(output_predict_file, "w") as writer:
logger.info("***** Predict Results *****")
writer.write("index\tprediction\n")
for index, item in enumerate(predictions):
item = label_list[item]
writer.write(f"{index}\t{item}\n")
kwargs = {"finetuned_from": model_args.model_name_or_path, "tasks": "text-classification"}
if training_args.push_to_hub:
trainer.push_to_hub(**kwargs)
else:
trainer.create_model_card(**kwargs)
def _mp_fn(index):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/tapex/requirements.txt
|
numpy
datasets
pandas
nltk
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/bertology/run_bertology.py
|
#!/usr/bin/env python3
# Copyright 2018 CMU and The HuggingFace Inc. team.
#
# 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.
""" Bertology: this script shows how you can explore the internals of the models in the library to:
- compute the entropy of the head attentions
- compute the importance of each head
- prune (remove) the low importance head.
Some parts of this script are adapted from the code of Michel et al. (http://arxiv.org/abs/1905.10650)
which is available at https://github.com/pmichel31415/are-16-heads-really-better-than-1
"""
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, SequentialSampler, Subset
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm
import transformers
from transformers import (
AutoConfig,
AutoModelForSequenceClassification,
AutoTokenizer,
GlueDataset,
default_data_collator,
glue_compute_metrics,
glue_output_modes,
glue_processors,
set_seed,
)
from transformers.trainer_utils import is_main_process
logger = logging.getLogger(__name__)
def entropy(p):
"""Compute the entropy of a probability distribution"""
plogp = p * torch.log(p)
plogp[p == 0] = 0
return -plogp.sum(dim=-1)
def print_2d_tensor(tensor):
"""Print a 2D tensor"""
logger.info("lv, h >\t" + "\t".join(f"{x + 1}" for x in range(len(tensor))))
for row in range(len(tensor)):
if tensor.dtype != torch.long:
logger.info(f"layer {row + 1}:\t" + "\t".join(f"{x:.5f}" for x in tensor[row].cpu().data))
else:
logger.info(f"layer {row + 1}:\t" + "\t".join(f"{x:d}" for x in tensor[row].cpu().data))
def compute_heads_importance(
args, model, eval_dataloader, compute_entropy=True, compute_importance=True, head_mask=None, actually_pruned=False
):
"""This method shows how to compute:
- head attention entropy
- head importance scores according to http://arxiv.org/abs/1905.10650
"""
# Prepare our tensors
n_layers, n_heads = model.config.num_hidden_layers, model.config.num_attention_heads
head_importance = torch.zeros(n_layers, n_heads).to(args.device)
attn_entropy = torch.zeros(n_layers, n_heads).to(args.device)
if head_mask is None:
head_mask = torch.ones(n_layers, n_heads).to(args.device)
head_mask.requires_grad_(requires_grad=True)
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
head_mask = None
preds = None
labels = None
tot_tokens = 0.0
for step, inputs in enumerate(tqdm(eval_dataloader, desc="Iteration", disable=args.local_rank not in [-1, 0])):
for k, v in inputs.items():
inputs[k] = v.to(args.device)
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
outputs = model(**inputs, head_mask=head_mask)
loss, logits, all_attentions = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
if compute_entropy:
for layer, attn in enumerate(all_attentions):
masked_entropy = entropy(attn.detach()) * inputs["attention_mask"].float().unsqueeze(1)
attn_entropy[layer] += masked_entropy.sum(-1).sum(0).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
# Also store our logits/labels if we want to compute metrics afterwards
if preds is None:
preds = logits.detach().cpu().numpy()
labels = inputs["labels"].detach().cpu().numpy()
else:
preds = np.append(preds, logits.detach().cpu().numpy(), axis=0)
labels = np.append(labels, inputs["labels"].detach().cpu().numpy(), axis=0)
tot_tokens += inputs["attention_mask"].float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
exponent = 2
norm_by_layer = torch.pow(torch.pow(head_importance, exponent).sum(-1), 1 / exponent)
head_importance /= norm_by_layer.unsqueeze(-1) + 1e-20
if not args.dont_normalize_global_importance:
head_importance = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print/save matrices
np.save(os.path.join(args.output_dir, "attn_entropy.npy"), attn_entropy.detach().cpu().numpy())
np.save(os.path.join(args.output_dir, "head_importance.npy"), head_importance.detach().cpu().numpy())
logger.info("Attention entropies")
print_2d_tensor(attn_entropy)
logger.info("Head importance scores")
print_2d_tensor(head_importance)
logger.info("Head ranked by importance scores")
head_ranks = torch.zeros(head_importance.numel(), dtype=torch.long, device=args.device)
head_ranks[head_importance.view(-1).sort(descending=True)[1]] = torch.arange(
head_importance.numel(), device=args.device
)
head_ranks = head_ranks.view_as(head_importance)
print_2d_tensor(head_ranks)
return attn_entropy, head_importance, preds, labels
def mask_heads(args, model, eval_dataloader):
"""This method shows how to mask head (set some heads to zero), to test the effect on the network,
based on the head importance scores, as described in Michel et al. (http://arxiv.org/abs/1905.10650)
"""
_, head_importance, preds, labels = compute_heads_importance(args, model, eval_dataloader, compute_entropy=False)
preds = np.argmax(preds, axis=1) if args.output_mode == "classification" else np.squeeze(preds)
original_score = glue_compute_metrics(args.task_name, preds, labels)[args.metric_name]
logger.info("Pruning: original score: %f, threshold: %f", original_score, original_score * args.masking_threshold)
new_head_mask = torch.ones_like(head_importance)
num_to_mask = max(1, int(new_head_mask.numel() * args.masking_amount))
current_score = original_score
while current_score >= original_score * args.masking_threshold:
head_mask = new_head_mask.clone() # save current head mask
# heads from least important to most - keep only not-masked heads
head_importance[head_mask == 0.0] = float("Inf")
current_heads_to_mask = head_importance.view(-1).sort()[1]
if len(current_heads_to_mask) <= num_to_mask:
break
# mask heads
current_heads_to_mask = current_heads_to_mask[:num_to_mask]
logger.info("Heads to mask: %s", str(current_heads_to_mask.tolist()))
new_head_mask = new_head_mask.view(-1)
new_head_mask[current_heads_to_mask] = 0.0
new_head_mask = new_head_mask.view_as(head_mask)
new_head_mask = new_head_mask.clone().detach()
print_2d_tensor(new_head_mask)
# Compute metric and head importance again
_, head_importance, preds, labels = compute_heads_importance(
args, model, eval_dataloader, compute_entropy=False, head_mask=new_head_mask
)
preds = np.argmax(preds, axis=1) if args.output_mode == "classification" else np.squeeze(preds)
current_score = glue_compute_metrics(args.task_name, preds, labels)[args.metric_name]
logger.info(
"Masking: current score: %f, remaining heads %d (%.1f percents)",
current_score,
new_head_mask.sum(),
new_head_mask.sum() / new_head_mask.numel() * 100,
)
logger.info("Final head mask")
print_2d_tensor(head_mask)
np.save(os.path.join(args.output_dir, "head_mask.npy"), head_mask.detach().cpu().numpy())
return head_mask
def prune_heads(args, model, eval_dataloader, head_mask):
"""This method shows how to prune head (remove heads weights) based on
the head importance scores as described in Michel et al. (http://arxiv.org/abs/1905.10650)
"""
# Try pruning and test time speedup
# Pruning is like masking but we actually remove the masked weights
before_time = datetime.now()
_, _, preds, labels = compute_heads_importance(
args, model, eval_dataloader, compute_entropy=False, compute_importance=False, head_mask=head_mask
)
preds = np.argmax(preds, axis=1) if args.output_mode == "classification" else np.squeeze(preds)
score_masking = glue_compute_metrics(args.task_name, preds, labels)[args.metric_name]
original_time = datetime.now() - before_time
original_num_params = sum(p.numel() for p in model.parameters())
heads_to_prune = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(head_mask))
}
assert sum(len(h) for h in heads_to_prune.values()) == (1 - head_mask.long()).sum().item()
model.prune_heads(heads_to_prune)
pruned_num_params = sum(p.numel() for p in model.parameters())
before_time = datetime.now()
_, _, preds, labels = compute_heads_importance(
args,
model,
eval_dataloader,
compute_entropy=False,
compute_importance=False,
head_mask=None,
actually_pruned=True,
)
preds = np.argmax(preds, axis=1) if args.output_mode == "classification" else np.squeeze(preds)
score_pruning = glue_compute_metrics(args.task_name, preds, labels)[args.metric_name]
new_time = datetime.now() - before_time
logger.info(
"Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)",
original_num_params,
pruned_num_params,
pruned_num_params / original_num_params * 100,
)
logger.info("Pruning: score with masking: %f score with pruning: %f", score_masking, score_pruning)
logger.info("Pruning: speed ratio (new timing / original timing): %f percents", original_time / new_time * 100)
def main():
parser = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--data_dir",
default=None,
type=str,
required=True,
help="The input data dir. Should contain the .tsv files (or other data files) for the task.",
)
parser.add_argument(
"--model_name_or_path",
default=None,
type=str,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models",
)
parser.add_argument(
"--task_name",
default=None,
type=str,
required=True,
help="The name of the task to train selected in the list: " + ", ".join(glue_processors.keys()),
)
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help="The output directory where the model predictions and checkpoints will be written.",
)
# Other parameters
parser.add_argument(
"--config_name",
default="",
type=str,
help="Pretrained config name or path if not the same as model_name_or_path",
)
parser.add_argument(
"--tokenizer_name",
default="",
type=str,
help="Pretrained tokenizer name or path if not the same as model_name_or_path",
)
parser.add_argument(
"--cache_dir",
default=None,
type=str,
help="Where do you want to store the pre-trained models downloaded from huggingface.co",
)
parser.add_argument(
"--data_subset", type=int, default=-1, help="If > 0: limit the data to a subset of data_subset instances."
)
parser.add_argument(
"--overwrite_output_dir", action="store_true", help="Whether to overwrite data in output directory"
)
parser.add_argument(
"--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
)
parser.add_argument(
"--dont_normalize_importance_by_layer", action="store_true", help="Don't normalize importance score by layers"
)
parser.add_argument(
"--dont_normalize_global_importance",
action="store_true",
help="Don't normalize all importance scores between 0 and 1",
)
parser.add_argument(
"--try_masking", action="store_true", help="Whether to try to mask head until a threshold of accuracy."
)
parser.add_argument(
"--masking_threshold",
default=0.9,
type=float,
help="masking threshold in term of metrics (stop masking when metric < threshold * original metric value).",
)
parser.add_argument(
"--masking_amount", default=0.1, type=float, help="Amount to heads to masking at each masking step."
)
parser.add_argument("--metric_name", default="acc", type=str, help="Metric to use for head masking.")
parser.add_argument(
"--max_seq_length",
default=128,
type=int,
help=(
"The maximum total input sequence length after WordPiece tokenization. \n"
"Sequences longer than this will be truncated, sequences shorter padded."
),
)
parser.add_argument("--batch_size", default=1, type=int, help="Batch size.")
parser.add_argument("--seed", type=int, default=42)
parser.add_argument("--local_rank", type=int, default=-1, help="local_rank for distributed training on gpus")
parser.add_argument("--no_cuda", action="store_true", help="Whether not to use CUDA when available")
parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
args = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
args.device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank)
args.device = torch.device("cuda", args.local_rank)
args.n_gpu = 1
torch.distributed.init_process_group(backend="nccl") # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN)
logger.info("device: {} n_gpu: {}, distributed: {}".format(args.device, args.n_gpu, bool(args.local_rank != -1)))
# Set the verbosity to info of the Transformers logger (on main process only):
if is_main_process(args.local_rank):
transformers.utils.logging.set_verbosity_info()
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Set seeds
set_seed(args.seed)
# Prepare GLUE task
args.task_name = args.task_name.lower()
if args.task_name not in glue_processors:
raise ValueError("Task not found: %s" % (args.task_name))
processor = glue_processors[args.task_name]()
args.output_mode = glue_output_modes[args.task_name]
label_list = processor.get_labels()
num_labels = len(label_list)
# 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(
args.config_name if args.config_name else args.model_name_or_path,
num_labels=num_labels,
finetuning_task=args.task_name,
output_attentions=True,
cache_dir=args.cache_dir,
)
tokenizer = AutoTokenizer.from_pretrained(
args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
cache_dir=args.cache_dir,
)
model = AutoModelForSequenceClassification.from_pretrained(
args.model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
cache_dir=args.cache_dir,
)
# Distributed and parallel training
model.to(args.device)
if args.local_rank != -1:
model = nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=True
)
elif args.n_gpu > 1:
model = nn.DataParallel(model)
# Print/save training arguments
os.makedirs(args.output_dir, exist_ok=True)
torch.save(args, os.path.join(args.output_dir, "run_args.bin"))
logger.info("Training/evaluation parameters %s", args)
# Prepare dataset for the GLUE task
eval_dataset = GlueDataset(args, tokenizer=tokenizer, mode="dev")
if args.data_subset > 0:
eval_dataset = Subset(eval_dataset, list(range(min(args.data_subset, len(eval_dataset)))))
eval_sampler = SequentialSampler(eval_dataset) if args.local_rank == -1 else DistributedSampler(eval_dataset)
eval_dataloader = DataLoader(
eval_dataset, sampler=eval_sampler, batch_size=args.batch_size, collate_fn=default_data_collator
)
# Compute head entropy and importance score
compute_heads_importance(args, model, eval_dataloader)
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
head_mask = mask_heads(args, model, eval_dataloader)
prune_heads(args, model, eval_dataloader, head_mask)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/bertology/requirements.txt
|
transformers == 3.5.1
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/bertology/run_prune_gpt.py
|
#!/usr/bin/env python3
""" This script is adapted from the Bertology pruning code (https://github.com/huggingface/transformers/blob/783d7d2629e97c5f0c5f9ef01b8c66410275c204/examples/research_projects/bertology/run_bertology.py)
to prune GPT-like models. The author is @altsoph.
"""
import argparse
import logging
import os
from datetime import datetime
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, TensorDataset
from tqdm import tqdm
from transformers import GPT2LMHeadModel
logger = logging.getLogger(__name__)
def save_model(model, dirpath):
# save results
if os.path.exists(dirpath):
if os.path.exists(os.path.join(dirpath, "config.json")) and os.path.isfile(
os.path.join(dirpath, "config.json")
):
os.remove(os.path.join(dirpath, "config.json"))
if os.path.exists(os.path.join(dirpath, "pytorch_model.bin")) and os.path.isfile(
os.path.join(dirpath, "pytorch_model.bin")
):
os.remove(os.path.join(dirpath, "pytorch_model.bin"))
else:
os.makedirs(dirpath)
model.save_pretrained(dirpath)
def entropy(p, unlogit=False):
"""Compute the entropy of a probability distribution"""
exponent = 2
if unlogit:
p = torch.pow(p, exponent)
plogp = p * torch.log(p)
plogp[p == 0] = 0
return -plogp.sum(dim=-1)
def print_2d_tensor(tensor):
"""Print a 2D tensor"""
logger.info("lv, h >\t" + "\t".join(f"{x + 1}" for x in range(len(tensor))))
for row in range(len(tensor)):
if tensor.dtype != torch.long:
logger.info(f"layer {row + 1}:\t" + "\t".join(f"{x:.5f}" for x in tensor[row].cpu().data))
else:
logger.info(f"layer {row + 1}:\t" + "\t".join(f"{x:d}" for x in tensor[row].cpu().data))
def compute_heads_importance(
args, model, eval_dataloader, compute_entropy=True, compute_importance=True, head_mask=None, actually_pruned=False
):
"""This method shows how to compute:
- head attention entropy
- head importance scores according to http://arxiv.org/abs/1905.10650
"""
# Prepare our tensors
n_layers, n_heads = model.config.num_hidden_layers, model.config.num_attention_heads
head_importance = torch.zeros(n_layers, n_heads).to(args.device)
attn_entropy = torch.zeros(n_layers, n_heads).to(args.device)
if head_mask is None:
head_mask = torch.ones(n_layers, n_heads).to(args.device)
head_mask.requires_grad_(requires_grad=True)
# If actually pruned attention multi-head, set head mask to None to avoid shape mismatch
if actually_pruned:
head_mask = None
tot_tokens = 0.0
total_loss = 0.0
for step, inputs in enumerate(tqdm(eval_dataloader, desc="Iteration", disable=args.local_rank not in [-1, 0])):
inputs = tuple(t.to(args.device) for t in inputs)
(input_ids,) = inputs
# Do a forward pass (not with torch.no_grad() since we need gradients for importance score - see below)
outputs = model(input_ids, labels=input_ids, head_mask=head_mask)
# (loss), lm_logits, presents, (all hidden_states), (attentions)
loss, _, all_attentions = (
outputs[0],
outputs[1],
outputs[-1],
) # Loss and logits are the first, attention the last
loss.backward() # Backpropagate to populate the gradients in the head mask
total_loss += loss.detach().cpu().numpy()
if compute_entropy:
for layer, attn in enumerate(all_attentions):
masked_entropy = entropy(attn.detach(), True)
attn_entropy[layer] += masked_entropy.sum(-1).sum(0).sum(0).detach()
if compute_importance:
head_importance += head_mask.grad.abs().detach()
tot_tokens += torch.ones_like(input_ids).float().detach().sum().data
# Normalize
attn_entropy /= tot_tokens
head_importance /= tot_tokens
# Layerwise importance normalization
if not args.dont_normalize_importance_by_layer:
exponent = 2
norm_by_layer = torch.pow(torch.pow(head_importance, exponent).sum(-1), 1 / exponent)
head_importance /= norm_by_layer.unsqueeze(-1) + 1e-20
if not args.dont_normalize_global_importance:
head_importance = (head_importance - head_importance.min()) / (head_importance.max() - head_importance.min())
# Print matrices
if compute_entropy:
logger.info("Attention entropies")
print_2d_tensor(attn_entropy)
if compute_importance:
logger.info("Head importance scores")
print_2d_tensor(head_importance)
logger.info("Head ranked by importance scores")
head_ranks = torch.zeros(head_importance.numel(), dtype=torch.long, device=args.device)
head_ranks[head_importance.view(-1).sort(descending=True)[1]] = torch.arange(
head_importance.numel(), device=args.device
)
head_ranks = head_ranks.view_as(head_importance)
print_2d_tensor(head_ranks)
return attn_entropy, head_importance, total_loss
def mask_heads(args, model, eval_dataloader):
"""This method shows how to mask head (set some heads to zero), to test the effect on the network,
based on the head importance scores, as described in Michel et al. (http://arxiv.org/abs/1905.10650)
"""
_, head_importance, loss = compute_heads_importance(args, model, eval_dataloader, compute_entropy=False)
original_score = 1 / loss # instead of downsteam score use the LM loss
logger.info("Pruning: original score: %f, threshold: %f", original_score, original_score * args.masking_threshold)
new_head_mask = torch.ones_like(head_importance)
num_to_mask = max(1, int(new_head_mask.numel() * args.masking_amount))
current_score = original_score
while current_score >= original_score * args.masking_threshold:
head_mask = new_head_mask.clone().detach() # save current head mask
# heads from least important to most - keep only not-masked heads
head_importance[head_mask == 0.0] = float("Inf")
current_heads_to_mask = head_importance.view(-1).sort()[1]
if len(current_heads_to_mask) <= num_to_mask:
print("BREAK BY num_to_mask")
break
# mask heads
current_heads_to_mask = current_heads_to_mask[:num_to_mask]
logger.info("Heads to mask: %s", str(current_heads_to_mask.tolist()))
new_head_mask = new_head_mask.view(-1)
new_head_mask[current_heads_to_mask] = 0.0
new_head_mask = new_head_mask.view_as(head_mask)
new_head_mask = new_head_mask.clone().detach()
print_2d_tensor(new_head_mask)
# Compute metric and head importance again
_, head_importance, loss = compute_heads_importance(
args, model, eval_dataloader, compute_entropy=False, head_mask=new_head_mask
)
current_score = 1 / loss
logger.info(
"Masking: current score: %f, remaining heads %d (%.1f percents)",
current_score,
new_head_mask.sum(),
new_head_mask.sum() / new_head_mask.numel() * 100,
)
logger.info("Final head mask")
print_2d_tensor(head_mask)
np.save(os.path.join(args.output_dir, "head_mask.npy"), head_mask.detach().cpu().numpy())
return head_mask
def prune_heads(args, model, eval_dataloader, head_mask):
"""This method shows how to prune head (remove heads weights) based on
the head importance scores as described in Michel et al. (http://arxiv.org/abs/1905.10650)
"""
# Try pruning and test time speedup
# Pruning is like masking but we actually remove the masked weights
before_time = datetime.now()
_, _, loss = compute_heads_importance(
args, model, eval_dataloader, compute_entropy=False, compute_importance=False, head_mask=head_mask
)
score_masking = 1 / loss
original_time = datetime.now() - before_time
original_num_params = sum(p.numel() for p in model.parameters())
heads_to_prune = {
layer: (1 - head_mask[layer].long()).nonzero().squeeze().tolist() for layer in range(len(head_mask))
}
for k, v in heads_to_prune.items():
if isinstance(v, int):
heads_to_prune[k] = [
v,
]
assert sum(len(h) for h in heads_to_prune.values()) == (1 - head_mask.long()).sum().item()
model.prune_heads(heads_to_prune)
pruned_num_params = sum(p.numel() for p in model.parameters())
before_time = datetime.now()
_, _, loss = compute_heads_importance(
args,
model,
eval_dataloader,
compute_entropy=False,
compute_importance=False,
head_mask=None,
actually_pruned=True,
)
score_pruning = 1 / loss
new_time = datetime.now() - before_time
logger.info(
"Pruning: original num of params: %.2e, after pruning %.2e (%.1f percents)",
original_num_params,
pruned_num_params,
pruned_num_params / original_num_params * 100,
)
logger.info("Pruning: score with masking: %f score with pruning: %f", score_masking, score_pruning)
logger.info("Pruning: speed ratio (original timing / new timing): %f percents", original_time / new_time * 100)
save_model(model, args.output_dir)
def main():
parser = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--data_dir",
default=None,
type=str,
required=True,
help="The input data dir. Should contain the .tsv files (or other data files) for the task.",
)
parser.add_argument(
"--model_name_or_path",
default=None,
type=str,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models",
)
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help="The output directory where the model predictions and checkpoints will be written.",
)
# Other parameters
parser.add_argument(
"--config_name",
default="",
type=str,
help="Pretrained config name or path if not the same as model_name_or_path",
)
parser.add_argument(
"--tokenizer_name",
default="",
type=str,
help="Pretrained tokenizer name or path if not the same as model_name_or_path",
)
parser.add_argument(
"--cache_dir",
default=None,
type=str,
help="Where do you want to store the pre-trained models downloaded from s3",
)
parser.add_argument(
"--data_subset", type=int, default=-1, help="If > 0: limit the data to a subset of data_subset instances."
)
parser.add_argument(
"--overwrite_output_dir", action="store_true", help="Whether to overwrite data in output directory"
)
parser.add_argument(
"--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
)
parser.add_argument(
"--dont_normalize_importance_by_layer", action="store_true", help="Don't normalize importance score by layers"
)
parser.add_argument(
"--dont_normalize_global_importance",
action="store_true",
help="Don't normalize all importance scores between 0 and 1",
)
parser.add_argument(
"--try_masking", action="store_true", help="Whether to try to mask head until a threshold of accuracy."
)
parser.add_argument(
"--masking_threshold",
default=0.9,
type=float,
help="masking threshold in term of metrics (stop masking when metric < threshold * original metric value).",
)
parser.add_argument(
"--masking_amount", default=0.1, type=float, help="Amount to heads to masking at each masking step."
)
parser.add_argument("--metric_name", default="acc", type=str, help="Metric to use for head masking.")
parser.add_argument(
"--max_seq_length",
default=128,
type=int,
help=(
"The maximum total input sequence length after WordPiece tokenization. \n"
"Sequences longer than this will be truncated, sequences shorter padded."
),
)
parser.add_argument("--batch_size", default=1, type=int, help="Batch size.")
parser.add_argument("--seed", type=int, default=42)
parser.add_argument("--local_rank", type=int, default=-1, help="local_rank for distributed training on gpus")
parser.add_argument("--no_cuda", action="store_true", help="Whether not to use CUDA when available")
parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
args = parser.parse_args()
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
# Setup devices and distributed training
if args.local_rank == -1 or args.no_cuda:
args.device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = 0 if args.no_cuda else torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank)
args.device = torch.device("cuda", args.local_rank)
args.n_gpu = 1
torch.distributed.init_process_group(backend="nccl") # Initializes the distributed backend
# Setup logging
logging.basicConfig(level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN)
logger.info("device: {} n_gpu: {}, distributed: {}".format(args.device, args.n_gpu, bool(args.local_rank != -1)))
model = GPT2LMHeadModel.from_pretrained(args.model_name_or_path)
# Distributed and parallel training
model.to(args.device)
if args.local_rank != -1:
model = nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=True
)
elif args.n_gpu > 1:
model = nn.DataParallel(model)
# Print/save training arguments
os.makedirs(args.output_dir, exist_ok=True)
torch.save(args, os.path.join(args.output_dir, "run_args.bin"))
logger.info("Training/evaluation parameters %s", args)
# Prepare dataset
numpy_data = np.concatenate(
[
np.loadtxt(args.data_dir, dtype=np.int64),
]
)
train_tensor_dataset = (torch.from_numpy(numpy_data),)
train_data = TensorDataset(*train_tensor_dataset)
train_sampler = RandomSampler(train_data)
eval_dataloader = DataLoader(train_data, sampler=train_sampler, batch_size=args.batch_size)
# Compute head entropy and importance score
compute_heads_importance(args, model, eval_dataloader)
# Try head masking (set heads to zero until the score goes under a threshole)
# and head pruning (remove masked heads and see the effect on the network)
if args.try_masking and args.masking_threshold > 0.0 and args.masking_threshold < 1.0:
head_mask = mask_heads(args, model, eval_dataloader)
prune_heads(args, model, eval_dataloader, head_mask)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/self-training-text-classification/README.md
|
# Self-training
This is an implementation of the self-training algorithm (without task augmentation) in the [EMNLP 2021](https://2021.emnlp.org/) paper: [STraTA: Self-Training with Task Augmentation for Better Few-shot Learning](https://arxiv.org/abs/2109.06270). Please check out https://github.com/google-research/google-research/tree/master/STraTA for the original codebase.
**Note**: The code can be used as a tool for automatic data labeling.
## Table of Contents
* [Installation](#installation)
* [Self-training](#self-training)
* [Running self-training with a base model](#running-self-training-with-a-base-model)
* [Hyperparameters for self-training](#hyperparameters-for-self-training)
* [Distributed training](#distributed-training)
* [Demo](#demo)
* [How to cite](#how-to-cite)
## Installation
This repository is tested on Python 3.8+, PyTorch 1.10+, and the 🤗 Transformers 4.16+.
You should install all necessary Python packages in a [virtual environment](https://docs.python.org/3/library/venv.html). If you are unfamiliar with Python virtual environments, please check out the [user guide](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/).
Below, we create a virtual environment with the [Anaconda Python distribution](https://www.anaconda.com/products/distribution) and activate it.
```sh
conda create -n strata python=3.9
conda activate strata
```
Next, you need to install 🤗 Transformers. Please refer to [🤗 Transformers installation page](https://github.com/huggingface/transformers#installation) for a detailed guide.
```sh
pip install transformers
```
Finally, install all necessary Python packages for our self-training algorithm.
```sh
pip install -r STraTA/selftraining/requirements.txt
```
This will install PyTorch as a backend.
## Self-training
### Running self-training with a base model
The following example code shows how to run our self-training algorithm with a base model (e.g., `BERT`) on the `SciTail` science entailment dataset, which has two classes `['entails', 'neutral']`. We assume that you have a data directory that includes some training data (e.g., `train.csv`), evaluation data (e.g., `eval.csv`), and unlabeled data (e.g., `infer.csv`).
```python
import os
from selftraining import selftrain
data_dir = '/path/to/your/data/dir'
parameters_dict = {
'max_selftrain_iterations': 100,
'model_name_or_path': '/path/to/your/base/model', # could be the id of a model hosted by 🤗 Transformers
'output_dir': '/path/to/your/output/dir',
'train_file': os.path.join(data_dir, 'train.csv'),
'infer_file': os.path.join(data_dir, 'infer.csv'),
'eval_file': os.path.join(data_dir, 'eval.csv'),
'evaluation_strategy': 'steps',
'task_name': 'scitail',
'label_list': ['entails', 'neutral'],
'per_device_train_batch_size': 32,
'per_device_eval_batch_size': 8,
'max_length': 128,
'learning_rate': 2e-5,
'max_steps': 100000,
'eval_steps': 1,
'early_stopping_patience': 50,
'overwrite_output_dir': True,
'do_filter_by_confidence': False,
# 'confidence_threshold': 0.3,
'do_filter_by_val_performance': True,
'finetune_on_labeled_data': False,
'seed': 42,
}
selftrain(**parameters_dict)
```
**Note**: We checkpoint periodically during self-training. In case of preemptions, just re-run the above script and self-training will resume from the latest iteration.
### Hyperparameters for self-training
If you have development data, you might want to tune some hyperparameters for self-training.
Below are hyperparameters that could provide additional gains for your task.
- `finetune_on_labeled_data`: If set to `True`, the resulting model from each self-training iteration is further fine-tuned on the original labeled data before the next self-training iteration. Intuitively, this would give the model a chance to "correct" ifself after being trained on pseudo-labeled data.
- `do_filter_by_confidence`: If set to `True`, the pseudo-labeled data in each self-training iteration is filtered based on the model confidence. For instance, if `confidence_threshold` is set to `0.3`, pseudo-labeled examples with a confidence score less than or equal to `0.3` will be discarded. Note that `confidence_threshold` should be greater or equal to `1/num_labels`, where `num_labels` is the number of class labels. Filtering out the lowest-confidence pseudo-labeled examples could be helpful in some cases.
- `do_filter_by_val_performance`: If set to `True`, the pseudo-labeled data in each self-training iteration is filtered based on the current validation performance. For instance, if your validation performance is 80% accuracy, you might want to get rid of 20% of the pseudo-labeled data with the lowest the confidence scores.
### Distributed training
We strongly recommend distributed training with multiple accelerators. To activate distributed training, please try one of the following methods:
1. Run `accelerate config` and answer to the questions asked. This will save a `default_config.yaml` file in your cache folder for 🤗 Accelerate. Now, you can run your script with the following command:
```sh
accelerate launch your_script.py --args_to_your_script
```
2. Run your script with the following command:
```sh
python -m torch.distributed.launch --nnodes="{$NUM_NODES}" --nproc_per_node="{$NUM_TRAINERS}" --your_script.py --args_to_your_script
```
3. Run your script with the following command:
```sh
torchrun --nnodes="{$NUM_NODES}" --nproc_per_node="{$NUM_TRAINERS}" --your_script.py --args_to_your_script
```
## Demo
Please check out `run.sh` to see how to perform our self-training algorithm with a `BERT` Base model on the SciTail science entailment dataset using 8 labeled examples per class. You can configure your training environment by specifying `NUM_NODES` and `NUM_TRAINERS` (number of processes per node). To launch the script, simply run `source run.sh`.
## How to cite
If you extend or use this code, please cite the [paper](https://arxiv.org/abs/2109.06270) where it was introduced:
```bibtex
@inproceedings{vu-etal-2021-strata,
title = "{ST}ra{TA}: Self-Training with Task Augmentation for Better Few-shot Learning",
author = "Vu, Tu and
Luong, Minh-Thang and
Le, Quoc and
Simon, Grady and
Iyyer, Mohit",
booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2021",
address = "Online and Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.emnlp-main.462",
doi = "10.18653/v1/2021.emnlp-main.462",
pages = "5715--5731",
}
```
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/self-training-text-classification/selftraining.py
|
# coding=utf-8
# Copyright 2022 The Google Research Authors.
#
# 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.
"""Self-training for sequence classification."""
import argparse
import dataclasses
import json
import logging
import os
import shutil
from typing import List, Optional
import datasets
from accelerate import Accelerator
from datasets import load_dataset
from finetuning import finetune
from tqdm.auto import tqdm
import transformers
from transformers import AutoConfig, set_seed
from transformers.trainer_utils import IntervalStrategy
logger = logging.getLogger(__name__)
MODEL_BIN_FILE = "pytorch_model.bin"
@dataclasses.dataclass
class STModelArguments:
"""Arguments pertaining to which config/tokenizer/model we are going to fine-tune from."""
model_name_or_path: str = dataclasses.field(
metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models."}
)
cache_dir: Optional[str] = dataclasses.field(
default=None,
metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co."},
)
@dataclasses.dataclass
class STDataArguments:
"""Arguments pertaining to what data we are going to input our model for training and evaluation."""
train_file: str = dataclasses.field(metadata={"help": "A csv or a json file containing the training data."})
infer_file: str = dataclasses.field(metadata={"help": "A csv or a json file containing the data to predict on."})
eval_file: Optional[str] = dataclasses.field(
default=None, metadata={"help": "A csv or a json file containing the validation data."}
)
task_name: Optional[str] = dataclasses.field(
default=None,
metadata={"help": "The name of the task to train on."},
)
label_list: Optional[List[str]] = dataclasses.field(
default=None, metadata={"help": "The list of labels for the task."}
)
@dataclasses.dataclass
class STTrainingArguments:
"""Training arguments pertaining to the training loop itself."""
output_dir: str = dataclasses.field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."}
)
eval_metric: Optional[str] = dataclasses.field(
default="accuracy", metadata={"help": "The evaluation metric used for the task."}
)
evaluation_strategy: Optional[str] = dataclasses.field(
default="no",
metadata={
"help": 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
},
)
early_stopping_patience: Optional[int] = dataclasses.field(
default=10,
metadata={"help": "Number of evaluation calls with no improvement after which training will be stopped."},
)
early_stopping_threshold: Optional[float] = dataclasses.field(
default=0.0,
metadata={
"help": "How much the specified evaluation metric must improve to satisfy early stopping conditions."
},
)
do_filter_by_confidence: Optional[bool] = dataclasses.field(
default=False,
metadata={"help": "Whether to filter the pseudo-labeled data based on the confidence score."},
)
do_filter_by_val_performance: Optional[bool] = dataclasses.field(
default=False,
metadata={"help": "Whether to filter the pseudo-labeled data based on the validation performance."},
)
finetune_on_labeled_data: Optional[bool] = dataclasses.field(
default=False,
metadata={"help": "Whether to fine-tune on labeled data after pseudo training."},
)
confidence_threshold: Optional[float] = dataclasses.field(
default=0.0,
metadata={"help": "Confidence threshold for pseudo-labeled data filtering."},
)
max_selftrain_iterations: Optional[int] = dataclasses.field(
default=100,
metadata={"help": "Number of evaluation calls with no improvement after which training will be stopped."},
)
seed: Optional[int] = dataclasses.field(
default=None,
metadata={"help": "Random seed for initialization."},
)
def create_pseudo_labeled_data(args, infer_input, infer_output, eval_result, id2label, next_data_dir):
"""Create pseudeo labeled data for the next self-training iteration."""
dataset = datasets.concatenate_datasets([infer_input, infer_output], axis=1)
if args.do_filter_by_confidence:
dataset = dataset.filter(lambda example: example["probability"] > args.confidence_threshold)
if args.do_filter_by_val_performance:
assert eval_result >= 0.0 and eval_result <= 1.0
num_selected_rows = int(eval_result * len(dataset))
print(num_selected_rows)
dataset = dataset.sort("probability", reverse=True)
dataset = dataset.select(range(num_selected_rows))
dataset = dataset.remove_columns(["label", "probability"])
dataset = dataset.rename_column("prediction", "label")
dataset = dataset.map(lambda example: {"label": id2label[example["label"]]})
dataset = dataset.shuffle(seed=args.seed)
pseudo_labeled_data_file = os.path.join(next_data_dir, f"train_pseudo.{args.data_file_extension}")
if args.data_file_extension == "csv":
dataset.to_csv(pseudo_labeled_data_file, index=False)
else:
dataset.to_json(pseudo_labeled_data_file)
def selftrain(model_name_or_path, train_file, infer_file, output_dir, **kwargs):
"""Self-training a pre-trained model on a downstream task.
Args:
model_name_or_path: Path to pretrained model or model identifier from
huggingface.co/models.
train_file: A csv or a json file containing the training data.
infer_file: A csv or a json file containing the data to predict on.
output_dir: The output directory where the model predictions and checkpoints
will be written.
**kwargs: Dictionary of key/value pairs with which to update the
configuration object after loading. The values in kwargs of any keys which
are configuration attributes will be used to override the loaded values.
"""
# Initialize the accelerator. We will let the accelerator handle device
# placement for us.
accelerator = Accelerator()
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
logger.info(accelerator.state)
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
model_args = STModelArguments(model_name_or_path=model_name_or_path)
data_args = STDataArguments(train_file=train_file, infer_file=infer_file)
training_args = STTrainingArguments(output_dir=output_dir)
args = argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(arg_class).items():
setattr(args, key, value)
for key, value in kwargs.items():
if hasattr(args, key):
setattr(args, key, value)
# Sanity checks
data_files = {}
args.data_file_extension = None
# You need to provide the training data and the data to predict on
assert args.train_file is not None
assert args.infer_file is not None
data_files["train"] = args.train_file
data_files["infer"] = args.infer_file
if args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
data_files["eval"] = args.eval_file
for key in data_files:
extension = data_files[key].split(".")[-1]
assert extension in ["csv", "json"], f"`{key}_file` should be a csv or a json file."
if args.data_file_extension is None:
args.data_file_extension = extension
else:
assert extension == args.data_file_extension, f"`{key}_file` should be a {args.data_file_extension} file`."
assert (
args.eval_metric in datasets.list_metrics()
), f"{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}."
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed)
logger.info("Creating the initial data directory for self-training...")
data_dir_format = f"{args.output_dir}/self-train_iter-{{}}".format
initial_data_dir = data_dir_format(0)
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir, exist_ok=True)
os.makedirs(initial_data_dir, exist_ok=True)
accelerator.wait_for_everyone()
best_iteration = None
best_eval_result = None
early_stopping_patience_counter = 0
should_training_stop = False
# Show the progress bar
progress_bar = tqdm(range(args.max_selftrain_iterations), disable=not accelerator.is_local_main_process)
# Self-train
for iteration in range(0, int(args.max_selftrain_iterations)):
current_data_dir = data_dir_format(iteration)
assert os.path.exists(current_data_dir)
# Stage 1: initial fine-tuning for iteration = 0 or pseudo-training for
# iteration > 0
current_output_dir = os.path.join(current_data_dir, "stage-1")
arguments_dict = {
"accelerator": accelerator,
"model_name_or_path": args.model_name_or_path,
"cache_dir": args.cache_dir,
"do_train": True,
"train_file": data_files["train"] if iteration == 0 else data_files["train_pseudo"],
"do_eval": True if args.eval_file is not None else False,
"eval_file": data_files["eval"],
"do_predict": True,
"infer_file": data_files["infer"],
"task_name": args.task_name,
"label_list": args.label_list,
"output_dir": current_output_dir,
"eval_metric": args.eval_metric,
"evaluation_strategy": args.evaluation_strategy,
"early_stopping_patience": args.early_stopping_patience,
"early_stopping_threshold": args.early_stopping_threshold,
"seed": args.seed,
}
# Add additional training arguments
for key, value in kwargs.items():
if key not in arguments_dict and not hasattr(training_args, key):
arguments_dict.update({key: value})
model_bin_file_path = os.path.join(current_output_dir, "best-checkpoint", MODEL_BIN_FILE)
if os.path.exists(model_bin_file_path):
logger.info(
"Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 1.",
model_bin_file_path,
iteration,
)
else:
logger.info("***** Running self-training: iteration: %d, stage: 1 *****", iteration)
finetune(**arguments_dict)
accelerator.wait_for_everyone()
assert os.path.exists(model_bin_file_path)
logger.info("Self-training job completed: iteration: %d, stage: 1.", iteration)
if iteration > 0 and args.finetune_on_labeled_data:
# Stage 2 (optional): fine-tuning on the original labeled data
model_path = os.path.join(current_output_dir, "best-checkpoint")
current_output_dir = os.path.join(current_data_dir, "stage-2")
# Update arguments_dict
arguments_dict["model_name_or_path"] = model_path
arguments_dict["train_file"] = data_files["train"]
arguments_dict["output_dir"] = current_output_dir
model_bin_file_path = os.path.join(current_output_dir, "best-checkpoint", MODEL_BIN_FILE)
if os.path.exists(model_bin_file_path):
logger.info(
"Found existing model checkpoint at %s. Skipping self-training: iteration: %d, stage: 2.",
model_bin_file_path,
iteration,
)
else:
logger.info("***** Running self-training: iteration: %d, stage: 2 *****", iteration)
finetune(**arguments_dict)
accelerator.wait_for_everyone()
assert os.path.exists(model_bin_file_path)
logger.info("Self-training job completed: iteration: %d, stage: 2.", iteration)
new_iteration = iteration
next_data_dir = data_dir_format(iteration + 1)
config = AutoConfig.from_pretrained(os.path.join(current_output_dir, "best-checkpoint"))
id2label = config.id2label
eval_results_file = os.path.join(current_output_dir, "eval_results_best-checkpoint.json")
test_results_file = os.path.join(current_output_dir, "test_results_best-checkpoint.json")
assert os.path.exists(eval_results_file)
with open(eval_results_file, "r") as f:
eval_result = float(json.load(f)[args.eval_metric])
infer_output_file = os.path.join(current_output_dir, "infer_output_best-checkpoint.csv")
assert os.path.exists(infer_output_file)
# Loading the dataset from local csv or json files.
infer_input = load_dataset(args.data_file_extension, data_files={"data": data_files["infer"]})["data"]
infer_output = load_dataset("csv", data_files={"data": infer_output_file})["data"]
if accelerator.is_main_process:
os.makedirs(next_data_dir, exist_ok=True)
shutil.copy(eval_results_file, os.path.join(output_dir, f"eval_results_iter-{iteration}.json"))
if os.path.exists(test_results_file):
shutil.copy(eval_results_file, os.path.join(output_dir, f"test_results_iter-{iteration}.json"))
create_pseudo_labeled_data(args, infer_input, infer_output, eval_result, id2label, next_data_dir)
accelerator.wait_for_everyone()
data_files["train_pseudo"] = os.path.join(next_data_dir, f"train_pseudo.{args.data_file_extension}")
if args.evaluation_strategy != IntervalStrategy.NO.value:
new_eval_result = eval_result
if best_iteration is None:
best_iteration = new_iteration
best_eval_result = new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
best_iteration = new_iteration
best_eval_result = new_eval_result
early_stopping_patience_counter = 0
else:
if new_eval_result == best_eval_result:
best_iteration = new_iteration
best_eval_result = new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
should_training_stop = True
progress_bar.update(1)
if should_training_stop:
break
if best_iteration is not None:
# Save the best iteration
logger.info("Best iteration: %d", best_iteration)
logger.info("Best evaluation result: %s = %f", args.eval_metric, best_eval_result)
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(output_dir, f"eval_results_iter-{iteration}.json"),
os.path.join(output_dir, "eval_results_best-iteration.json"),
)
else:
# Assume that the last iteration is the best
logger.info("Best iteration: %d", args.max_selftrain_iterations - 1)
logger.info("Best evaluation result: %s = %f", args.eval_metric, eval_result)
accelerator.wait_for_everyone()
if accelerator.is_main_process:
shutil.copy(
os.path.join(output_dir, f"eval_results_iter-{args.max_selftrain_iterations - 1}.json"),
os.path.join(output_dir, "eval_results_best-iteration.json"),
)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/self-training-text-classification/requirements.txt
|
accelerate
datasets >= 1.8.0
protobuf
scikit-learn
scipy
sentencepiece != 0.1.92
torch >= 1.3
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/self-training-text-classification/run.sh
|
# Copyright 2022 The Google Research Authors.
#
# 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.
#!/bin/bash
# Create a virtual environment
conda deactivate
conda update conda -y
conda update anaconda -y
pip install --upgrade pip
python3 -m pip install --user virtualenv
conda create -n strata python=3.9 -y
conda activate strata
# Install all necessary packages
pip install transformers
pip install -r requirements.txt
# Download and prepare data
WORK_DIR="/tmp/strata"
rm -rf "${WORK_DIR}" && mkdir -p "${WORK_DIR}"
wget https://storage.googleapis.com/gresearch/strata/demo.zip -P "${WORK_DIR}"
DEMO_ZIP_FILE="${WORK_DIR}/demo.zip"
unzip "${DEMO_ZIP_FILE}" -d "${WORK_DIR}" && rm "${DEMO_ZIP_FILE}"
DATA_DIR="${WORK_DIR}/demo/scitail-8"
OUTPUT_DIR="/tmp/output"
rm -rf "${OUTPUT_DIR}" && mkdir -p "${OUTPUT_DIR}"
# Specific hyperparameters
MODEL_NAME_OR_PATH="bert-base-uncased"
NUM_NODES=1
NUM_TRAINERS=4
LAUNCH_SCRIPT="torchrun --nnodes='${NUM_NODES}' --nproc_per_node='${NUM_TRAINERS}' python -c"
MAX_SELFTRAIN_ITERATIONS=100
TRAIN_FILE="train.csv"
INFER_FILE="infer.csv"
EVAL_FILE="eval_256.csv"
MAX_STEPS=100000
# Start self-training
${LAUNCH_SCRIPT} "
import os
from selftraining import selftrain
data_dir = '${DATA_DIR}'
parameters_dict = {
'max_selftrain_iterations': ${MAX_SELFTRAIN_ITERATIONS},
'model_name_or_path': '${MODEL_NAME_OR_PATH}',
'output_dir': '${OUTPUT_DIR}',
'train_file': os.path.join(data_dir, '${TRAIN_FILE}'),
'infer_file': os.path.join(data_dir, '${INFER_FILE}'),
'eval_file': os.path.join(data_dir, '${EVAL_FILE}'),
'evaluation_strategy': 'steps',
'task_name': 'scitail',
'label_list': ['entails', 'neutral'],
'per_device_train_batch_size': 32,
'per_device_eval_batch_size': 8,
'max_length': 128,
'learning_rate': 2e-5,
'max_steps': ${MAX_STEPS},
'eval_steps': 1,
'early_stopping_patience': 50,
'overwrite_output_dir': True,
'do_filter_by_confidence': False,
'do_filter_by_val_performance': True,
'finetune_on_labeled_data': False,
'seed': 42,
}
selftrain(**parameters_dict)
"
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/self-training-text-classification/finetuning.py
|
# coding=utf-8
# Copyright 2022 The Google Research Authors.
#
# 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.
"""Fine-tuning the library models for sequence classification."""
import argparse
import dataclasses
import json
import logging
import math
import os
import random
import shutil
from typing import List, Optional
import datasets
import numpy as np
import pandas as pd
import torch
from datasets import load_dataset, load_metric
from torch.utils.data import DataLoader
from tqdm.auto import tqdm
from transformers import (
AdamW,
AutoConfig,
AutoModelForSequenceClassification,
AutoTokenizer,
DataCollatorWithPadding,
default_data_collator,
get_scheduler,
set_seed,
)
from transformers.file_utils import ExplicitEnum
from transformers.trainer_utils import IntervalStrategy
logger = logging.getLogger(__name__)
class Split(ExplicitEnum):
TRAIN = "train"
EVAL = "eval"
TEST = "test"
INFER = "infer"
@dataclasses.dataclass
class FTModelArguments:
"""Arguments pertaining to which config/tokenizer/model we are going to fine-tune from."""
model_name_or_path: str = dataclasses.field(
metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models."}
)
use_fast_tokenizer: Optional[bool] = dataclasses.field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
cache_dir: Optional[str] = dataclasses.field(
default=None,
metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co."},
)
@dataclasses.dataclass
class FTDataArguments:
"""Arguments pertaining to what data we are going to input our model for training and evaluation."""
train_file: str = dataclasses.field(
default=None, metadata={"help": "A csv or a json file containing the training data."}
)
eval_file: Optional[str] = dataclasses.field(
default=None, metadata={"help": "A csv or a json file containing the validation data."}
)
test_file: Optional[str] = dataclasses.field(
default=None, metadata={"help": "A csv or a json file containing the test data."}
)
infer_file: Optional[str] = dataclasses.field(
default=None, metadata={"help": "A csv or a json file containing the data to predict on."}
)
task_name: Optional[str] = dataclasses.field(
default=None,
metadata={"help": "The name of the task to train on."},
)
label_list: Optional[List[str]] = dataclasses.field(
default=None, metadata={"help": "The list of labels for the task."}
)
max_length: Optional[int] = dataclasses.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."
)
},
)
pad_to_max_length: Optional[bool] = dataclasses.field(
default=False,
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."
)
},
)
@dataclasses.dataclass
class FTTrainingArguments:
"""Training arguments pertaining to the training loop itself."""
output_dir: str = dataclasses.field(
metadata={"help": "The output directory where the model predictions and checkpoints will be written."}
)
do_train: Optional[bool] = dataclasses.field(
default=False,
metadata={"help": "Whether to run training or not."},
)
do_eval: Optional[bool] = dataclasses.field(
default=False,
metadata={"help": "Whether to run evaluation on the validation set or not."},
)
do_predict: Optional[bool] = dataclasses.field(
default=False,
metadata={"help": "Whether to run inference on the inference set or not."},
)
seed: Optional[int] = dataclasses.field(
default=42,
metadata={"help": "Random seed that will be set at the beginning of training."},
)
per_device_train_batch_size: Optional[int] = dataclasses.field(
default=8,
metadata={"help": "The batch size per GPU/TPU core/CPU for training."},
)
per_device_eval_batch_size: Optional[int] = dataclasses.field(
default=8,
metadata={"help": "The batch size per GPU/TPU core/CPU for evaluation."},
)
weight_decay: Optional[float] = dataclasses.field(
default=0.0,
metadata={
"help": (
"The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights in"
" [`AdamW`] optimizer."
)
},
)
learning_rate: Optional[float] = dataclasses.field(
default=5e-5,
metadata={"help": "The initial learning rate for [`AdamW`] optimizer."},
)
gradient_accumulation_steps: Optional[int] = dataclasses.field(
default=1,
metadata={
"help": (
"Number of updates steps to accumulate the gradients for, before performing a backward/update pass."
)
},
)
max_steps: Optional[int] = dataclasses.field(
default=-1,
metadata={
"help": (
"If set to a positive number, the total number of training steps to perform. Overrides"
" `num_train_epochs`."
)
},
)
lr_scheduler_type: Optional[str] = dataclasses.field(
default="linear", metadata={"help": "The scheduler type to use."}
)
warmup_steps: Optional[int] = dataclasses.field(
default=1,
metadata={
"help": (
"Number of steps used for a linear warmup from 0 to `learning_rate`. Overrides any effect of"
" `warmup_ratio`."
)
},
)
evaluation_strategy: Optional[str] = dataclasses.field(
default="no",
metadata={
"help": 'The evaluation strategy to adopt during training. Possible values are: ["no", "step", "epoch]'
},
)
eval_steps: Optional[int] = dataclasses.field(
default=1,
metadata={"help": 'Number of update steps between two evaluations if `evaluation_strategy="steps"`.'},
)
eval_metric: Optional[str] = dataclasses.field(
default="accuracy", metadata={"help": "The evaluation metric used for the task."}
)
keep_checkpoint_max: Optional[int] = dataclasses.field(
default=1,
metadata={"help": "The maximum number of best checkpoint files to keep."},
)
early_stopping_patience: Optional[int] = dataclasses.field(
default=10,
metadata={"help": "Number of evaluation calls with no improvement after which training will be stopped."},
)
early_stopping_threshold: Optional[float] = dataclasses.field(
default=0.0,
metadata={
"help": "How much the specified evaluation metric must improve to satisfy early stopping conditions."
},
)
def train(args, accelerator, model, tokenizer, train_dataloader, optimizer, lr_scheduler, eval_dataloader=None):
"""Train a model on the given training data."""
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
logger.info("***** Running training *****")
logger.info(" Num examples = %d", args.num_examples[Split.TRAIN.value])
logger.info(" Instantaneous batch size per device = %d", args.per_device_train_batch_size)
logger.info(" Total train batch size (w. parallel, distributed & accumulation) = %d", total_batch_size)
logger.info(" Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
logger.info(" Total optimization steps = %d", args.max_steps)
# Only show the progress bar once on each machine.
progress_bar = tqdm(range(args.max_steps), disable=not accelerator.is_local_main_process)
checkpoints = None
eval_results = None
best_checkpoint = None
best_eval_result = None
early_stopping_patience_counter = 0
should_training_stop = False
epoch = 0
completed_steps = 0
train_loss = 0.0
model.zero_grad()
for _ in range(args.num_train_epochs):
epoch += 1
model.train()
for step, batch in enumerate(train_dataloader):
outputs = model(**batch)
loss = outputs.loss
loss = loss / args.gradient_accumulation_steps
accelerator.backward(loss)
train_loss += loss.item()
if step % args.gradient_accumulation_steps == 0 or step == len(train_dataloader) - 1:
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
progress_bar.update(1)
completed_steps += 1
# Evaluate during training
if (
eval_dataloader is not None
and args.evaluation_strategy == IntervalStrategy.STEPS.value
and args.eval_steps > 0
and completed_steps % args.eval_steps == 0
):
accelerator.wait_for_everyone()
new_checkpoint = f"checkpoint-{IntervalStrategy.STEPS.value}-{completed_steps}"
new_eval_result = evaluate(args, accelerator, eval_dataloader, "eval", model, new_checkpoint)[
args.eval_metric
]
logger.info(
"Evaluation result at step %d: %s = %f", completed_steps, args.eval_metric, new_eval_result
)
if checkpoints is None:
checkpoints = np.array([new_checkpoint])
eval_results = np.array([new_eval_result])
best_checkpoint = new_checkpoint
best_eval_result = new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
best_checkpoint = new_checkpoint
best_eval_result = new_eval_result
early_stopping_patience_counter = 0
else:
if new_eval_result == best_eval_result:
best_checkpoint = new_checkpoint
best_eval_result = new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
should_training_stop = True
checkpoints = np.append(checkpoints, [new_checkpoint], axis=0)
eval_results = np.append(eval_results, [new_eval_result], axis=0)
sorted_ids = np.argsort(eval_results)
eval_results = eval_results[sorted_ids]
checkpoints = checkpoints[sorted_ids]
if len(checkpoints) > args.keep_checkpoint_max:
# Delete the current worst checkpoint
checkpoint_to_remove, *checkpoints = checkpoints
eval_results = eval_results[1:]
if checkpoint_to_remove != new_checkpoint:
if accelerator.is_main_process:
shutil.rmtree(os.path.join(args.output_dir, checkpoint_to_remove), ignore_errors=True)
accelerator.wait_for_everyone()
if new_checkpoint in checkpoints:
# Save model checkpoint
checkpoint_output_dir = os.path.join(args.output_dir, new_checkpoint)
if accelerator.is_main_process:
if not os.path.exists(checkpoint_output_dir):
os.makedirs(checkpoint_output_dir)
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(checkpoint_output_dir, save_function=accelerator.save)
if accelerator.is_main_process:
tokenizer.save_pretrained(checkpoint_output_dir)
logger.info("Saving model checkpoint to %s", checkpoint_output_dir)
if completed_steps >= args.max_steps:
break
if should_training_stop:
break
# Evaluate during training
if eval_dataloader is not None and args.evaluation_strategy == IntervalStrategy.EPOCH.value:
accelerator.wait_for_everyone()
new_checkpoint = f"checkpoint-{IntervalStrategy.EPOCH.value}-{epoch}"
new_eval_result = evaluate(args, accelerator, eval_dataloader, "eval", model, new_checkpoint)[
args.eval_metric
]
logger.info("Evaluation result at epoch %d: %s = %f", epoch, args.eval_metric, new_eval_result)
if checkpoints is None:
checkpoints = np.array([new_checkpoint])
eval_results = np.array([new_eval_result])
best_checkpoint = new_checkpoint
best_eval_result = new_eval_result
else:
if new_eval_result - best_eval_result > args.early_stopping_threshold:
best_checkpoint = new_checkpoint
best_eval_result = new_eval_result
early_stopping_patience_counter = 0
else:
if new_eval_result == best_eval_result:
best_checkpoint = new_checkpoint
best_eval_result = new_eval_result
early_stopping_patience_counter += 1
if early_stopping_patience_counter >= args.early_stopping_patience:
should_training_stop = True
checkpoints = np.append(checkpoints, [new_checkpoint], axis=0)
eval_results = np.append(eval_results, [new_eval_result], axis=0)
sorted_ids = np.argsort(eval_results)
eval_results = eval_results[sorted_ids]
checkpoints = checkpoints[sorted_ids]
if len(checkpoints) > args.keep_checkpoint_max:
# Delete the current worst checkpoint
checkpoint_to_remove, *checkpoints = checkpoints
eval_results = eval_results[1:]
if checkpoint_to_remove != new_checkpoint:
if accelerator.is_main_process:
shutil.rmtree(os.path.join(args.output_dir, checkpoint_to_remove), ignore_errors=True)
accelerator.wait_for_everyone()
if new_checkpoint in checkpoints:
# Save model checkpoint
checkpoint_output_dir = os.path.join(args.output_dir, new_checkpoint)
if accelerator.is_main_process:
if not os.path.exists(checkpoint_output_dir):
os.makedirs(checkpoint_output_dir)
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(checkpoint_output_dir, save_function=accelerator.save)
if accelerator.is_main_process:
tokenizer.save_pretrained(checkpoint_output_dir)
logger.info("Saving model checkpoint to %s", checkpoint_output_dir)
if completed_steps >= args.max_steps:
break
if should_training_stop:
break
if best_checkpoint is not None:
# Save the best checkpoint
logger.info("Best checkpoint: %s", best_checkpoint)
logger.info("Best evaluation result: %s = %f", args.eval_metric, best_eval_result)
best_checkpoint_output_dir = os.path.join(args.output_dir, best_checkpoint)
if accelerator.is_main_process:
shutil.move(best_checkpoint_output_dir, os.path.join(args.output_dir, "best-checkpoint"))
shutil.rmtree(best_checkpoint_output_dir, ignore_errors=True)
accelerator.wait_for_everyone()
else:
# Assume that the last checkpoint is the best checkpoint and save it
checkpoint_output_dir = os.path.join(args.output_dir, "best-checkpoint")
if not os.path.exists(checkpoint_output_dir):
os.makedirs(checkpoint_output_dir)
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
unwrapped_model.save_pretrained(checkpoint_output_dir, save_function=accelerator.save)
if accelerator.is_main_process:
tokenizer.save_pretrained(checkpoint_output_dir)
logger.info("Saving model checkpoint to %s", checkpoint_output_dir)
return completed_steps, train_loss / completed_steps
def evaluate(args, accelerator, dataloader, eval_set, model, checkpoint, has_labels=True, write_to_file=True):
"""Evaluate a model checkpoint on the given evaluation data."""
num_examples = args.num_examples[eval_set]
eval_metric = None
completed_steps = 0
eval_loss = 0.0
all_predictions = None
all_references = None
all_probabilities = None
if has_labels:
# Get the metric function
eval_metric = load_metric(args.eval_metric)
eval_results = {}
model.eval()
for _, batch in enumerate(dataloader):
with torch.no_grad():
outputs = model(**batch)
eval_loss += outputs.loss.item()
logits = outputs.logits
predictions = logits.argmax(dim=-1) if not args.is_regression else logits.squeeze()
predictions = accelerator.gather(predictions)
if all_predictions is None:
all_predictions = predictions.detach().cpu().numpy()
else:
all_predictions = np.append(all_predictions, predictions.detach().cpu().numpy(), axis=0)
if not args.is_regression:
probabilities = logits.softmax(dim=-1).max(dim=-1).values
probabilities = accelerator.gather(probabilities)
if all_probabilities is None:
all_probabilities = probabilities.detach().cpu().numpy()
else:
all_probabilities = np.append(all_probabilities, probabilities.detach().cpu().numpy(), axis=0)
if has_labels:
references = batch["labels"]
references = accelerator.gather(references)
if all_references is None:
all_references = references.detach().cpu().numpy()
else:
all_references = np.append(all_references, references.detach().cpu().numpy(), axis=0)
eval_metric.add_batch(
predictions=predictions,
references=references,
)
completed_steps += 1
if has_labels:
eval_results.update(eval_metric.compute())
eval_results["completed_steps"] = completed_steps
eval_results["avg_eval_loss"] = eval_loss / completed_steps
if write_to_file:
accelerator.wait_for_everyone()
if accelerator.is_main_process:
results_file = os.path.join(args.output_dir, f"{eval_set}_results_{checkpoint}.json")
with open(results_file, "w") as f:
json.dump(eval_results, f, indent=4, sort_keys=True)
if write_to_file:
accelerator.wait_for_everyone()
if accelerator.is_main_process:
output_file = os.path.join(args.output_dir, f"{eval_set}_output_{checkpoint}.csv")
if not args.is_regression:
assert len(all_predictions) == len(all_probabilities)
df = pd.DataFrame(list(zip(all_predictions, all_probabilities)), columns=["prediction", "probability"])
else:
df = pd.DataFrame(all_predictions, columns=["prediction"])
df = df.head(num_examples)
df.to_csv(output_file, header=True, index=False)
return eval_results
def load_from_pretrained(args, pretrained_model_name_or_path):
"""Load the pretrained model and tokenizer."""
# In distributed training, the .from_pretrained methods guarantee that only
# one local process can concurrently perform this procedure.
config = AutoConfig.from_pretrained(
pretrained_model_name_or_path,
num_labels=args.num_labels if hasattr(args, "num_labels") else None,
finetuning_task=args.task_name.lower(),
cache_dir=args.cache_dir,
)
tokenizer = AutoTokenizer.from_pretrained(
pretrained_model_name_or_path, use_fast=args.use_fast_tokenizer, cache_dir=args.cache_dir
)
model = AutoModelForSequenceClassification.from_pretrained(
pretrained_model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
ignore_mismatched_sizes=True,
cache_dir=args.cache_dir,
)
return config, tokenizer, model
def finetune(accelerator, model_name_or_path, train_file, output_dir, **kwargs):
"""Fine-tuning a pre-trained model on a downstream task.
Args:
accelerator: An instance of an accelerator for distributed training (on
multi-GPU, TPU) or mixed precision training.
model_name_or_path: Path to pretrained model or model identifier from
huggingface.co/models.
train_file: A csv or a json file containing the training data.
output_dir: The output directory where the model predictions and checkpoints
will be written.
**kwargs: Dictionary of key/value pairs with which to update the
configuration object after loading. The values in kwargs of any keys which
are configuration attributes will be used to override the loaded values.
"""
# Make one log on every process with the configuration for debugging.
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
logger.info(accelerator.state)
# Setup logging, we only want one process per machine to log things on the
# screen. accelerator.is_local_main_process is only True for one process per
# machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
model_args = FTModelArguments(model_name_or_path=model_name_or_path)
data_args = FTDataArguments(train_file=train_file)
training_args = FTTrainingArguments(output_dir=output_dir)
args = argparse.Namespace()
for arg_class in (model_args, data_args, training_args):
for key, value in vars(arg_class).items():
setattr(args, key, value)
for key, value in kwargs.items():
if hasattr(args, key):
setattr(args, key, value)
# Sanity checks
data_files = {}
args.data_file_extension = None
# You need to provide the training data as we always run training
args.do_train = True
assert args.train_file is not None
data_files[Split.TRAIN.value] = args.train_file
if args.do_eval or args.evaluation_strategy != IntervalStrategy.NO.value:
assert args.eval_file is not None
data_files[Split.EVAL.value] = args.eval_file
if args.do_eval and args.test_file is not None:
data_files[Split.TEST.value] = args.test_file
if args.do_predict:
assert args.infer_file is not None
data_files[Split.INFER.value] = args.infer_file
for key in data_files:
extension = data_files[key].split(".")[-1]
assert extension in ["csv", "json"], f"`{key}_file` should be a csv or a json file."
if args.data_file_extension is None:
args.data_file_extension = extension
else:
assert extension == args.data_file_extension, f"`{key}_file` should be a {args.data_file_extension} file`."
assert (
args.eval_metric in datasets.list_metrics()
), f"{args.eval_metric} not in the list of supported metrics {datasets.list_metrics()}."
# Handle the output directory creation
if accelerator.is_main_process:
if args.output_dir is not None:
os.makedirs(args.output_dir, exist_ok=True)
accelerator.wait_for_everyone()
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed)
# You need to provide your CSV/JSON data files.
#
# For CSV/JSON files, this script will use as labels the column called 'label'
# and as pair of sentences the sentences in columns called 'sentence1' and
# 'sentence2' if these columns exist or the first two columns not named
# 'label' if at least two columns are provided.
#
# If the CSVs/JSONs contain only one non-label column, the script does single
# sentence classification on this single column.
#
# In distributed training, the load_dataset function guarantees that only one
# local process can download the dataset.
# Loading the dataset from local csv or json files.
raw_datasets = load_dataset(args.data_file_extension, data_files=data_files)
# Labels
is_regression = raw_datasets[Split.TRAIN.value].features["label"].dtype in ["float32", "float64"]
args.is_regression = is_regression
if args.is_regression:
label_list = None
num_labels = 1
else:
label_list = args.label_list
assert label_list is not None
label_list.sort() # Let's sort it for determinism
num_labels = len(label_list)
args.num_labels = num_labels
# Load pre-trained model
config, tokenizer, model = load_from_pretrained(args, args.model_name_or_path)
# Preprocessing the datasets
non_label_column_names = [name for name in raw_datasets[Split.TRAIN.value].column_names if name != "label"]
if "sentence1" in non_label_column_names and "sentence2" in non_label_column_names:
sentence1_key, sentence2_key = "sentence1", "sentence2"
else:
if len(non_label_column_names) >= 2:
sentence1_key, sentence2_key = non_label_column_names[:2]
else:
sentence1_key, sentence2_key = non_label_column_names[0], None
label_to_id = {v: i for i, v in enumerate(label_list)}
config.label2id = label_to_id
config.id2label = {id: label for label, id in config.label2id.items()}
padding = "max_length" if args.pad_to_max_length else False
def preprocess_function(examples):
# Tokenize the texts
texts = (
(examples[sentence1_key],) if sentence2_key is None else (examples[sentence1_key], examples[sentence2_key])
)
result = tokenizer(*texts, padding=padding, max_length=args.max_length, truncation=True)
if "label" in examples:
if label_to_id is not None:
# Map labels to IDs (not necessary for GLUE tasks)
result["labels"] = [label_to_id[l] for l in examples["label"]]
else:
# In all cases, rename the column to labels because the model will
# expect that.
result["labels"] = examples["label"]
return result
with accelerator.main_process_first():
processed_datasets = raw_datasets.map(
preprocess_function,
batched=True,
remove_columns=raw_datasets[Split.TRAIN.value].column_names,
desc="Running tokenizer on dataset",
)
num_examples = {}
splits = [s.value for s in Split]
for split in splits:
if split in processed_datasets:
num_examples[split] = len(processed_datasets[split])
args.num_examples = num_examples
train_dataset = processed_datasets[Split.TRAIN.value]
eval_dataset = processed_datasets[Split.EVAL.value] if Split.EVAL.value in processed_datasets else None
test_dataset = processed_datasets[Split.TEST.value] if Split.TEST.value in processed_datasets else None
infer_dataset = processed_datasets[Split.INFER.value] if Split.INFER.value in processed_datasets else None
# Log a few random samples from the training set:
for index in random.sample(range(len(train_dataset)), 3):
logger.info("Sample %d of the training set: %s.", index, train_dataset[index])
# DataLoaders creation:
if args.pad_to_max_length:
# If padding was already done ot max length, we use the default data
# collator that will just convert everything to tensors.
data_collator = default_data_collator
else:
# Otherwise, `DataCollatorWithPadding` will apply dynamic padding for us (by
# padding to the maximum length of the samples passed). When using mixed
# precision, we add `pad_to_multiple_of=8` to pad all tensors to multiple of
# 8s, which will enable the use of Tensor Cores on NVIDIA hardware with
# compute capability >= 7.5 (Volta).
data_collator = DataCollatorWithPadding(tokenizer, pad_to_multiple_of=(8 if accelerator.use_fp16 else None))
train_dataloader = DataLoader(
train_dataset,
batch_size=args.per_device_train_batch_size,
shuffle=True,
collate_fn=data_collator,
)
eval_dataloader, test_dataloader, infer_dataloader = None, None, None
if eval_dataset is not None:
eval_dataloader = DataLoader(
eval_dataset, batch_size=args.per_device_eval_batch_size, collate_fn=data_collator
)
if test_dataset is not None:
test_dataloader = DataLoader(
test_dataset, batch_size=args.per_device_eval_batch_size, collate_fn=data_collator
)
if infer_dataset is not None:
infer_dataloader = DataLoader(
infer_dataset, batch_size=args.per_device_eval_batch_size, collate_fn=data_collator
)
# Optimizer
# Split weights in two groups, one with weight decay and the other not.
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{
"params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{
"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)],
"weight_decay": 0.0,
},
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
# Prepare everything with our `accelerator`.
model, optimizer, train_dataloader, eval_dataloader, test_dataloader, infer_dataloader = accelerator.prepare(
model, optimizer, train_dataloader, eval_dataloader, test_dataloader, infer_dataloader
)
# Note -> the training dataloader needs to be prepared before we grab its
# length below (cause its length will be shorter in multiprocess)
# Scheduler and math around the number of training steps.
num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
if args.max_steps == -1:
args.max_steps = args.num_train_epochs * num_update_steps_per_epoch
else:
args.num_train_epochs = math.ceil(args.max_steps / num_update_steps_per_epoch)
lr_scheduler = get_scheduler(
name=args.lr_scheduler_type,
optimizer=optimizer,
num_warmup_steps=args.warmup_steps,
num_training_steps=args.max_steps,
)
# Train
completed_steps, avg_train_loss = train(
args, accelerator, model, tokenizer, train_dataloader, optimizer, lr_scheduler, eval_dataloader
)
accelerator.wait_for_everyone()
logger.info("Training job completed: completed_steps = %d, avg_train_loss = %f", completed_steps, avg_train_loss)
args.model_name_or_path = os.path.join(args.output_dir, "best-checkpoint")
logger.info("Loading the best checkpoint: %s", args.model_name_or_path)
config, tokenizer, model = load_from_pretrained(args, args.model_name_or_path)
model = accelerator.prepare(model)
if args.do_eval:
# Evaluate
if eval_dataloader is not None:
logger.info("***** Running evaluation on the eval data using the best checkpoint *****")
eval_results = evaluate(args, accelerator, eval_dataloader, Split.EVAL.value, model, "best-checkpoint")
avg_eval_loss = eval_results["avg_eval_loss"]
eval_metric = eval_results[args.eval_metric]
logger.info("Evaluation job completed: avg_eval_loss = %f", avg_eval_loss)
logger.info("Evaluation result for the best checkpoint: %s = %f", args.eval_metric, eval_metric)
if test_dataloader is not None:
logger.info("***** Running evaluation on the test data using the best checkpoint *****")
eval_results = evaluate(args, accelerator, test_dataloader, Split.TEST.value, model, "best-checkpoint")
avg_eval_loss = eval_results["avg_eval_loss"]
eval_metric = eval_results[args.eval_metric]
logger.info("Test job completed: avg_test_loss = %f", avg_eval_loss)
logger.info("Test result for the best checkpoint: %s = %f", args.eval_metric, eval_metric)
if args.do_predict:
# Predict
if infer_dataloader is not None:
logger.info("***** Running inference using the best checkpoint *****")
evaluate(
args, accelerator, infer_dataloader, Split.INFER.value, model, "best-checkpoint", has_labels=False
)
logger.info("Inference job completed.")
# Release all references to the internal objects stored and call the garbage
# collector. You should call this method between two trainings with different
# models/optimizers.
accelerator.free_memory()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/lm_seqs_dataset.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team and Facebook, Inc.
#
# 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.
""" Dataset to distilled models
adapted in part from Facebook, Inc XLM model (https://github.com/facebookresearch/XLM)
"""
import numpy as np
import torch
from torch.utils.data import Dataset
from utils import logger
class LmSeqsDataset(Dataset):
"""Custom Dataset wrapping language modeling sequences.
Each sample will be retrieved by indexing the list of token_ids and their corresponding lengths.
Input:
------
params: `NameSpace` parameters
data: `List[np.array[int]]
"""
def __init__(self, params, data):
self.params = params
self.token_ids = np.array(data)
self.lengths = np.array([len(t) for t in data])
self.check()
self.remove_long_sequences()
self.remove_empty_sequences()
self.remove_unknown_sequences()
self.check()
self.print_statistics()
def __getitem__(self, index):
return (self.token_ids[index], self.lengths[index])
def __len__(self):
return len(self.lengths)
def check(self):
"""
Some sanity checks
"""
assert len(self.token_ids) == len(self.lengths)
assert all(self.lengths[i] == len(self.token_ids[i]) for i in range(len(self.lengths)))
def remove_long_sequences(self):
"""
Sequences that are too long are split by chunk of max_model_input_size.
"""
max_len = self.params.max_model_input_size
indices = self.lengths > max_len
logger.info(f"Splitting {sum(indices)} too long sequences.")
def divide_chunks(l, n):
return [l[i : i + n] for i in range(0, len(l), n)]
new_tok_ids = []
new_lengths = []
if self.params.mlm:
cls_id, sep_id = self.params.special_tok_ids["cls_token"], self.params.special_tok_ids["sep_token"]
else:
cls_id, sep_id = self.params.special_tok_ids["bos_token"], self.params.special_tok_ids["eos_token"]
for seq_, len_ in zip(self.token_ids, self.lengths):
assert (seq_[0] == cls_id) and (seq_[-1] == sep_id), seq_
if len_ <= max_len:
new_tok_ids.append(seq_)
new_lengths.append(len_)
else:
sub_seqs = []
for sub_s in divide_chunks(seq_, max_len - 2):
if sub_s[0] != cls_id:
sub_s = np.insert(sub_s, 0, cls_id)
if sub_s[-1] != sep_id:
sub_s = np.insert(sub_s, len(sub_s), sep_id)
assert len(sub_s) <= max_len
assert (sub_s[0] == cls_id) and (sub_s[-1] == sep_id), sub_s
sub_seqs.append(sub_s)
new_tok_ids.extend(sub_seqs)
new_lengths.extend([len(l) for l in sub_seqs])
self.token_ids = np.array(new_tok_ids)
self.lengths = np.array(new_lengths)
def remove_empty_sequences(self):
"""
Too short sequences are simply removed. This could be tuned.
"""
init_size = len(self)
indices = self.lengths > 11
self.token_ids = self.token_ids[indices]
self.lengths = self.lengths[indices]
new_size = len(self)
logger.info(f"Remove {init_size - new_size} too short (<=11 tokens) sequences.")
def remove_unknown_sequences(self):
"""
Remove sequences with a (too) high level of unknown tokens.
"""
if "unk_token" not in self.params.special_tok_ids:
return
else:
unk_token_id = self.params.special_tok_ids["unk_token"]
init_size = len(self)
unk_occs = np.array([np.count_nonzero(a == unk_token_id) for a in self.token_ids])
indices = (unk_occs / self.lengths) < 0.5
self.token_ids = self.token_ids[indices]
self.lengths = self.lengths[indices]
new_size = len(self)
logger.info(f"Remove {init_size - new_size} sequences with a high level of unknown tokens (50%).")
def print_statistics(self):
"""
Print some statistics on the corpus. Only the master process.
"""
if not self.params.is_master:
return
logger.info(f"{len(self)} sequences")
# data_len = sum(self.lengths)
# nb_unique_tokens = len(Counter(list(chain(*self.token_ids))))
# logger.info(f'{data_len} tokens ({nb_unique_tokens} unique)')
# unk_idx = self.params.special_tok_ids['unk_token']
# nb_unknown = sum([(t==unk_idx).sum() for t in self.token_ids])
# logger.info(f'{nb_unknown} unknown tokens (covering {100*nb_unknown/data_len:.2f}% of the data)')
def batch_sequences(self, batch):
"""
Do the padding and transform into torch.tensor.
"""
token_ids = [t[0] for t in batch]
lengths = [t[1] for t in batch]
assert len(token_ids) == len(lengths)
# Max for paddings
max_seq_len_ = max(lengths)
# Pad token ids
if self.params.mlm:
pad_idx = self.params.special_tok_ids["pad_token"]
else:
pad_idx = self.params.special_tok_ids["unk_token"]
tk_ = [list(t.astype(int)) + [pad_idx] * (max_seq_len_ - len(t)) for t in token_ids]
assert len(tk_) == len(token_ids)
assert all(len(t) == max_seq_len_ for t in tk_)
tk_t = torch.tensor(tk_) # (bs, max_seq_len_)
lg_t = torch.tensor(lengths) # (bs)
return tk_t, lg_t
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/README.md
|
# Distil*
Author: @VictorSanh
This folder contains the original code used to train Distil* as well as examples showcasing how to use DistilBERT, DistilRoBERTa and DistilGPT2.
**January 20, 2020 - Bug fixing** We have recently discovered and fixed [a bug](https://github.com/huggingface/transformers/commit/48cbf267c988b56c71a2380f748a3e6092ccaed3) in the evaluation of our `run_*.py` scripts that caused the reported metrics to be over-estimated on average. We have updated all the metrics with the latest runs.
**December 6, 2019 - Update** We release **DistilmBERT**: 92% of `bert-base-multilingual-cased` on XNLI. The model supports 104 different languages listed [here](https://github.com/google-research/bert/blob/master/multilingual.md#list-of-languages).
**November 19, 2019 - Update** We release German **DistilBERT**: 98.8% of `bert-base-german-dbmdz-cased` on NER tasks.
**October 23, 2019 - Update** We release **DistilRoBERTa**: 95% of `RoBERTa-base`'s performance on GLUE, twice as fast as RoBERTa while being 35% smaller.
**October 3, 2019 - Update** We release our [NeurIPS workshop paper](https://arxiv.org/abs/1910.01108) explaining our approach on **DistilBERT**. It includes updated results and further experiments. We applied the same method to GPT2 and release the weights of **DistilGPT2**. DistilGPT2 is two times faster and 33% smaller than GPT2. **The paper supersedes our [previous blogpost](https://medium.com/huggingface/distilbert-8cf3380435b5) with a different distillation loss and better performances. Please use the paper as a reference when comparing/reporting results on DistilBERT.**
**September 19, 2019 - Update:** We fixed bugs in the code and released an updated version of the weights trained with a modification of the distillation loss. DistilBERT now reaches 99% of `BERT-base`'s performance on GLUE, and 86.9 F1 score on SQuAD v1.1 dev set (compared to 88.5 for `BERT-base`). We will publish a formal write-up of our approach in the near future!
## What is Distil*
Distil* is a class of compressed models that started with DistilBERT. DistilBERT stands for Distilled-BERT. DistilBERT is a small, fast, cheap and light Transformer model based on Bert architecture. It has 40% less parameters than `bert-base-uncased`, runs 60% faster while preserving 97% of BERT's performances as measured on the GLUE language understanding benchmark. DistilBERT is trained using knowledge distillation, a technique to compress a large model called the teacher into a smaller model called the student. By distillating Bert, we obtain a smaller Transformer model that bears a lot of similarities with the original BERT model while being lighter, smaller and faster to run. DistilBERT is thus an interesting option to put large-scaled trained Transformer model into production.
We have applied the same method to other Transformer architectures and released the weights:
- GPT2: on the [WikiText-103](https://blog.einstein.ai/the-wikitext-long-term-dependency-language-modeling-dataset/) benchmark, GPT2 reaches a perplexity on the test set of 16.3 compared to 21.1 for **DistilGPT2** (after fine-tuning on the train set).
- RoBERTa: **DistilRoBERTa** reaches 95% of `RoBERTa-base`'s performance on GLUE while being twice faster and 35% smaller.
- German BERT: **German DistilBERT** reaches 99% of `bert-base-german-dbmdz-cased`'s performance on German NER (CoNLL-2003).
- Multilingual BERT: **DistilmBERT** reaches 92% of Multilingual BERT's performance on XNLI while being twice faster and 25% smaller. The model supports 104 languages listed [here](https://github.com/google-research/bert/blob/master/multilingual.md#list-of-languages).
For more information on DistilBERT, please refer to our [NeurIPS workshop paper](https://arxiv.org/abs/1910.01108).
Here are the results on the dev sets of GLUE:
| Model | Macro-score | CoLA | MNLI | MRPC | QNLI | QQP | RTE | SST-2| STS-B| WNLI |
| :---: | :---: | :---:| :---:| :---:| :---:| :---:| :---:| :---:| :---:| :---: |
| BERT-base-uncased | **79.5** | 56.3 | 84.7 | 88.6 | 91.8 | 89.6 | 69.3 | 92.7 | 89.0 | 53.5 |
| DistilBERT-base-uncased | **77.0** | 51.3 | 82.1 | 87.5 | 89.2 | 88.5 | 59.9 | 91.3 | 86.9 | 56.3 |
| BERT-base-cased | **78.2** | 58.2 | 83.9 | 87.8 | 91.0 | 89.2 | 66.1 | 91.7 | 89.2 | 46.5 |
| DistilBERT-base-cased | **75.9** | 47.2 | 81.5 | 85.6 | 88.2 | 87.8 | 60.6 | 90.4 | 85.5 | 56.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| RoBERTa-base (reported) | **83.2**/**86.4**<sup>2</sup> | 63.6 | 87.6 | 90.2 | 92.8 | 91.9 | 78.7 | 94.8 | 91.2 | 57.7<sup>3</sup> |
| DistilRoBERTa<sup>1</sup> | **79.0**/**82.3**<sup>2</sup> | 59.3 | 84.0 | 86.6 | 90.8 | 89.4 | 67.9 | 92.5 | 88.3 | 52.1 |
<sup>1</sup> We did not use the MNLI checkpoint for fine-tuning but directly perform transfer learning on the pre-trained DistilRoBERTa.
<sup>2</sup> Macro-score computed without WNLI.
<sup>3</sup> We compute this score ourselves for completeness.
Here are the results on the *test* sets for 6 of the languages available in XNLI. The results are computed in the zero shot setting (trained on the English portion and evaluated on the target language portion):
| Model | English | Spanish | Chinese | German | Arabic | Urdu |
| :---: | :---: | :---: | :---: | :---: | :---: | :---:|
| mBERT base cased (computed) | 82.1 | 74.6 | 69.1 | 72.3 | 66.4 | 58.5 |
| mBERT base uncased (reported)| 81.4 | 74.3 | 63.8 | 70.5 | 62.1 | 58.3 |
| DistilmBERT | 78.2 | 69.1 | 64.0 | 66.3 | 59.1 | 54.7 |
## Setup
This part of the library has only be tested with Python3.6+. There are few specific dependencies to install before launching a distillation, you can install them with the command `pip install -r requirements.txt`.
**Important note:** The training scripts have been updated to support PyTorch v1.2.0 (there are breaking changes compared to v1.1.0).
## How to use DistilBERT
Transformers includes five pre-trained Distil* models, currently only provided for English and German (we are investigating the possibility to train and release a multilingual version of DistilBERT):
- `distilbert-base-uncased`: DistilBERT English language model pretrained on the same data used to pretrain Bert (concatenation of the Toronto Book Corpus and full English Wikipedia) using distillation with the supervision of the `bert-base-uncased` version of Bert. The model has 6 layers, 768 dimension and 12 heads, totalizing 66M parameters.
- `distilbert-base-uncased-distilled-squad`: A finetuned version of `distilbert-base-uncased` finetuned using (a second step of) knowledge distillation on SQuAD 1.0. This model reaches a F1 score of 86.9 on the dev set (for comparison, Bert `bert-base-uncased` version reaches a 88.5 F1 score).
- `distilbert-base-cased`: DistilBERT English language model pretrained on the same data used to pretrain Bert (concatenation of the Toronto Book Corpus and full English Wikipedia) using distillation with the supervision of the `bert-base-cased` version of Bert. The model has 6 layers, 768 dimension and 12 heads, totalizing 65M parameters.
- `distilbert-base-cased-distilled-squad`: A finetuned version of `distilbert-base-cased` finetuned using (a second step of) knowledge distillation on SQuAD 1.0. This model reaches a F1 score of 87.1 on the dev set (for comparison, Bert `bert-base-cased` version reaches a 88.7 F1 score).
- `distilbert-base-german-cased`: DistilBERT German language model pretrained on 1/2 of the data used to pretrain Bert using distillation with the supervision of the `bert-base-german-dbmdz-cased` version of German DBMDZ Bert. For NER tasks the model reaches a F1 score of 83.49 on the CoNLL-2003 test set (for comparison, `bert-base-german-dbmdz-cased` reaches a 84.52 F1 score), and a F1 score of 85.23 on the GermEval 2014 test set (`bert-base-german-dbmdz-cased` reaches a 86.89 F1 score).
- `distilgpt2`: DistilGPT2 English language model pretrained with the supervision of `gpt2` (the smallest version of GPT2) on [OpenWebTextCorpus](https://skylion007.github.io/OpenWebTextCorpus/), a reproduction of OpenAI's WebText dataset. The model has 6 layers, 768 dimension and 12 heads, totalizing 82M parameters (compared to 124M parameters for GPT2). On average, DistilGPT2 is two times faster than GPT2.
- `distilroberta-base`: DistilRoBERTa English language model pretrained with the supervision of `roberta-base` solely on [OpenWebTextCorpus](https://skylion007.github.io/OpenWebTextCorpus/), a reproduction of OpenAI's WebText dataset (it is ~4 times less training data than the teacher RoBERTa). The model has 6 layers, 768 dimension and 12 heads, totalizing 82M parameters (compared to 125M parameters for RoBERTa-base). On average DistilRoBERTa is twice as fast as Roberta-base.
- `distilbert-base-multilingual-cased`: DistilmBERT multilingual model pretrained with the supervision of `bert-base-multilingual-cased` on the concatenation of Wikipedia in 104 different languages. The model supports the 104 languages listed [here](https://github.com/google-research/bert/blob/master/multilingual.md#list-of-languages). The model has 6 layers, 768 dimension and 12 heads, totalizing 134M parameters (compared to 177M parameters for mBERT-base). On average DistilmBERT is twice as fast as mBERT-base.
Using DistilBERT is very similar to using BERT. DistilBERT share the same tokenizer as BERT's `bert-base-uncased` even though we provide a link to this tokenizer under the `DistilBertTokenizer` name to have a consistent naming between the library models.
```python
tokenizer = DistilBertTokenizer.from_pretrained('distilbert-base-cased')
model = DistilBertModel.from_pretrained('distilbert-base-cased')
input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)
outputs = model(input_ids)
last_hidden_states = outputs[0] # The last hidden-state is the first element of the output tuple
```
Similarly, using the other Distil* models simply consists in calling the base classes with a different pretrained checkpoint:
- DistilBERT uncased: `model = DistilBertModel.from_pretrained('distilbert-base-uncased')`
- DistilGPT2: `model = GPT2Model.from_pretrained('distilgpt2')`
- DistilRoBERTa: `model = RobertaModel.from_pretrained('distilroberta-base')`
- DistilmBERT: `model = DistilBertModel.from_pretrained('distilbert-base-multilingual-cased')`
## How to train Distil*
In the following, we will explain how you can train DistilBERT.
### A. Preparing the data
The weights we release are trained using a concatenation of Toronto Book Corpus and English Wikipedia (same training data as the English version of BERT).
To avoid processing the data several time, we do it once and for all before the training. From now on, will suppose that you have a text file `dump.txt` which contains one sequence per line (a sequence being composed of one of several coherent sentences).
First, we will binarize the data, i.e. tokenize the data and convert each token in an index in our model's vocabulary.
```bash
python scripts/binarized_data.py \
--file_path data/dump.txt \
--tokenizer_type bert \
--tokenizer_name bert-base-uncased \
--dump_file data/binarized_text
```
Our implementation of masked language modeling loss follows [XLM](https://github.com/facebookresearch/XLM)'s one and smooths the probability of masking with a factor that put more emphasis on rare words. Thus we count the occurrences of each tokens in the data:
```bash
python scripts/token_counts.py \
--data_file data/binarized_text.bert-base-uncased.pickle \
--token_counts_dump data/token_counts.bert-base-uncased.pickle \
--vocab_size 30522
```
### B. Training
Training with distillation is really simple once you have pre-processed the data:
```bash
python train.py \
--student_type distilbert \
--student_config training_configs/distilbert-base-uncased.json \
--teacher_type bert \
--teacher_name bert-base-uncased \
--alpha_ce 5.0 --alpha_mlm 2.0 --alpha_cos 1.0 --alpha_clm 0.0 --mlm \
--freeze_pos_embs \
--dump_path serialization_dir/my_first_training \
--data_file data/binarized_text.bert-base-uncased.pickle \
--token_counts data/token_counts.bert-base-uncased.pickle \
--force # overwrites the `dump_path` if it already exists.
```
By default, this will launch a training on a single GPU (even if more are available on the cluster). Other parameters are available in the command line, please look in `train.py` or run `python train.py --help` to list them.
We highly encourage you to use distributed training for training DistilBERT as the training corpus is quite large. Here's an example that runs a distributed training on a single node having 4 GPUs:
```bash
export NODE_RANK=0
export N_NODES=1
export N_GPU_NODE=4
export WORLD_SIZE=4
export MASTER_PORT=<AN_OPEN_PORT>
export MASTER_ADDR=<I.P.>
pkill -f 'python -u train.py'
python -m torch.distributed.launch \
--nproc_per_node=$N_GPU_NODE \
--nnodes=$N_NODES \
--node_rank $NODE_RANK \
--master_addr $MASTER_ADDR \
--master_port $MASTER_PORT \
train.py \
--force \
--n_gpu $WORLD_SIZE \
--student_type distilbert \
--student_config training_configs/distilbert-base-uncased.json \
--teacher_type bert \
--teacher_name bert-base-uncased \
--alpha_ce 0.33 --alpha_mlm 0.33 --alpha_cos 0.33 --alpha_clm 0.0 --mlm \
--freeze_pos_embs \
--dump_path serialization_dir/my_first_training \
--data_file data/binarized_text.bert-base-uncased.pickle \
--token_counts data/token_counts.bert-base-uncased.pickle
```
**Tips:** Starting distilled training with good initialization of the model weights is crucial to reach decent performance. In our experiments, we initialized our model from a few layers of the teacher (Bert) itself! Please refer to `scripts/extract.py` and `scripts/extract_distilbert.py` to create a valid initialization checkpoint and use `--student_pretrained_weights` argument to use this initialization for the distilled training!
Happy distillation!
## Citation
If you find the resource useful, you should cite the following paper:
```
@inproceedings{sanh2019distilbert,
title={DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter},
author={Sanh, Victor and Debut, Lysandre and Chaumond, Julien and Wolf, Thomas},
booktitle={NeurIPS EMC^2 Workshop},
year={2019}
}
```
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/distiller.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team and Facebook, Inc.
#
# 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.
""" The distiller to distil the student.
Adapted in part from Facebook, Inc XLM model (https://github.com/facebookresearch/XLM)
"""
import math
import os
import time
import psutil
import torch
from grouped_batch_sampler import GroupedBatchSampler, create_lengths_groups
from lm_seqs_dataset import LmSeqsDataset
from torch import nn
from torch.optim import AdamW
from torch.utils.data import BatchSampler, DataLoader, RandomSampler
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm
from transformers import get_linear_schedule_with_warmup
from utils import logger
try:
from torch.utils.tensorboard import SummaryWriter
except ImportError:
from tensorboardX import SummaryWriter
class Distiller:
def __init__(
self, params: dict, dataset: LmSeqsDataset, token_probs: torch.tensor, student: nn.Module, teacher: nn.Module
):
logger.info("Initializing Distiller")
self.params = params
self.dump_path = params.dump_path
self.multi_gpu = params.multi_gpu
self.fp16 = params.fp16
self.student = student
self.teacher = teacher
self.student_config = student.config
self.vocab_size = student.config.vocab_size
if params.n_gpu <= 1:
sampler = RandomSampler(dataset)
else:
sampler = DistributedSampler(dataset)
if params.group_by_size:
groups = create_lengths_groups(lengths=dataset.lengths, k=params.max_model_input_size)
sampler = GroupedBatchSampler(sampler=sampler, group_ids=groups, batch_size=params.batch_size)
else:
sampler = BatchSampler(sampler=sampler, batch_size=params.batch_size, drop_last=False)
self.dataloader = DataLoader(dataset=dataset, batch_sampler=sampler, collate_fn=dataset.batch_sequences)
self.temperature = params.temperature
assert self.temperature > 0.0
self.alpha_ce = params.alpha_ce
self.alpha_mlm = params.alpha_mlm
self.alpha_clm = params.alpha_clm
self.alpha_mse = params.alpha_mse
self.alpha_cos = params.alpha_cos
self.mlm = params.mlm
if self.mlm:
logger.info("Using MLM loss for LM step.")
self.mlm_mask_prop = params.mlm_mask_prop
assert 0.0 <= self.mlm_mask_prop <= 1.0
assert params.word_mask + params.word_keep + params.word_rand == 1.0
self.pred_probs = torch.FloatTensor([params.word_mask, params.word_keep, params.word_rand])
self.pred_probs = self.pred_probs.to(f"cuda:{params.local_rank}") if params.n_gpu > 0 else self.pred_probs
self.token_probs = token_probs.to(f"cuda:{params.local_rank}") if params.n_gpu > 0 else token_probs
if self.fp16:
self.pred_probs = self.pred_probs.half()
self.token_probs = self.token_probs.half()
else:
logger.info("Using CLM loss for LM step.")
self.epoch = 0
self.n_iter = 0
self.n_total_iter = 0
self.n_sequences_epoch = 0
self.total_loss_epoch = 0
self.last_loss = 0
self.last_loss_ce = 0
self.last_loss_mlm = 0
self.last_loss_clm = 0
if self.alpha_mse > 0.0:
self.last_loss_mse = 0
if self.alpha_cos > 0.0:
self.last_loss_cos = 0
self.last_log = 0
self.ce_loss_fct = nn.KLDivLoss(reduction="batchmean")
self.lm_loss_fct = nn.CrossEntropyLoss(ignore_index=-100)
if self.alpha_mse > 0.0:
self.mse_loss_fct = nn.MSELoss(reduction="sum")
if self.alpha_cos > 0.0:
self.cosine_loss_fct = nn.CosineEmbeddingLoss(reduction="mean")
logger.info("--- Initializing model optimizer")
assert params.gradient_accumulation_steps >= 1
self.num_steps_epoch = len(self.dataloader)
num_train_optimization_steps = (
int(self.num_steps_epoch / params.gradient_accumulation_steps * params.n_epoch) + 1
)
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{
"params": [
p for n, p in student.named_parameters() if not any(nd in n for nd in no_decay) and p.requires_grad
],
"weight_decay": params.weight_decay,
},
{
"params": [
p for n, p in student.named_parameters() if any(nd in n for nd in no_decay) and p.requires_grad
],
"weight_decay": 0.0,
},
]
logger.info(
"------ Number of trainable parameters (student): %i"
% sum([p.numel() for p in self.student.parameters() if p.requires_grad])
)
logger.info("------ Number of parameters (student): %i" % sum([p.numel() for p in self.student.parameters()]))
self.optimizer = AdamW(
optimizer_grouped_parameters, lr=params.learning_rate, eps=params.adam_epsilon, betas=(0.9, 0.98)
)
warmup_steps = math.ceil(num_train_optimization_steps * params.warmup_prop)
self.scheduler = get_linear_schedule_with_warmup(
self.optimizer, num_warmup_steps=warmup_steps, num_training_steps=num_train_optimization_steps
)
if self.fp16:
try:
from apex import amp
except ImportError:
raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
logger.info(f"Using fp16 training: {self.params.fp16_opt_level} level")
self.student, self.optimizer = amp.initialize(
self.student, self.optimizer, opt_level=self.params.fp16_opt_level
)
self.teacher = self.teacher.half()
if self.multi_gpu:
if self.fp16:
from apex.parallel import DistributedDataParallel
logger.info("Using apex.parallel.DistributedDataParallel for distributed training.")
self.student = DistributedDataParallel(self.student)
else:
from torch.nn.parallel import DistributedDataParallel
logger.info("Using nn.parallel.DistributedDataParallel for distributed training.")
self.student = DistributedDataParallel(
self.student,
device_ids=[params.local_rank],
output_device=params.local_rank,
find_unused_parameters=True,
)
self.is_master = params.is_master
if self.is_master:
logger.info("--- Initializing Tensorboard")
self.tensorboard = SummaryWriter(log_dir=os.path.join(self.dump_path, "log", "train"))
self.tensorboard.add_text(tag="config/training", text_string=str(self.params), global_step=0)
self.tensorboard.add_text(tag="config/student", text_string=str(self.student_config), global_step=0)
def prepare_batch_mlm(self, batch):
"""
Prepare the batch: from the token_ids and the lengths, compute the attention mask and the masked label for MLM.
Input:
------
batch: `Tuple`
token_ids: `torch.tensor(bs, seq_length)` - The token ids for each of the sequence. It is padded.
lengths: `torch.tensor(bs)` - The lengths of each of the sequences in the batch.
Output:
-------
token_ids: `torch.tensor(bs, seq_length)` - The token ids after the modifications for MLM.
attn_mask: `torch.tensor(bs, seq_length)` - The attention mask for the self-attention.
mlm_labels: `torch.tensor(bs, seq_length)` - The masked language modeling labels. There is a -100 where there is nothing to predict.
"""
token_ids, lengths = batch
token_ids, lengths = self.round_batch(x=token_ids, lengths=lengths)
assert token_ids.size(0) == lengths.size(0)
attn_mask = torch.arange(token_ids.size(1), dtype=torch.long, device=lengths.device) < lengths[:, None]
bs, max_seq_len = token_ids.size()
mlm_labels = token_ids.new(token_ids.size()).copy_(token_ids)
x_prob = self.token_probs[token_ids.flatten()]
n_tgt = math.ceil(self.mlm_mask_prop * lengths.sum().item())
tgt_ids = torch.multinomial(x_prob / x_prob.sum(), n_tgt, replacement=False)
pred_mask = torch.zeros(
bs * max_seq_len, dtype=torch.bool, device=token_ids.device
) # previously `dtype=torch.uint8`, cf pytorch 1.2.0 compatibility
pred_mask[tgt_ids] = 1
pred_mask = pred_mask.view(bs, max_seq_len)
pred_mask[token_ids == self.params.special_tok_ids["pad_token"]] = 0
# mask a number of words == 0 [8] (faster with fp16)
if self.fp16:
n1 = pred_mask.sum().item()
if n1 > 8:
pred_mask = pred_mask.view(-1)
n2 = max(n1 % 8, 8 * (n1 // 8))
if n2 != n1:
pred_mask[torch.nonzero(pred_mask).view(-1)[: n1 - n2]] = 0
pred_mask = pred_mask.view(bs, max_seq_len)
assert pred_mask.sum().item() % 8 == 0, pred_mask.sum().item()
_token_ids_real = token_ids[pred_mask]
_token_ids_rand = _token_ids_real.clone().random_(self.vocab_size)
_token_ids_mask = _token_ids_real.clone().fill_(self.params.special_tok_ids["mask_token"])
probs = torch.multinomial(self.pred_probs, len(_token_ids_real), replacement=True)
_token_ids = (
_token_ids_mask * (probs == 0).long()
+ _token_ids_real * (probs == 1).long()
+ _token_ids_rand * (probs == 2).long()
)
token_ids = token_ids.masked_scatter(pred_mask, _token_ids)
mlm_labels[~pred_mask] = -100 # previously `mlm_labels[1-pred_mask] = -1`, cf pytorch 1.2.0 compatibility
# sanity checks
assert 0 <= token_ids.min() <= token_ids.max() < self.vocab_size
return token_ids, attn_mask, mlm_labels
def prepare_batch_clm(self, batch):
"""
Prepare the batch: from the token_ids and the lengths, compute the attention mask and the labels for CLM.
Input:
------
batch: `Tuple`
token_ids: `torch.tensor(bs, seq_length)` - The token ids for each of the sequence. It is padded.
lengths: `torch.tensor(bs)` - The lengths of each of the sequences in the batch.
Output:
-------
token_ids: `torch.tensor(bs, seq_length)` - The token ids after the modifications for MLM.
attn_mask: `torch.tensor(bs, seq_length)` - The attention mask for the self-attention.
clm_labels: `torch.tensor(bs, seq_length)` - The causal language modeling labels. There is a -100 where there is nothing to predict.
"""
token_ids, lengths = batch
token_ids, lengths = self.round_batch(x=token_ids, lengths=lengths)
assert token_ids.size(0) == lengths.size(0)
attn_mask = torch.arange(token_ids.size(1), dtype=torch.long, device=lengths.device) < lengths[:, None]
clm_labels = token_ids.new(token_ids.size()).copy_(token_ids)
clm_labels[~attn_mask] = -100 # previously `clm_labels[1-attn_mask] = -1`, cf pytorch 1.2.0 compatibility
# sanity checks
assert 0 <= token_ids.min() <= token_ids.max() < self.vocab_size
return token_ids, attn_mask, clm_labels
def round_batch(self, x: torch.tensor, lengths: torch.tensor):
"""
For float16 only.
Sub-sample sentences in a batch, and add padding, so that each dimension is a multiple of 8.
Input:
------
x: `torch.tensor(bs, seq_length)` - The token ids.
lengths: `torch.tensor(bs, seq_length)` - The lengths of each of the sequence in the batch.
Output:
-------
x: `torch.tensor(new_bs, new_seq_length)` - The updated token ids.
lengths: `torch.tensor(new_bs, new_seq_length)` - The updated lengths.
"""
if not self.fp16 or len(lengths) < 8:
return x, lengths
# number of sentences == 0 [8]
bs1 = len(lengths)
bs2 = 8 * (bs1 // 8)
assert bs2 > 0 and bs2 % 8 == 0
if bs1 != bs2:
idx = torch.randperm(bs1)[:bs2]
lengths = lengths[idx]
slen = lengths.max().item()
x = x[idx, :slen]
else:
idx = None
# sequence length == 0 [8]
ml1 = x.size(1)
if ml1 % 8 != 0:
pad = 8 - (ml1 % 8)
ml2 = ml1 + pad
if self.mlm:
pad_id = self.params.special_tok_ids["pad_token"]
else:
pad_id = self.params.special_tok_ids["unk_token"]
padding_tensor = torch.zeros(bs2, pad, dtype=torch.long, device=x.device).fill_(pad_id)
x = torch.cat([x, padding_tensor], 1)
assert x.size() == (bs2, ml2)
assert x.size(0) % 8 == 0
assert x.size(1) % 8 == 0
return x, lengths
def train(self):
"""
The real training loop.
"""
if self.is_master:
logger.info("Starting training")
self.last_log = time.time()
self.student.train()
self.teacher.eval()
for _ in range(self.params.n_epoch):
if self.is_master:
logger.info(f"--- Starting epoch {self.epoch}/{self.params.n_epoch-1}")
if self.multi_gpu:
torch.distributed.barrier()
iter_bar = tqdm(self.dataloader, desc="-Iter", disable=self.params.local_rank not in [-1, 0])
for batch in iter_bar:
if self.params.n_gpu > 0:
batch = tuple(t.to(f"cuda:{self.params.local_rank}") for t in batch)
if self.mlm:
token_ids, attn_mask, lm_labels = self.prepare_batch_mlm(batch=batch)
else:
token_ids, attn_mask, lm_labels = self.prepare_batch_clm(batch=batch)
self.step(input_ids=token_ids, attention_mask=attn_mask, lm_labels=lm_labels)
iter_bar.update()
iter_bar.set_postfix(
{"Last_loss": f"{self.last_loss:.2f}", "Avg_cum_loss": f"{self.total_loss_epoch/self.n_iter:.2f}"}
)
iter_bar.close()
if self.is_master:
logger.info(f"--- Ending epoch {self.epoch}/{self.params.n_epoch-1}")
self.end_epoch()
if self.is_master:
logger.info("Save very last checkpoint as `pytorch_model.bin`.")
self.save_checkpoint(checkpoint_name="pytorch_model.bin")
logger.info("Training is finished")
def step(self, input_ids: torch.tensor, attention_mask: torch.tensor, lm_labels: torch.tensor):
"""
One optimization step: forward of student AND teacher, backward on the loss (for gradient accumulation),
and possibly a parameter update (depending on the gradient accumulation).
Input:
------
input_ids: `torch.tensor(bs, seq_length)` - The token ids.
attention_mask: `torch.tensor(bs, seq_length)` - The attention mask for self attention.
lm_labels: `torch.tensor(bs, seq_length)` - The language modeling labels (mlm labels for MLM and clm labels for CLM).
"""
if self.mlm:
student_outputs = self.student(
input_ids=input_ids, attention_mask=attention_mask
) # (bs, seq_length, voc_size)
with torch.no_grad():
teacher_outputs = self.teacher(
input_ids=input_ids, attention_mask=attention_mask
) # (bs, seq_length, voc_size)
else:
student_outputs = self.student(input_ids=input_ids, attention_mask=None) # (bs, seq_length, voc_size)
with torch.no_grad():
teacher_outputs = self.teacher(input_ids=input_ids, attention_mask=None) # (bs, seq_length, voc_size)
s_logits, s_hidden_states = student_outputs["logits"], student_outputs["hidden_states"]
t_logits, t_hidden_states = teacher_outputs["logits"], teacher_outputs["hidden_states"]
assert s_logits.size() == t_logits.size()
# https://github.com/peterliht/knowledge-distillation-pytorch/blob/master/model/net.py#L100
# https://github.com/peterliht/knowledge-distillation-pytorch/issues/2
if self.params.restrict_ce_to_mask:
mask = (lm_labels > -1).unsqueeze(-1).expand_as(s_logits) # (bs, seq_length, voc_size)
else:
mask = attention_mask.unsqueeze(-1).expand_as(s_logits) # (bs, seq_length, voc_size)
s_logits_slct = torch.masked_select(s_logits, mask) # (bs * seq_length * voc_size) modulo the 1s in mask
s_logits_slct = s_logits_slct.view(-1, s_logits.size(-1)) # (bs * seq_length, voc_size) modulo the 1s in mask
t_logits_slct = torch.masked_select(t_logits, mask) # (bs * seq_length * voc_size) modulo the 1s in mask
t_logits_slct = t_logits_slct.view(-1, s_logits.size(-1)) # (bs * seq_length, voc_size) modulo the 1s in mask
assert t_logits_slct.size() == s_logits_slct.size()
loss_ce = (
self.ce_loss_fct(
nn.functional.log_softmax(s_logits_slct / self.temperature, dim=-1),
nn.functional.softmax(t_logits_slct / self.temperature, dim=-1),
)
* (self.temperature) ** 2
)
loss = self.alpha_ce * loss_ce
if self.alpha_mlm > 0.0:
loss_mlm = self.lm_loss_fct(s_logits.view(-1, s_logits.size(-1)), lm_labels.view(-1))
loss += self.alpha_mlm * loss_mlm
if self.alpha_clm > 0.0:
shift_logits = s_logits[..., :-1, :].contiguous()
shift_labels = lm_labels[..., 1:].contiguous()
loss_clm = self.lm_loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))
loss += self.alpha_clm * loss_clm
if self.alpha_mse > 0.0:
loss_mse = self.mse_loss_fct(s_logits_slct, t_logits_slct) / s_logits_slct.size(
0
) # Reproducing batchmean reduction
loss += self.alpha_mse * loss_mse
if self.alpha_cos > 0.0:
s_hidden_states = s_hidden_states[-1] # (bs, seq_length, dim)
t_hidden_states = t_hidden_states[-1] # (bs, seq_length, dim)
mask = attention_mask.unsqueeze(-1).expand_as(s_hidden_states) # (bs, seq_length, dim)
assert s_hidden_states.size() == t_hidden_states.size()
dim = s_hidden_states.size(-1)
s_hidden_states_slct = torch.masked_select(s_hidden_states, mask) # (bs * seq_length * dim)
s_hidden_states_slct = s_hidden_states_slct.view(-1, dim) # (bs * seq_length, dim)
t_hidden_states_slct = torch.masked_select(t_hidden_states, mask) # (bs * seq_length * dim)
t_hidden_states_slct = t_hidden_states_slct.view(-1, dim) # (bs * seq_length, dim)
target = s_hidden_states_slct.new(s_hidden_states_slct.size(0)).fill_(1) # (bs * seq_length,)
loss_cos = self.cosine_loss_fct(s_hidden_states_slct, t_hidden_states_slct, target)
loss += self.alpha_cos * loss_cos
self.total_loss_epoch += loss.item()
self.last_loss = loss.item()
self.last_loss_ce = loss_ce.item()
if self.alpha_mlm > 0.0:
self.last_loss_mlm = loss_mlm.item()
if self.alpha_clm > 0.0:
self.last_loss_clm = loss_clm.item()
if self.alpha_mse > 0.0:
self.last_loss_mse = loss_mse.item()
if self.alpha_cos > 0.0:
self.last_loss_cos = loss_cos.item()
self.optimize(loss)
self.n_sequences_epoch += input_ids.size(0)
def optimize(self, loss):
"""
Normalization on the loss (gradient accumulation or distributed training), followed by
backward pass on the loss, possibly followed by a parameter update (depending on the gradient accumulation).
Also update the metrics for tensorboard.
"""
# Check for NaN
if (loss != loss).data.any():
logger.error("NaN detected")
exit()
if self.multi_gpu:
loss = loss.mean()
if self.params.gradient_accumulation_steps > 1:
loss = loss / self.params.gradient_accumulation_steps
if self.fp16:
from apex import amp
with amp.scale_loss(loss, self.optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
self.iter()
if self.n_iter % self.params.gradient_accumulation_steps == 0:
if self.fp16:
nn.utils.clip_grad_norm_(amp.master_params(self.optimizer), self.params.max_grad_norm)
else:
nn.utils.clip_grad_norm_(self.student.parameters(), self.params.max_grad_norm)
self.optimizer.step()
self.optimizer.zero_grad()
self.scheduler.step()
def iter(self):
"""
Update global counts, write to tensorboard and save checkpoint.
"""
self.n_iter += 1
self.n_total_iter += 1
if self.n_total_iter % self.params.log_interval == 0:
self.log_tensorboard()
self.last_log = time.time()
if self.n_total_iter % self.params.checkpoint_interval == 0:
self.save_checkpoint()
def log_tensorboard(self):
"""
Log into tensorboard. Only by the master process.
"""
if not self.is_master:
return
for param_name, param in self.student.named_parameters():
self.tensorboard.add_scalar(
tag="parameter_mean/" + param_name, scalar_value=param.data.mean(), global_step=self.n_total_iter
)
self.tensorboard.add_scalar(
tag="parameter_std/" + param_name, scalar_value=param.data.std(), global_step=self.n_total_iter
)
if param.grad is None:
continue
self.tensorboard.add_scalar(
tag="grad_mean/" + param_name, scalar_value=param.grad.data.mean(), global_step=self.n_total_iter
)
self.tensorboard.add_scalar(
tag="grad_std/" + param_name, scalar_value=param.grad.data.std(), global_step=self.n_total_iter
)
self.tensorboard.add_scalar(
tag="losses/cum_avg_loss_epoch",
scalar_value=self.total_loss_epoch / self.n_iter,
global_step=self.n_total_iter,
)
self.tensorboard.add_scalar(tag="losses/loss", scalar_value=self.last_loss, global_step=self.n_total_iter)
self.tensorboard.add_scalar(
tag="losses/loss_ce", scalar_value=self.last_loss_ce, global_step=self.n_total_iter
)
if self.alpha_mlm > 0.0:
self.tensorboard.add_scalar(
tag="losses/loss_mlm", scalar_value=self.last_loss_mlm, global_step=self.n_total_iter
)
if self.alpha_clm > 0.0:
self.tensorboard.add_scalar(
tag="losses/loss_clm", scalar_value=self.last_loss_clm, global_step=self.n_total_iter
)
if self.alpha_mse > 0.0:
self.tensorboard.add_scalar(
tag="losses/loss_mse", scalar_value=self.last_loss_mse, global_step=self.n_total_iter
)
if self.alpha_cos > 0.0:
self.tensorboard.add_scalar(
tag="losses/loss_cos", scalar_value=self.last_loss_cos, global_step=self.n_total_iter
)
self.tensorboard.add_scalar(
tag="learning_rate/lr", scalar_value=self.scheduler.get_lr()[0], global_step=self.n_total_iter
)
self.tensorboard.add_scalar(
tag="global/memory_usage",
scalar_value=psutil.virtual_memory()._asdict()["used"] / 1_000_000,
global_step=self.n_total_iter,
)
self.tensorboard.add_scalar(
tag="global/speed", scalar_value=time.time() - self.last_log, global_step=self.n_total_iter
)
def end_epoch(self):
"""
Finally arrived at the end of epoch (full pass on dataset).
Do some tensorboard logging and checkpoint saving.
"""
logger.info(f"{self.n_sequences_epoch} sequences have been trained during this epoch.")
if self.is_master:
self.save_checkpoint(checkpoint_name=f"model_epoch_{self.epoch}.pth")
self.tensorboard.add_scalar(
tag="epoch/loss", scalar_value=self.total_loss_epoch / self.n_iter, global_step=self.epoch
)
self.epoch += 1
self.n_sequences_epoch = 0
self.n_iter = 0
self.total_loss_epoch = 0
def save_checkpoint(self, checkpoint_name: str = "checkpoint.pth"):
"""
Save the current state. Only by the master process.
"""
if not self.is_master:
return
mdl_to_save = self.student.module if hasattr(self.student, "module") else self.student
mdl_to_save.config.save_pretrained(self.dump_path)
state_dict = mdl_to_save.state_dict()
torch.save(state_dict, os.path.join(self.dump_path, checkpoint_name))
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/utils.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team and Facebook, Inc.
#
# 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.
""" Utils to train DistilBERT
adapted in part from Facebook, Inc XLM model (https://github.com/facebookresearch/XLM)
"""
import json
import logging
import os
import socket
import git
import numpy as np
import torch
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - PID: %(process)d - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
logger = logging.getLogger(__name__)
def git_log(folder_path: str):
"""
Log commit info.
"""
repo = git.Repo(search_parent_directories=True)
repo_infos = {
"repo_id": str(repo),
"repo_sha": str(repo.head.object.hexsha),
"repo_branch": str(repo.active_branch),
}
with open(os.path.join(folder_path, "git_log.json"), "w") as f:
json.dump(repo_infos, f, indent=4)
def init_gpu_params(params):
"""
Handle single and multi-GPU / multi-node.
"""
if params.n_gpu <= 0:
params.local_rank = 0
params.master_port = -1
params.is_master = True
params.multi_gpu = False
return
assert torch.cuda.is_available()
logger.info("Initializing GPUs")
if params.n_gpu > 1:
assert params.local_rank != -1
params.world_size = int(os.environ["WORLD_SIZE"])
params.n_gpu_per_node = int(os.environ["N_GPU_NODE"])
params.global_rank = int(os.environ["RANK"])
# number of nodes / node ID
params.n_nodes = params.world_size // params.n_gpu_per_node
params.node_id = params.global_rank // params.n_gpu_per_node
params.multi_gpu = True
assert params.n_nodes == int(os.environ["N_NODES"])
assert params.node_id == int(os.environ["NODE_RANK"])
# local job (single GPU)
else:
assert params.local_rank == -1
params.n_nodes = 1
params.node_id = 0
params.local_rank = 0
params.global_rank = 0
params.world_size = 1
params.n_gpu_per_node = 1
params.multi_gpu = False
# sanity checks
assert params.n_nodes >= 1
assert 0 <= params.node_id < params.n_nodes
assert 0 <= params.local_rank <= params.global_rank < params.world_size
assert params.world_size == params.n_nodes * params.n_gpu_per_node
# define whether this is the master process / if we are in multi-node distributed mode
params.is_master = params.node_id == 0 and params.local_rank == 0
params.multi_node = params.n_nodes > 1
# summary
PREFIX = f"--- Global rank: {params.global_rank} - "
logger.info(PREFIX + "Number of nodes: %i" % params.n_nodes)
logger.info(PREFIX + "Node ID : %i" % params.node_id)
logger.info(PREFIX + "Local rank : %i" % params.local_rank)
logger.info(PREFIX + "World size : %i" % params.world_size)
logger.info(PREFIX + "GPUs per node : %i" % params.n_gpu_per_node)
logger.info(PREFIX + "Master : %s" % str(params.is_master))
logger.info(PREFIX + "Multi-node : %s" % str(params.multi_node))
logger.info(PREFIX + "Multi-GPU : %s" % str(params.multi_gpu))
logger.info(PREFIX + "Hostname : %s" % socket.gethostname())
# set GPU device
torch.cuda.set_device(params.local_rank)
# initialize multi-GPU
if params.multi_gpu:
logger.info("Initializing PyTorch distributed")
torch.distributed.init_process_group(
init_method="env://",
backend="nccl",
)
def set_seed(args):
"""
Set the random seed.
"""
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if args.n_gpu > 0:
torch.cuda.manual_seed_all(args.seed)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/train.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team.
#
# 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.
"""
Training the distilled model.
Supported architectures include: BERT -> DistilBERT, RoBERTa -> DistilRoBERTa, GPT2 -> DistilGPT2.
"""
import argparse
import json
import os
import pickle
import shutil
import numpy as np
import torch
from distiller import Distiller
from lm_seqs_dataset import LmSeqsDataset
from transformers import (
BertConfig,
BertForMaskedLM,
BertTokenizer,
DistilBertConfig,
DistilBertForMaskedLM,
DistilBertTokenizer,
GPT2Config,
GPT2LMHeadModel,
GPT2Tokenizer,
RobertaConfig,
RobertaForMaskedLM,
RobertaTokenizer,
)
from utils import git_log, init_gpu_params, logger, set_seed
MODEL_CLASSES = {
"distilbert": (DistilBertConfig, DistilBertForMaskedLM, DistilBertTokenizer),
"roberta": (RobertaConfig, RobertaForMaskedLM, RobertaTokenizer),
"bert": (BertConfig, BertForMaskedLM, BertTokenizer),
"gpt2": (GPT2Config, GPT2LMHeadModel, GPT2Tokenizer),
}
def sanity_checks(args):
"""
A bunch of args sanity checks to perform even starting...
"""
assert (args.mlm and args.alpha_mlm > 0.0) or (not args.mlm and args.alpha_mlm == 0.0)
assert (args.alpha_mlm > 0.0 and args.alpha_clm == 0.0) or (args.alpha_mlm == 0.0 and args.alpha_clm > 0.0)
if args.mlm:
assert os.path.isfile(args.token_counts)
assert (args.student_type in ["roberta", "distilbert"]) and (args.teacher_type in ["roberta", "bert"])
else:
assert (args.student_type in ["gpt2"]) and (args.teacher_type in ["gpt2"])
assert args.teacher_type == args.student_type or (
args.student_type == "distilbert" and args.teacher_type == "bert"
)
assert os.path.isfile(args.student_config)
if args.student_pretrained_weights is not None:
assert os.path.isfile(args.student_pretrained_weights)
if args.freeze_token_type_embds:
assert args.student_type in ["roberta"]
assert args.alpha_ce >= 0.0
assert args.alpha_mlm >= 0.0
assert args.alpha_clm >= 0.0
assert args.alpha_mse >= 0.0
assert args.alpha_cos >= 0.0
assert args.alpha_ce + args.alpha_mlm + args.alpha_clm + args.alpha_mse + args.alpha_cos > 0.0
def freeze_pos_embeddings(student, args):
if args.student_type == "roberta":
student.roberta.embeddings.position_embeddings.weight.requires_grad = False
elif args.student_type == "gpt2":
student.transformer.wpe.weight.requires_grad = False
def freeze_token_type_embeddings(student, args):
if args.student_type == "roberta":
student.roberta.embeddings.token_type_embeddings.weight.requires_grad = False
def main():
parser = argparse.ArgumentParser(description="Training")
parser.add_argument("--force", action="store_true", help="Overwrite dump_path if it already exists.")
parser.add_argument(
"--dump_path", type=str, required=True, help="The output directory (log, checkpoints, parameters, etc.)"
)
parser.add_argument(
"--data_file",
type=str,
required=True,
help="The binarized file (tokenized + tokens_to_ids) and grouped by sequence.",
)
parser.add_argument(
"--student_type",
type=str,
choices=["distilbert", "roberta", "gpt2"],
required=True,
help="The student type (DistilBERT, RoBERTa).",
)
parser.add_argument("--student_config", type=str, required=True, help="Path to the student configuration.")
parser.add_argument(
"--student_pretrained_weights", default=None, type=str, help="Load student initialization checkpoint."
)
parser.add_argument(
"--teacher_type", choices=["bert", "roberta", "gpt2"], required=True, help="Teacher type (BERT, RoBERTa)."
)
parser.add_argument("--teacher_name", type=str, required=True, help="The teacher model.")
parser.add_argument("--temperature", default=2.0, type=float, help="Temperature for the softmax temperature.")
parser.add_argument(
"--alpha_ce", default=0.5, type=float, help="Linear weight for the distillation loss. Must be >=0."
)
parser.add_argument(
"--alpha_mlm",
default=0.0,
type=float,
help="Linear weight for the MLM loss. Must be >=0. Should be used in conjunction with `mlm` flag.",
)
parser.add_argument("--alpha_clm", default=0.5, type=float, help="Linear weight for the CLM loss. Must be >=0.")
parser.add_argument("--alpha_mse", default=0.0, type=float, help="Linear weight of the MSE loss. Must be >=0.")
parser.add_argument(
"--alpha_cos", default=0.0, type=float, help="Linear weight of the cosine embedding loss. Must be >=0."
)
parser.add_argument(
"--mlm", action="store_true", help="The LM step: MLM or CLM. If `mlm` is True, the MLM is used over CLM."
)
parser.add_argument(
"--mlm_mask_prop",
default=0.15,
type=float,
help="Proportion of tokens for which we need to make a prediction.",
)
parser.add_argument("--word_mask", default=0.8, type=float, help="Proportion of tokens to mask out.")
parser.add_argument("--word_keep", default=0.1, type=float, help="Proportion of tokens to keep.")
parser.add_argument("--word_rand", default=0.1, type=float, help="Proportion of tokens to randomly replace.")
parser.add_argument(
"--mlm_smoothing",
default=0.7,
type=float,
help="Smoothing parameter to emphasize more rare tokens (see XLM, similar to word2vec).",
)
parser.add_argument("--token_counts", type=str, help="The token counts in the data_file for MLM.")
parser.add_argument(
"--restrict_ce_to_mask",
action="store_true",
help="If true, compute the distillation loss only the [MLM] prediction distribution.",
)
parser.add_argument(
"--freeze_pos_embs",
action="store_true",
help="Freeze positional embeddings during distillation. For student_type in ['roberta', 'gpt2'] only.",
)
parser.add_argument(
"--freeze_token_type_embds",
action="store_true",
help="Freeze token type embeddings during distillation if existent. For student_type in ['roberta'] only.",
)
parser.add_argument("--n_epoch", type=int, default=3, help="Number of pass on the whole dataset.")
parser.add_argument("--batch_size", type=int, default=5, help="Batch size (for each process).")
parser.add_argument(
"--group_by_size",
action="store_false",
help="If true, group sequences that have similar length into the same batch. Default is true.",
)
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=50,
help="Gradient accumulation for larger training batches.",
)
parser.add_argument("--warmup_prop", default=0.05, type=float, help="Linear warmup proportion.")
parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight decay if we apply some.")
parser.add_argument("--learning_rate", default=5e-4, type=float, help="The initial learning rate for Adam.")
parser.add_argument("--adam_epsilon", default=1e-6, type=float, help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=5.0, type=float, help="Max gradient norm.")
parser.add_argument("--initializer_range", default=0.02, type=float, help="Random initialization range.")
parser.add_argument(
"--fp16",
action="store_true",
help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit",
)
parser.add_argument(
"--fp16_opt_level",
type=str,
default="O1",
help=(
"For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
"See details at https://nvidia.github.io/apex/amp.html"
),
)
parser.add_argument("--n_gpu", type=int, default=1, help="Number of GPUs in the node.")
parser.add_argument("--local_rank", type=int, default=-1, help="Distributed training - Local rank")
parser.add_argument("--seed", type=int, default=56, help="Random seed")
parser.add_argument("--log_interval", type=int, default=500, help="Tensorboard logging interval.")
parser.add_argument("--checkpoint_interval", type=int, default=4000, help="Checkpoint interval.")
args = parser.parse_args()
sanity_checks(args)
# ARGS #
init_gpu_params(args)
set_seed(args)
if args.is_master:
if os.path.exists(args.dump_path):
if not args.force:
raise ValueError(
f"Serialization dir {args.dump_path} already exists, but you have not precised wheter to overwrite"
" itUse `--force` if you want to overwrite it"
)
else:
shutil.rmtree(args.dump_path)
if not os.path.exists(args.dump_path):
os.makedirs(args.dump_path)
logger.info(f"Experiment will be dumped and logged in {args.dump_path}")
# SAVE PARAMS #
logger.info(f"Param: {args}")
with open(os.path.join(args.dump_path, "parameters.json"), "w") as f:
json.dump(vars(args), f, indent=4)
git_log(args.dump_path)
student_config_class, student_model_class, _ = MODEL_CLASSES[args.student_type]
teacher_config_class, teacher_model_class, teacher_tokenizer_class = MODEL_CLASSES[args.teacher_type]
# TOKENIZER #
tokenizer = teacher_tokenizer_class.from_pretrained(args.teacher_name)
special_tok_ids = {}
for tok_name, tok_symbol in tokenizer.special_tokens_map.items():
idx = tokenizer.all_special_tokens.index(tok_symbol)
special_tok_ids[tok_name] = tokenizer.all_special_ids[idx]
logger.info(f"Special tokens {special_tok_ids}")
args.special_tok_ids = special_tok_ids
args.max_model_input_size = tokenizer.max_model_input_sizes[args.teacher_name]
# DATA LOADER #
logger.info(f"Loading data from {args.data_file}")
with open(args.data_file, "rb") as fp:
data = pickle.load(fp)
if args.mlm:
logger.info(f"Loading token counts from {args.token_counts} (already pre-computed)")
with open(args.token_counts, "rb") as fp:
counts = pickle.load(fp)
token_probs = np.maximum(counts, 1) ** -args.mlm_smoothing
for idx in special_tok_ids.values():
token_probs[idx] = 0.0 # do not predict special tokens
token_probs = torch.from_numpy(token_probs)
else:
token_probs = None
train_lm_seq_dataset = LmSeqsDataset(params=args, data=data)
logger.info("Data loader created.")
# STUDENT #
logger.info(f"Loading student config from {args.student_config}")
stu_architecture_config = student_config_class.from_pretrained(args.student_config)
stu_architecture_config.output_hidden_states = True
if args.student_pretrained_weights is not None:
logger.info(f"Loading pretrained weights from {args.student_pretrained_weights}")
student = student_model_class.from_pretrained(args.student_pretrained_weights, config=stu_architecture_config)
else:
student = student_model_class(stu_architecture_config)
if args.n_gpu > 0:
student.to(f"cuda:{args.local_rank}")
logger.info("Student loaded.")
# TEACHER #
teacher = teacher_model_class.from_pretrained(args.teacher_name, output_hidden_states=True)
if args.n_gpu > 0:
teacher.to(f"cuda:{args.local_rank}")
logger.info(f"Teacher loaded from {args.teacher_name}.")
# FREEZING #
if args.freeze_pos_embs:
freeze_pos_embeddings(student, args)
if args.freeze_token_type_embds:
freeze_token_type_embeddings(student, args)
# SANITY CHECKS #
assert student.config.vocab_size == teacher.config.vocab_size
assert student.config.hidden_size == teacher.config.hidden_size
assert student.config.max_position_embeddings == teacher.config.max_position_embeddings
if args.mlm:
assert token_probs.size(0) == stu_architecture_config.vocab_size
# DISTILLER #
torch.cuda.empty_cache()
distiller = Distiller(
params=args, dataset=train_lm_seq_dataset, token_probs=token_probs, student=student, teacher=teacher
)
distiller.train()
logger.info("Let's go get some drinks.")
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/grouped_batch_sampler.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team and Facebook, Inc.
#
# 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.
""" Adapted from PyTorch Vision (https://github.com/pytorch/vision/blob/master/references/detection/group_by_aspect_ratio.py)
"""
import bisect
import copy
from collections import defaultdict
import numpy as np
from torch.utils.data import BatchSampler, Sampler
from utils import logger
def _quantize(x, bins):
bins = copy.deepcopy(bins)
bins = sorted(bins)
quantized = [bisect.bisect_right(bins, y) for y in x]
return quantized
def create_lengths_groups(lengths, k=0):
bins = np.arange(start=3, stop=k, step=4).tolist() if k > 0 else [10]
groups = _quantize(lengths, bins)
# count number of elements per group
counts = np.unique(groups, return_counts=True)[1]
fbins = [0] + bins + [np.inf]
logger.info("Using {} as bins for aspect lengths quantization".format(fbins))
logger.info("Count of instances per bin: {}".format(counts))
return groups
class GroupedBatchSampler(BatchSampler):
"""
Wraps another sampler to yield a mini-batch of indices.
It enforces that the batch only contain elements from the same group.
It also tries to provide mini-batches which follows an ordering which is
as close as possible to the ordering from the original sampler.
Arguments:
sampler (Sampler): Base sampler.
group_ids (list[int]): If the sampler produces indices in range [0, N),
`group_ids` must be a list of `N` ints which contains the group id of each sample.
The group ids must be a continuous set of integers starting from
0, i.e. they must be in the range [0, num_groups).
batch_size (int): Size of mini-batch.
"""
def __init__(self, sampler, group_ids, batch_size):
if not isinstance(sampler, Sampler):
raise ValueError(
"sampler should be an instance of torch.utils.data.Sampler, but got sampler={}".format(sampler)
)
self.sampler = sampler
self.group_ids = group_ids
self.batch_size = batch_size
def __iter__(self):
buffer_per_group = defaultdict(list)
samples_per_group = defaultdict(list)
num_batches = 0
for idx in self.sampler:
group_id = self.group_ids[idx]
buffer_per_group[group_id].append(idx)
samples_per_group[group_id].append(idx)
if len(buffer_per_group[group_id]) == self.batch_size:
yield buffer_per_group[group_id] # TODO
num_batches += 1
del buffer_per_group[group_id]
assert len(buffer_per_group[group_id]) < self.batch_size
# now we have run out of elements that satisfy
# the group criteria, let's return the remaining
# elements so that the size of the sampler is
# deterministic
expected_num_batches = len(self)
num_remaining = expected_num_batches - num_batches
if num_remaining > 0:
# for the remaining batches, group the batches by similar lengths
batch_idx = []
for group_id, idxs in sorted(buffer_per_group.items(), key=lambda x: x[0]):
batch_idx.extend(idxs)
if len(batch_idx) >= self.batch_size:
yield batch_idx[: self.batch_size]
batch_idx = batch_idx[self.batch_size :]
num_remaining -= 1
if len(batch_idx) > 0:
yield batch_idx
num_remaining -= 1
assert num_remaining == 0
def __len__(self):
"""
Return the number of mini-batches rather than the number of samples.
"""
return (len(self.sampler) + self.batch_size - 1) // self.batch_size
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/requirements.txt
|
transformers
gitpython==3.1.32
tensorboard>=1.14.0
tensorboardX==1.8
psutil==5.6.6
scipy>=1.4.1
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/distillation/run_squad_w_distillation.py
|
# 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.
""" This is the exact same script as `examples/question-answering/run_squad.py` (as of 2020, January 8th) with an additional and optional step of distillation."""
import argparse
import glob
import logging
import os
import random
import timeit
import numpy as np
import torch
from torch import nn
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm, trange
import transformers
from transformers import (
WEIGHTS_NAME,
AdamW,
BertConfig,
BertForQuestionAnswering,
BertTokenizer,
DistilBertConfig,
DistilBertForQuestionAnswering,
DistilBertTokenizer,
RobertaConfig,
RobertaForQuestionAnswering,
RobertaTokenizer,
XLMConfig,
XLMForQuestionAnswering,
XLMTokenizer,
XLNetConfig,
XLNetForQuestionAnswering,
XLNetTokenizer,
get_linear_schedule_with_warmup,
squad_convert_examples_to_features,
)
from transformers.data.metrics.squad_metrics import (
compute_predictions_log_probs,
compute_predictions_logits,
squad_evaluate,
)
from transformers.data.processors.squad import SquadResult, SquadV1Processor, SquadV2Processor
from transformers.trainer_utils import is_main_process
try:
from torch.utils.tensorboard import SummaryWriter
except ImportError:
from tensorboardX import SummaryWriter
logger = logging.getLogger(__name__)
MODEL_CLASSES = {
"bert": (BertConfig, BertForQuestionAnswering, BertTokenizer),
"xlnet": (XLNetConfig, XLNetForQuestionAnswering, XLNetTokenizer),
"xlm": (XLMConfig, XLMForQuestionAnswering, XLMTokenizer),
"distilbert": (DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer),
"roberta": (RobertaConfig, RobertaForQuestionAnswering, RobertaTokenizer),
}
def set_seed(args):
random.seed(args.seed)
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if args.n_gpu > 0:
torch.cuda.manual_seed_all(args.seed)
def to_list(tensor):
return tensor.detach().cpu().tolist()
def train(args, train_dataset, model, tokenizer, teacher=None):
"""Train the model"""
if args.local_rank in [-1, 0]:
tb_writer = SummaryWriter()
args.train_batch_size = args.per_gpu_train_batch_size * max(1, args.n_gpu)
train_sampler = RandomSampler(train_dataset) if args.local_rank == -1 else DistributedSampler(train_dataset)
train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size)
if args.max_steps > 0:
t_total = args.max_steps
args.num_train_epochs = args.max_steps // (len(train_dataloader) // args.gradient_accumulation_steps) + 1
else:
t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs
# Prepare optimizer and schedule (linear warmup and decay)
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{
"params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], "weight_decay": 0.0},
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
scheduler = get_linear_schedule_with_warmup(
optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
)
# Check if saved optimizer or scheduler states exist
if os.path.isfile(os.path.join(args.model_name_or_path, "optimizer.pt")) and os.path.isfile(
os.path.join(args.model_name_or_path, "scheduler.pt")
):
# Load in optimizer and scheduler states
optimizer.load_state_dict(torch.load(os.path.join(args.model_name_or_path, "optimizer.pt")))
scheduler.load_state_dict(torch.load(os.path.join(args.model_name_or_path, "scheduler.pt")))
if args.fp16:
try:
from apex import amp
except ImportError:
raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
model, optimizer = amp.initialize(model, optimizer, opt_level=args.fp16_opt_level)
# multi-gpu training (should be after apex fp16 initialization)
if args.n_gpu > 1:
model = nn.DataParallel(model)
# Distributed training (should be after apex fp16 initialization)
if args.local_rank != -1:
model = nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=True
)
# Train!
logger.info("***** Running training *****")
logger.info(" Num examples = %d", len(train_dataset))
logger.info(" Num Epochs = %d", args.num_train_epochs)
logger.info(" Instantaneous batch size per GPU = %d", args.per_gpu_train_batch_size)
logger.info(
" Total train batch size (w. parallel, distributed & accumulation) = %d",
args.train_batch_size
* args.gradient_accumulation_steps
* (torch.distributed.get_world_size() if args.local_rank != -1 else 1),
)
logger.info(" Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
logger.info(" Total optimization steps = %d", t_total)
global_step = 1
epochs_trained = 0
steps_trained_in_current_epoch = 0
# Check if continuing training from a checkpoint
if os.path.exists(args.model_name_or_path):
try:
# set global_step to gobal_step of last saved checkpoint from model path
checkpoint_suffix = args.model_name_or_path.split("-")[-1].split("/")[0]
global_step = int(checkpoint_suffix)
epochs_trained = global_step // (len(train_dataloader) // args.gradient_accumulation_steps)
steps_trained_in_current_epoch = global_step % (len(train_dataloader) // args.gradient_accumulation_steps)
logger.info(" Continuing training from checkpoint, will skip to saved global_step")
logger.info(" Continuing training from epoch %d", epochs_trained)
logger.info(" Continuing training from global step %d", global_step)
logger.info(" Will skip the first %d steps in the first epoch", steps_trained_in_current_epoch)
except ValueError:
logger.info(" Starting fine-tuning.")
tr_loss, logging_loss = 0.0, 0.0
model.zero_grad()
train_iterator = trange(
epochs_trained, int(args.num_train_epochs), desc="Epoch", disable=args.local_rank not in [-1, 0]
)
# Added here for reproductibility
set_seed(args)
for _ in train_iterator:
epoch_iterator = tqdm(train_dataloader, desc="Iteration", disable=args.local_rank not in [-1, 0])
for step, batch in enumerate(epoch_iterator):
# Skip past any already trained steps if resuming training
if steps_trained_in_current_epoch > 0:
steps_trained_in_current_epoch -= 1
continue
model.train()
if teacher is not None:
teacher.eval()
batch = tuple(t.to(args.device) for t in batch)
inputs = {
"input_ids": batch[0],
"attention_mask": batch[1],
"start_positions": batch[3],
"end_positions": batch[4],
}
if args.model_type != "distilbert":
inputs["token_type_ids"] = None if args.model_type == "xlm" else batch[2]
if args.model_type in ["xlnet", "xlm"]:
inputs.update({"cls_index": batch[5], "p_mask": batch[6]})
if args.version_2_with_negative:
inputs.update({"is_impossible": batch[7]})
outputs = model(**inputs)
loss, start_logits_stu, end_logits_stu = outputs
# Distillation loss
if teacher is not None:
if "token_type_ids" not in inputs:
inputs["token_type_ids"] = None if args.teacher_type == "xlm" else batch[2]
with torch.no_grad():
start_logits_tea, end_logits_tea = teacher(
input_ids=inputs["input_ids"],
token_type_ids=inputs["token_type_ids"],
attention_mask=inputs["attention_mask"],
)
assert start_logits_tea.size() == start_logits_stu.size()
assert end_logits_tea.size() == end_logits_stu.size()
loss_fct = nn.KLDivLoss(reduction="batchmean")
loss_start = loss_fct(
nn.functional.log_softmax(start_logits_stu / args.temperature, dim=-1),
nn.functional.softmax(start_logits_tea / args.temperature, dim=-1),
) * (args.temperature**2)
loss_end = loss_fct(
nn.functional.log_softmax(end_logits_stu / args.temperature, dim=-1),
nn.functional.softmax(end_logits_tea / args.temperature, dim=-1),
) * (args.temperature**2)
loss_ce = (loss_start + loss_end) / 2.0
loss = args.alpha_ce * loss_ce + args.alpha_squad * loss
if args.n_gpu > 1:
loss = loss.mean() # mean() to average on multi-gpu parallel (not distributed) training
if args.gradient_accumulation_steps > 1:
loss = loss / args.gradient_accumulation_steps
if args.fp16:
with amp.scale_loss(loss, optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
tr_loss += loss.item()
if (step + 1) % args.gradient_accumulation_steps == 0:
if args.fp16:
nn.utils.clip_grad_norm_(amp.master_params(optimizer), args.max_grad_norm)
else:
nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)
optimizer.step()
scheduler.step() # Update learning rate schedule
model.zero_grad()
global_step += 1
# Log metrics
if args.local_rank in [-1, 0] and args.logging_steps > 0 and global_step % args.logging_steps == 0:
# Only evaluate when single GPU otherwise metrics may not average well
if args.local_rank == -1 and args.evaluate_during_training:
results = evaluate(args, model, tokenizer)
for key, value in results.items():
tb_writer.add_scalar("eval_{}".format(key), value, global_step)
tb_writer.add_scalar("lr", scheduler.get_lr()[0], global_step)
tb_writer.add_scalar("loss", (tr_loss - logging_loss) / args.logging_steps, global_step)
logging_loss = tr_loss
if args.local_rank in [-1, 0] and args.save_steps > 0 and global_step % args.save_steps == 0:
# Save model checkpoint
output_dir = os.path.join(args.output_dir, "checkpoint-{}".format(global_step))
if not os.path.exists(output_dir):
os.makedirs(output_dir)
model_to_save = (
model.module if hasattr(model, "module") else model
) # Take care of distributed/parallel training
model_to_save.save_pretrained(output_dir)
tokenizer.save_pretrained(output_dir)
torch.save(args, os.path.join(output_dir, "training_args.bin"))
logger.info("Saving model checkpoint to %s", output_dir)
torch.save(optimizer.state_dict(), os.path.join(output_dir, "optimizer.pt"))
torch.save(scheduler.state_dict(), os.path.join(output_dir, "scheduler.pt"))
logger.info("Saving optimizer and scheduler states to %s", output_dir)
if args.max_steps > 0 and global_step > args.max_steps:
epoch_iterator.close()
break
if args.max_steps > 0 and global_step > args.max_steps:
train_iterator.close()
break
if args.local_rank in [-1, 0]:
tb_writer.close()
return global_step, tr_loss / global_step
def evaluate(args, model, tokenizer, prefix=""):
dataset, examples, features = load_and_cache_examples(args, tokenizer, evaluate=True, output_examples=True)
if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
os.makedirs(args.output_dir)
args.eval_batch_size = args.per_gpu_eval_batch_size * max(1, args.n_gpu)
# Note that DistributedSampler samples randomly
eval_sampler = SequentialSampler(dataset)
eval_dataloader = DataLoader(dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)
# multi-gpu evaluate
if args.n_gpu > 1 and not isinstance(model, nn.DataParallel):
model = nn.DataParallel(model)
# Eval!
logger.info("***** Running evaluation {} *****".format(prefix))
logger.info(" Num examples = %d", len(dataset))
logger.info(" Batch size = %d", args.eval_batch_size)
all_results = []
start_time = timeit.default_timer()
for batch in tqdm(eval_dataloader, desc="Evaluating"):
model.eval()
batch = tuple(t.to(args.device) for t in batch)
with torch.no_grad():
inputs = {"input_ids": batch[0], "attention_mask": batch[1]}
if args.model_type != "distilbert":
inputs["token_type_ids"] = None if args.model_type == "xlm" else batch[2] # XLM don't use segment_ids
example_indices = batch[3]
if args.model_type in ["xlnet", "xlm"]:
inputs.update({"cls_index": batch[4], "p_mask": batch[5]})
outputs = model(**inputs)
for i, example_index in enumerate(example_indices):
eval_feature = features[example_index.item()]
unique_id = int(eval_feature.unique_id)
output = [to_list(output[i]) for output in outputs]
# Some models (XLNet, XLM) use 5 arguments for their predictions, while the other "simpler"
# models only use two.
if len(output) >= 5:
start_logits = output[0]
start_top_index = output[1]
end_logits = output[2]
end_top_index = output[3]
cls_logits = output[4]
result = SquadResult(
unique_id,
start_logits,
end_logits,
start_top_index=start_top_index,
end_top_index=end_top_index,
cls_logits=cls_logits,
)
else:
start_logits, end_logits = output
result = SquadResult(unique_id, start_logits, end_logits)
all_results.append(result)
evalTime = timeit.default_timer() - start_time
logger.info(" Evaluation done in total %f secs (%f sec per example)", evalTime, evalTime / len(dataset))
# Compute predictions
output_prediction_file = os.path.join(args.output_dir, "predictions_{}.json".format(prefix))
output_nbest_file = os.path.join(args.output_dir, "nbest_predictions_{}.json".format(prefix))
if args.version_2_with_negative:
output_null_log_odds_file = os.path.join(args.output_dir, "null_odds_{}.json".format(prefix))
else:
output_null_log_odds_file = None
if args.model_type in ["xlnet", "xlm"]:
# XLNet uses a more complex post-processing procedure
predictions = compute_predictions_log_probs(
examples,
features,
all_results,
args.n_best_size,
args.max_answer_length,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
model.config.start_n_top,
model.config.end_n_top,
args.version_2_with_negative,
tokenizer,
args.verbose_logging,
)
else:
predictions = compute_predictions_logits(
examples,
features,
all_results,
args.n_best_size,
args.max_answer_length,
args.do_lower_case,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
args.verbose_logging,
args.version_2_with_negative,
args.null_score_diff_threshold,
tokenizer,
)
# Compute the F1 and exact scores.
results = squad_evaluate(examples, predictions)
return results
def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=False):
if args.local_rank not in [-1, 0] and not evaluate:
# Make sure only the first process in distributed training process the dataset, and the others will use the cache
torch.distributed.barrier()
# Load data features from cache or dataset file
input_file = args.predict_file if evaluate else args.train_file
cached_features_file = os.path.join(
os.path.dirname(input_file),
"cached_distillation_{}_{}_{}".format(
"dev" if evaluate else "train",
list(filter(None, args.model_name_or_path.split("/"))).pop(),
str(args.max_seq_length),
),
)
if os.path.exists(cached_features_file) and not args.overwrite_cache:
logger.info("Loading features from cached file %s", cached_features_file)
features_and_dataset = torch.load(cached_features_file)
try:
features, dataset, examples = (
features_and_dataset["features"],
features_and_dataset["dataset"],
features_and_dataset["examples"],
)
except KeyError:
raise DeprecationWarning(
"You seem to be loading features from an older version of this script please delete the "
"file %s in order for it to be created again" % cached_features_file
)
else:
logger.info("Creating features from dataset file at %s", input_file)
processor = SquadV2Processor() if args.version_2_with_negative else SquadV1Processor()
if evaluate:
examples = processor.get_dev_examples(args.data_dir, filename=args.predict_file)
else:
examples = processor.get_train_examples(args.data_dir, filename=args.train_file)
features, dataset = squad_convert_examples_to_features(
examples=examples,
tokenizer=tokenizer,
max_seq_length=args.max_seq_length,
doc_stride=args.doc_stride,
max_query_length=args.max_query_length,
is_training=not evaluate,
return_dataset="pt",
threads=args.threads,
)
if args.local_rank in [-1, 0]:
logger.info("Saving features into cached file %s", cached_features_file)
torch.save({"features": features, "dataset": dataset, "examples": examples}, cached_features_file)
if args.local_rank == 0 and not evaluate:
# Make sure only the first process in distributed training process the dataset, and the others will use the cache
torch.distributed.barrier()
if output_examples:
return dataset, examples, features
return dataset
def main():
parser = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--model_type",
default=None,
type=str,
required=True,
help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
)
parser.add_argument(
"--model_name_or_path",
default=None,
type=str,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models",
)
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help="The output directory where the model checkpoints and predictions will be written.",
)
# Distillation parameters (optional)
parser.add_argument(
"--teacher_type",
default=None,
type=str,
help=(
"Teacher type. Teacher tokenizer and student (model) tokenizer must output the same tokenization. Only for"
" distillation."
),
)
parser.add_argument(
"--teacher_name_or_path",
default=None,
type=str,
help="Path to the already SQuAD fine-tuned teacher model. Only for distillation.",
)
parser.add_argument(
"--alpha_ce", default=0.5, type=float, help="Distillation loss linear weight. Only for distillation."
)
parser.add_argument(
"--alpha_squad", default=0.5, type=float, help="True SQuAD loss linear weight. Only for distillation."
)
parser.add_argument(
"--temperature", default=2.0, type=float, help="Distillation temperature. Only for distillation."
)
# Other parameters
parser.add_argument(
"--data_dir",
default=None,
type=str,
help="The input data dir. Should contain the .json files for the task."
+ "If no data dir or train/predict files are specified, will run with tensorflow_datasets.",
)
parser.add_argument(
"--train_file",
default=None,
type=str,
help="The input training file. If a data dir is specified, will look for the file there"
+ "If no data dir or train/predict files are specified, will run with tensorflow_datasets.",
)
parser.add_argument(
"--predict_file",
default=None,
type=str,
help="The input evaluation file. If a data dir is specified, will look for the file there"
+ "If no data dir or train/predict files are specified, will run with tensorflow_datasets.",
)
parser.add_argument(
"--config_name", default="", type=str, help="Pretrained config name or path if not the same as model_name"
)
parser.add_argument(
"--tokenizer_name",
default="",
type=str,
help="Pretrained tokenizer name or path if not the same as model_name",
)
parser.add_argument(
"--cache_dir",
default="",
type=str,
help="Where do you want to store the pre-trained models downloaded from huggingface.co",
)
parser.add_argument(
"--version_2_with_negative",
action="store_true",
help="If true, the SQuAD examples contain some that do not have an answer.",
)
parser.add_argument(
"--null_score_diff_threshold",
type=float,
default=0.0,
help="If null_score - best_non_null is greater than the threshold predict null.",
)
parser.add_argument(
"--max_seq_length",
default=384,
type=int,
help=(
"The maximum total input sequence length after WordPiece tokenization. Sequences "
"longer than this will be truncated, and sequences shorter than this will be padded."
),
)
parser.add_argument(
"--doc_stride",
default=128,
type=int,
help="When splitting up a long document into chunks, how much stride to take between chunks.",
)
parser.add_argument(
"--max_query_length",
default=64,
type=int,
help=(
"The maximum number of tokens for the question. Questions longer than this will "
"be truncated to this length."
),
)
parser.add_argument("--do_train", action="store_true", help="Whether to run training.")
parser.add_argument("--do_eval", action="store_true", help="Whether to run eval on the dev set.")
parser.add_argument(
"--evaluate_during_training", action="store_true", help="Rul evaluation during training at each logging step."
)
parser.add_argument(
"--do_lower_case", action="store_true", help="Set this flag if you are using an uncased model."
)
parser.add_argument("--per_gpu_train_batch_size", default=8, type=int, help="Batch size per GPU/CPU for training.")
parser.add_argument(
"--per_gpu_eval_batch_size", default=8, type=int, help="Batch size per GPU/CPU for evaluation."
)
parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight decay if we apply some.")
parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
parser.add_argument(
"--num_train_epochs", default=3.0, type=float, help="Total number of training epochs to perform."
)
parser.add_argument(
"--max_steps",
default=-1,
type=int,
help="If > 0: set total number of training steps to perform. Override num_train_epochs.",
)
parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")
parser.add_argument(
"--n_best_size",
default=20,
type=int,
help="The total number of n-best predictions to generate in the nbest_predictions.json output file.",
)
parser.add_argument(
"--max_answer_length",
default=30,
type=int,
help=(
"The maximum length of an answer that can be generated. This is needed because the start "
"and end predictions are not conditioned on one another."
),
)
parser.add_argument(
"--verbose_logging",
action="store_true",
help=(
"If true, all of the warnings related to data processing will be printed. "
"A number of warnings are expected for a normal SQuAD evaluation."
),
)
parser.add_argument("--logging_steps", type=int, default=50, help="Log every X updates steps.")
parser.add_argument("--save_steps", type=int, default=50, help="Save checkpoint every X updates steps.")
parser.add_argument(
"--eval_all_checkpoints",
action="store_true",
help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number",
)
parser.add_argument("--no_cuda", action="store_true", help="Whether not to use CUDA when available")
parser.add_argument(
"--overwrite_output_dir", action="store_true", help="Overwrite the content of the output directory"
)
parser.add_argument(
"--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
)
parser.add_argument("--seed", type=int, default=42, help="random seed for initialization")
parser.add_argument("--local_rank", type=int, default=-1, help="local_rank for distributed training on gpus")
parser.add_argument(
"--fp16",
action="store_true",
help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit",
)
parser.add_argument(
"--fp16_opt_level",
type=str,
default="O1",
help=(
"For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. "
"See details at https://nvidia.github.io/apex/amp.html"
),
)
parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--threads", type=int, default=1, help="multiple threads for converting example to features")
args = parser.parse_args()
if (
os.path.exists(args.output_dir)
and os.listdir(args.output_dir)
and args.do_train
and not args.overwrite_output_dir
):
raise ValueError(
"Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(
args.output_dir
)
)
# Setup distant debugging if needed
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
# Setup CUDA, GPU & distributed training
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = 0 if args.no_cuda else torch.cuda.device_count()
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
torch.distributed.init_process_group(backend="nccl")
args.n_gpu = 1
args.device = device
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN,
)
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank,
device,
args.n_gpu,
bool(args.local_rank != -1),
args.fp16,
)
# Set the verbosity to info of the Transformers logger (on main process only):
if is_main_process(args.local_rank):
transformers.utils.logging.set_verbosity_info()
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()
# Set seed
set_seed(args)
# Load pretrained model and tokenizer
if args.local_rank not in [-1, 0]:
# Make sure only the first process in distributed training will download model & vocab
torch.distributed.barrier()
args.model_type = args.model_type.lower()
config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
config = config_class.from_pretrained(
args.config_name if args.config_name else args.model_name_or_path,
cache_dir=args.cache_dir if args.cache_dir else None,
)
tokenizer = tokenizer_class.from_pretrained(
args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
do_lower_case=args.do_lower_case,
cache_dir=args.cache_dir if args.cache_dir else None,
)
model = model_class.from_pretrained(
args.model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
cache_dir=args.cache_dir if args.cache_dir else None,
)
if args.teacher_type is not None:
assert args.teacher_name_or_path is not None
assert args.alpha_ce > 0.0
assert args.alpha_ce + args.alpha_squad > 0.0
assert args.teacher_type != "distilbert", "We constraint teachers not to be of type DistilBERT."
teacher_config_class, teacher_model_class, _ = MODEL_CLASSES[args.teacher_type]
teacher_config = teacher_config_class.from_pretrained(
args.teacher_name_or_path, cache_dir=args.cache_dir if args.cache_dir else None
)
teacher = teacher_model_class.from_pretrained(
args.teacher_name_or_path, config=teacher_config, cache_dir=args.cache_dir if args.cache_dir else None
)
teacher.to(args.device)
else:
teacher = None
if args.local_rank == 0:
# Make sure only the first process in distributed training will download model & vocab
torch.distributed.barrier()
model.to(args.device)
logger.info("Training/evaluation parameters %s", args)
# Before we do anything with models, we want to ensure that we get fp16 execution of torch.einsum if args.fp16 is set.
# Otherwise it'll default to "promote" mode, and we'll get fp32 operations. Note that running `--fp16_opt_level="O2"` will
# remove the need for this code, but it is still valid.
if args.fp16:
try:
import apex
apex.amp.register_half_function(torch, "einsum")
except ImportError:
raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
# Training
if args.do_train:
train_dataset = load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=False)
global_step, tr_loss = train(args, train_dataset, model, tokenizer, teacher=teacher)
logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)
# Save the trained model and the tokenizer
if args.do_train and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
logger.info("Saving model checkpoint to %s", args.output_dir)
# Save a trained model, configuration and tokenizer using `save_pretrained()`.
# They can then be reloaded using `from_pretrained()`
model_to_save = (
model.module if hasattr(model, "module") else model
) # Take care of distributed/parallel training
model_to_save.save_pretrained(args.output_dir)
tokenizer.save_pretrained(args.output_dir)
# Good practice: save your training arguments together with the trained model
torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
# Load a trained model and vocabulary that you have fine-tuned
model = model_class.from_pretrained(args.output_dir)
tokenizer = tokenizer_class.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
model.to(args.device)
# Evaluation - we can ask to evaluate all the checkpoints (sub-directories) in a directory
results = {}
if args.do_eval and args.local_rank in [-1, 0]:
if args.do_train:
logger.info("Loading checkpoints saved during training for evaluation")
checkpoints = [args.output_dir]
if args.eval_all_checkpoints:
checkpoints = [
os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + "/**/" + WEIGHTS_NAME, recursive=True))
]
logger.info("Evaluate the following checkpoints: %s", checkpoints)
for checkpoint in checkpoints:
# Reload the model
global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
model = model_class.from_pretrained(checkpoint)
model.to(args.device)
# Evaluate
result = evaluate(args, model, tokenizer, prefix=global_step)
result = {k + ("_{}".format(global_step) if global_step else ""): v for k, v in result.items()}
results.update(result)
logger.info("Results: {}".format(results))
return results
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/scripts/extract_distilbert.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team.
#
# 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.
"""
Preprocessing script before training DistilBERT.
Specific to BERT -> DistilBERT.
"""
import argparse
import torch
from transformers import BertForMaskedLM
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description=(
"Extraction some layers of the full BertForMaskedLM or RObertaForMaskedLM for Transfer Learned"
" Distillation"
)
)
parser.add_argument("--model_type", default="bert", choices=["bert"])
parser.add_argument("--model_name", default="bert-base-uncased", type=str)
parser.add_argument("--dump_checkpoint", default="serialization_dir/tf_bert-base-uncased_0247911.pth", type=str)
parser.add_argument("--vocab_transform", action="store_true")
args = parser.parse_args()
if args.model_type == "bert":
model = BertForMaskedLM.from_pretrained(args.model_name)
prefix = "bert"
else:
raise ValueError('args.model_type should be "bert".')
state_dict = model.state_dict()
compressed_sd = {}
for w in ["word_embeddings", "position_embeddings"]:
compressed_sd[f"distilbert.embeddings.{w}.weight"] = state_dict[f"{prefix}.embeddings.{w}.weight"]
for w in ["weight", "bias"]:
compressed_sd[f"distilbert.embeddings.LayerNorm.{w}"] = state_dict[f"{prefix}.embeddings.LayerNorm.{w}"]
std_idx = 0
for teacher_idx in [0, 2, 4, 7, 9, 11]:
for w in ["weight", "bias"]:
compressed_sd[f"distilbert.transformer.layer.{std_idx}.attention.q_lin.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.attention.self.query.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.attention.k_lin.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.attention.self.key.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.attention.v_lin.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.attention.self.value.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.attention.out_lin.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.attention.output.dense.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.sa_layer_norm.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.attention.output.LayerNorm.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.ffn.lin1.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.intermediate.dense.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.ffn.lin2.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.output.dense.{w}"
]
compressed_sd[f"distilbert.transformer.layer.{std_idx}.output_layer_norm.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.output.LayerNorm.{w}"
]
std_idx += 1
compressed_sd["vocab_projector.weight"] = state_dict["cls.predictions.decoder.weight"]
compressed_sd["vocab_projector.bias"] = state_dict["cls.predictions.bias"]
if args.vocab_transform:
for w in ["weight", "bias"]:
compressed_sd[f"vocab_transform.{w}"] = state_dict[f"cls.predictions.transform.dense.{w}"]
compressed_sd[f"vocab_layer_norm.{w}"] = state_dict[f"cls.predictions.transform.LayerNorm.{w}"]
print(f"N layers selected for distillation: {std_idx}")
print(f"Number of params transferred for distillation: {len(compressed_sd.keys())}")
print(f"Save transferred checkpoint to {args.dump_checkpoint}.")
torch.save(compressed_sd, args.dump_checkpoint)
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/scripts/token_counts.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team.
#
# 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.
"""
Preprocessing script before training the distilled model.
"""
import argparse
import logging
import pickle
from collections import Counter
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO
)
logger = logging.getLogger(__name__)
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="Token Counts for smoothing the masking probabilities in MLM (cf XLM/word2vec)"
)
parser.add_argument(
"--data_file", type=str, default="data/dump.bert-base-uncased.pickle", help="The binarized dataset."
)
parser.add_argument(
"--token_counts_dump", type=str, default="data/token_counts.bert-base-uncased.pickle", help="The dump file."
)
parser.add_argument("--vocab_size", default=30522, type=int)
args = parser.parse_args()
logger.info(f"Loading data from {args.data_file}")
with open(args.data_file, "rb") as fp:
data = pickle.load(fp)
logger.info("Counting occurrences for MLM.")
counter = Counter()
for tk_ids in data:
counter.update(tk_ids)
counts = [0] * args.vocab_size
for k, v in counter.items():
counts[k] = v
logger.info(f"Dump to {args.token_counts_dump}")
with open(args.token_counts_dump, "wb") as handle:
pickle.dump(counts, handle, protocol=pickle.HIGHEST_PROTOCOL)
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/scripts/binarized_data.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team.
#
# 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.
"""
Preprocessing script before distillation.
"""
import argparse
import logging
import pickle
import random
import time
import numpy as np
from transformers import BertTokenizer, GPT2Tokenizer, RobertaTokenizer
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO
)
logger = logging.getLogger(__name__)
def main():
parser = argparse.ArgumentParser(
description="Preprocess the data to avoid re-doing it several times by (tokenization + token_to_ids)."
)
parser.add_argument("--file_path", type=str, default="data/dump.txt", help="The path to the data.")
parser.add_argument("--tokenizer_type", type=str, default="bert", choices=["bert", "roberta", "gpt2"])
parser.add_argument("--tokenizer_name", type=str, default="bert-base-uncased", help="The tokenizer to use.")
parser.add_argument("--dump_file", type=str, default="data/dump", help="The dump file prefix.")
args = parser.parse_args()
logger.info(f"Loading Tokenizer ({args.tokenizer_name})")
if args.tokenizer_type == "bert":
tokenizer = BertTokenizer.from_pretrained(args.tokenizer_name)
bos = tokenizer.special_tokens_map["cls_token"] # `[CLS]`
sep = tokenizer.special_tokens_map["sep_token"] # `[SEP]`
elif args.tokenizer_type == "roberta":
tokenizer = RobertaTokenizer.from_pretrained(args.tokenizer_name)
bos = tokenizer.special_tokens_map["cls_token"] # `<s>`
sep = tokenizer.special_tokens_map["sep_token"] # `</s>`
elif args.tokenizer_type == "gpt2":
tokenizer = GPT2Tokenizer.from_pretrained(args.tokenizer_name)
bos = tokenizer.special_tokens_map["bos_token"] # `<|endoftext|>`
sep = tokenizer.special_tokens_map["eos_token"] # `<|endoftext|>`
logger.info(f"Loading text from {args.file_path}")
with open(args.file_path, "r", encoding="utf8") as fp:
data = fp.readlines()
logger.info("Start encoding")
logger.info(f"{len(data)} examples to process.")
rslt = []
iter = 0
interval = 10000
start = time.time()
for text in data:
text = f"{bos} {text.strip()} {sep}"
token_ids = tokenizer.encode(text, add_special_tokens=False)
rslt.append(token_ids)
iter += 1
if iter % interval == 0:
end = time.time()
logger.info(f"{iter} examples processed. - {(end-start):.2f}s/{interval}expl")
start = time.time()
logger.info("Finished binarization")
logger.info(f"{len(data)} examples processed.")
dp_file = f"{args.dump_file}.{args.tokenizer_name}.pickle"
vocab_size = tokenizer.vocab_size
if vocab_size < (1 << 16):
rslt_ = [np.uint16(d) for d in rslt]
else:
rslt_ = [np.int32(d) for d in rslt]
random.shuffle(rslt_)
logger.info(f"Dump to {dp_file}")
with open(dp_file, "wb") as handle:
pickle.dump(rslt_, handle, protocol=pickle.HIGHEST_PROTOCOL)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/scripts/extract.py
|
# coding=utf-8
# Copyright 2019-present, the HuggingFace Inc. team.
#
# 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.
"""
Preprocessing script before training the distilled model.
Specific to RoBERTa -> DistilRoBERTa and GPT2 -> DistilGPT2.
"""
import argparse
import torch
from transformers import GPT2LMHeadModel, RobertaForMaskedLM
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description=(
"Extraction some layers of the full RobertaForMaskedLM or GPT2LMHeadModel for Transfer Learned"
" Distillation"
)
)
parser.add_argument("--model_type", default="roberta", choices=["roberta", "gpt2"])
parser.add_argument("--model_name", default="roberta-large", type=str)
parser.add_argument("--dump_checkpoint", default="serialization_dir/tf_roberta_048131723.pth", type=str)
parser.add_argument("--vocab_transform", action="store_true")
args = parser.parse_args()
if args.model_type == "roberta":
model = RobertaForMaskedLM.from_pretrained(args.model_name)
prefix = "roberta"
elif args.model_type == "gpt2":
model = GPT2LMHeadModel.from_pretrained(args.model_name)
prefix = "transformer"
state_dict = model.state_dict()
compressed_sd = {}
# Embeddings #
if args.model_type == "gpt2":
for param_name in ["wte.weight", "wpe.weight"]:
compressed_sd[f"{prefix}.{param_name}"] = state_dict[f"{prefix}.{param_name}"]
else:
for w in ["word_embeddings", "position_embeddings", "token_type_embeddings"]:
param_name = f"{prefix}.embeddings.{w}.weight"
compressed_sd[param_name] = state_dict[param_name]
for w in ["weight", "bias"]:
param_name = f"{prefix}.embeddings.LayerNorm.{w}"
compressed_sd[param_name] = state_dict[param_name]
# Transformer Blocks #
std_idx = 0
for teacher_idx in [0, 2, 4, 7, 9, 11]:
if args.model_type == "gpt2":
for layer in ["ln_1", "attn.c_attn", "attn.c_proj", "ln_2", "mlp.c_fc", "mlp.c_proj"]:
for w in ["weight", "bias"]:
compressed_sd[f"{prefix}.h.{std_idx}.{layer}.{w}"] = state_dict[
f"{prefix}.h.{teacher_idx}.{layer}.{w}"
]
compressed_sd[f"{prefix}.h.{std_idx}.attn.bias"] = state_dict[f"{prefix}.h.{teacher_idx}.attn.bias"]
else:
for layer in [
"attention.self.query",
"attention.self.key",
"attention.self.value",
"attention.output.dense",
"attention.output.LayerNorm",
"intermediate.dense",
"output.dense",
"output.LayerNorm",
]:
for w in ["weight", "bias"]:
compressed_sd[f"{prefix}.encoder.layer.{std_idx}.{layer}.{w}"] = state_dict[
f"{prefix}.encoder.layer.{teacher_idx}.{layer}.{w}"
]
std_idx += 1
# Language Modeling Head ###s
if args.model_type == "roberta":
for layer in ["lm_head.decoder.weight", "lm_head.bias"]:
compressed_sd[f"{layer}"] = state_dict[f"{layer}"]
if args.vocab_transform:
for w in ["weight", "bias"]:
compressed_sd[f"lm_head.dense.{w}"] = state_dict[f"lm_head.dense.{w}"]
compressed_sd[f"lm_head.layer_norm.{w}"] = state_dict[f"lm_head.layer_norm.{w}"]
elif args.model_type == "gpt2":
for w in ["weight", "bias"]:
compressed_sd[f"{prefix}.ln_f.{w}"] = state_dict[f"{prefix}.ln_f.{w}"]
compressed_sd["lm_head.weight"] = state_dict["lm_head.weight"]
print(f"N layers selected for distillation: {std_idx}")
print(f"Number of params transferred for distillation: {len(compressed_sd.keys())}")
print(f"Save transferred checkpoint to {args.dump_checkpoint}.")
torch.save(compressed_sd, args.dump_checkpoint)
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/training_configs/distilbert-base-cased.json
|
{
"activation": "gelu",
"attention_dropout": 0.1,
"dim": 768,
"dropout": 0.1,
"hidden_dim": 3072,
"initializer_range": 0.02,
"max_position_embeddings": 512,
"n_heads": 12,
"n_layers": 6,
"sinusoidal_pos_embds": true,
"tie_weights_": true,
"vocab_size": 28996
}
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/training_configs/distilbert-base-multilingual-cased.json
|
{
"activation": "gelu",
"attention_dropout": 0.1,
"dim": 768,
"dropout": 0.1,
"hidden_dim": 3072,
"initializer_range": 0.02,
"max_position_embeddings": 512,
"n_heads": 12,
"n_layers": 6,
"sinusoidal_pos_embds": true,
"tie_weights_": true,
"vocab_size": 119547
}
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/training_configs/distilroberta-base.json
|
{
"vocab_size": 50265,
"hidden_size": 768,
"num_hidden_layers": 6,
"num_attention_heads": 12,
"intermediate_size": 3072,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"attention_probs_dropout_prob": 0.1,
"max_position_embeddings": 514,
"type_vocab_size": 1,
"initializer_range": 0.02,
"layer_norm_eps": 0.00001
}
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/training_configs/distilgpt2.json
|
{
"initializer_range": 0.02,
"layer_norm_epsilon": 0.00001,
"n_embd": 768,
"n_head": 12,
"n_layer": 6,
"n_positions": 1024,
"vocab_size": 50257
}
| 0
|
hf_public_repos/transformers/examples/research_projects/distillation
|
hf_public_repos/transformers/examples/research_projects/distillation/training_configs/distilbert-base-uncased.json
|
{
"activation": "gelu",
"attention_dropout": 0.1,
"dim": 768,
"dropout": 0.1,
"hidden_dim": 3072,
"initializer_range": 0.02,
"max_position_embeddings": 512,
"n_heads": 12,
"n_layers": 6,
"sinusoidal_pos_embds": true,
"tie_weights_": true,
"vocab_size": 30522
}
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/mlm_wwm/README.md
|
<!---
Copyright 2020 The HuggingFace Team. 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.
-->
## Whole Word Mask Language Model
These scripts leverage the 🤗 Datasets library and the Trainer API. You can easily customize them to your needs if you
need extra processing on your datasets.
The following examples, will run on a datasets hosted on our [hub](https://huggingface.co/datasets) or with your own
text files for training and validation. We give examples of both below.
The BERT authors released a new version of BERT using Whole Word Masking in May 2019. Instead of masking randomly
selected tokens (which may be part of words), they mask randomly selected words (masking all the tokens corresponding
to that word). This technique has been refined for Chinese in [this paper](https://arxiv.org/abs/1906.08101).
To fine-tune a model using whole word masking, use the following script:
```bash
python run_mlm_wwm.py \
--model_name_or_path roberta-base \
--dataset_name wikitext \
--dataset_config_name wikitext-2-raw-v1 \
--do_train \
--do_eval \
--output_dir /tmp/test-mlm-wwm
```
For Chinese models, we need to generate a reference files (which requires the ltp library), because it's tokenized at
the character level.
**Q :** Why a reference file?
**A :** Suppose we have a Chinese sentence like: `我喜欢你` The original Chinese-BERT will tokenize it as
`['我','喜','欢','你']` (character level). But `喜欢` is a whole word. For whole word masking proxy, we need a result
like `['我','喜','##欢','你']`, so we need a reference file to tell the model which position of the BERT original token
should be added `##`.
**Q :** Why LTP ?
**A :** Cause the best known Chinese WWM BERT is [Chinese-BERT-wwm](https://github.com/ymcui/Chinese-BERT-wwm) by HIT.
It works well on so many Chines Task like CLUE (Chinese GLUE). They use LTP, so if we want to fine-tune their model,
we need LTP.
You could run the following:
```bash
export TRAIN_FILE=/path/to/train/file
export LTP_RESOURCE=/path/to/ltp/tokenizer
export BERT_RESOURCE=/path/to/bert/tokenizer
export SAVE_PATH=/path/to/data/ref.txt
python run_chinese_ref.py \
--file_name=$TRAIN_FILE \
--ltp=$LTP_RESOURCE \
--bert=$BERT_RESOURCE \
--save_path=$SAVE_PATH
```
Then you can run the script like this:
```bash
export TRAIN_FILE=/path/to/train/file
export VALIDATION_FILE=/path/to/validation/file
export TRAIN_REF_FILE=/path/to/train/chinese_ref/file
export VALIDATION_REF_FILE=/path/to/validation/chinese_ref/file
export OUTPUT_DIR=/tmp/test-mlm-wwm
python run_mlm_wwm.py \
--model_name_or_path roberta-base \
--train_file $TRAIN_FILE \
--validation_file $VALIDATION_FILE \
--train_ref_file $TRAIN_REF_FILE \
--validation_ref_file $VALIDATION_REF_FILE \
--do_train \
--do_eval \
--output_dir $OUTPUT_DIR
```
**Note1:** On TPU, you should the flag `--pad_to_max_length` to make sure all your batches have the same length.
**Note2:** And if you have any questions or something goes wrong when runing this code, don't hesitate to pin @wlhgtc.
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/mlm_wwm/run_chinese_ref.py
|
import argparse
import json
from typing import List
from ltp import LTP
from transformers.models.bert.tokenization_bert import BertTokenizer
def _is_chinese_char(cp):
"""Checks whether CP is the codepoint of a CJK character."""
# This defines a "chinese character" as anything in the CJK Unicode block:
# https://en.wikipedia.org/wiki/CJK_Unified_Ideographs_(Unicode_block)
#
# Note that the CJK Unicode block is NOT all Japanese and Korean characters,
# despite its name. The modern Korean Hangul alphabet is a different block,
# as is Japanese Hiragana and Katakana. Those alphabets are used to write
# space-separated words, so they are not treated specially and handled
# like the all of the other languages.
if (
(cp >= 0x4E00 and cp <= 0x9FFF)
or (cp >= 0x3400 and cp <= 0x4DBF) #
or (cp >= 0x20000 and cp <= 0x2A6DF) #
or (cp >= 0x2A700 and cp <= 0x2B73F) #
or (cp >= 0x2B740 and cp <= 0x2B81F) #
or (cp >= 0x2B820 and cp <= 0x2CEAF) #
or (cp >= 0xF900 and cp <= 0xFAFF)
or (cp >= 0x2F800 and cp <= 0x2FA1F) #
): #
return True
return False
def is_chinese(word: str):
# word like '180' or '身高' or '神'
for char in word:
char = ord(char)
if not _is_chinese_char(char):
return 0
return 1
def get_chinese_word(tokens: List[str]):
word_set = set()
for token in tokens:
chinese_word = len(token) > 1 and is_chinese(token)
if chinese_word:
word_set.add(token)
word_list = list(word_set)
return word_list
def add_sub_symbol(bert_tokens: List[str], chinese_word_set: set()):
if not chinese_word_set:
return bert_tokens
max_word_len = max([len(w) for w in chinese_word_set])
bert_word = bert_tokens
start, end = 0, len(bert_word)
while start < end:
single_word = True
if is_chinese(bert_word[start]):
l = min(end - start, max_word_len)
for i in range(l, 1, -1):
whole_word = "".join(bert_word[start : start + i])
if whole_word in chinese_word_set:
for j in range(start + 1, start + i):
bert_word[j] = "##" + bert_word[j]
start = start + i
single_word = False
break
if single_word:
start += 1
return bert_word
def prepare_ref(lines: List[str], ltp_tokenizer: LTP, bert_tokenizer: BertTokenizer):
ltp_res = []
for i in range(0, len(lines), 100):
res = ltp_tokenizer.pipeline(lines[i : i + 100], tasks=["cws"]).cws
res = [get_chinese_word(r) for r in res]
ltp_res.extend(res)
assert len(ltp_res) == len(lines)
bert_res = []
for i in range(0, len(lines), 100):
res = bert_tokenizer(lines[i : i + 100], add_special_tokens=True, truncation=True, max_length=512)
bert_res.extend(res["input_ids"])
assert len(bert_res) == len(lines)
ref_ids = []
for input_ids, chinese_word in zip(bert_res, ltp_res):
input_tokens = []
for id in input_ids:
token = bert_tokenizer._convert_id_to_token(id)
input_tokens.append(token)
input_tokens = add_sub_symbol(input_tokens, chinese_word)
ref_id = []
# We only save pos of chinese subwords start with ##, which mean is part of a whole word.
for i, token in enumerate(input_tokens):
if token[:2] == "##":
clean_token = token[2:]
# save chinese tokens' pos
if len(clean_token) == 1 and _is_chinese_char(ord(clean_token)):
ref_id.append(i)
ref_ids.append(ref_id)
assert len(ref_ids) == len(bert_res)
return ref_ids
def main(args):
# For Chinese (Ro)Bert, the best result is from : RoBERTa-wwm-ext (https://github.com/ymcui/Chinese-BERT-wwm)
# If we want to fine-tune these model, we have to use same tokenizer : LTP (https://github.com/HIT-SCIR/ltp)
with open(args.file_name, "r", encoding="utf-8") as f:
data = f.readlines()
data = [line.strip() for line in data if len(line) > 0 and not line.isspace()] # avoid delimiter like '\u2029'
ltp_tokenizer = LTP(args.ltp) # faster in GPU device
bert_tokenizer = BertTokenizer.from_pretrained(args.bert)
ref_ids = prepare_ref(data, ltp_tokenizer, bert_tokenizer)
with open(args.save_path, "w", encoding="utf-8") as f:
data = [json.dumps(ref) + "\n" for ref in ref_ids]
f.writelines(data)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="prepare_chinese_ref")
parser.add_argument(
"--file_name",
required=False,
type=str,
default="./resources/chinese-demo.txt",
help="file need process, same as training data in lm",
)
parser.add_argument(
"--ltp",
required=False,
type=str,
default="./resources/ltp",
help="resources for LTP tokenizer, usually a path",
)
parser.add_argument(
"--bert",
required=False,
type=str,
default="./resources/robert",
help="resources for Bert tokenizer",
)
parser.add_argument(
"--save_path",
required=False,
type=str,
default="./resources/ref.txt",
help="path to save res",
)
args = parser.parse_args()
main(args)
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/mlm_wwm/run_mlm_wwm.py
|
# coding=utf-8
# Copyright 2020 The HuggingFace Team 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.
"""
Fine-tuning the library models for masked language modeling (BERT, ALBERT, RoBERTa...) with whole word masking on a
text file or a dataset.
Here is the full list of checkpoints on the hub that can be fine-tuned by this script:
https://huggingface.co/models?filter=fill-mask
"""
# You can also adapt this script on your own masked language modeling task. Pointers for this are left as comments.
import json
import logging
import math
import os
import sys
from dataclasses import dataclass, field
from typing import Optional
from datasets import Dataset, load_dataset
import transformers
from transformers import (
CONFIG_MAPPING,
MODEL_FOR_MASKED_LM_MAPPING,
AutoConfig,
AutoModelForMaskedLM,
AutoTokenizer,
DataCollatorForWholeWordMask,
HfArgumentParser,
Trainer,
TrainingArguments,
set_seed,
)
from transformers.trainer_utils import get_last_checkpoint, is_main_process
logger = logging.getLogger(__name__)
MODEL_CONFIG_CLASSES = list(MODEL_FOR_MASKED_LM_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.
"""
model_name_or_path: Optional[str] = field(
default=None,
metadata={
"help": (
"The model checkpoint for weights initialization. Don't set if you want to train a model from scratch."
)
},
)
model_type: Optional[str] = field(
default=None,
metadata={"help": "If training from scratch, pass a model type from the list: " + ", ".join(MODEL_TYPES)},
)
config_overrides: Optional[str] = field(
default=None,
metadata={
"help": (
"Override some existing default config settings when a model is trained from scratch. Example: "
"n_embd=10,resid_pdrop=0.2,scale_attn_weights=false,summary_type=cls_index"
)
},
)
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 `huggingface-cli login` (necessary to use this script "
"with private models)."
)
},
)
def __post_init__(self):
if self.config_overrides is not None and (self.config_name is not None or self.model_name_or_path is not None):
raise ValueError(
"--config_overrides can't be used in combination with --config_name or --model_name_or_path"
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
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)."}
)
train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a text file)."})
validation_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input evaluation data file to evaluate the perplexity on (a text file)."},
)
train_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input train ref data file for whole word masking in Chinese."},
)
validation_ref_file: Optional[str] = field(
default=None,
metadata={"help": "An optional input validation ref data file for whole word masking in Chinese."},
)
overwrite_cache: bool = field(
default=False, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
validation_split_percentage: Optional[int] = field(
default=5,
metadata={
"help": "The percentage of the train set used as validation set in case there's no validation split"
},
)
max_seq_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated. Default to the max input length of the model."
)
},
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
mlm_probability: float = field(
default=0.15, metadata={"help": "Ratio of tokens to mask for masked language modeling loss"}
)
pad_to_max_length: bool = field(
default=False,
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."
)
},
)
def __post_init__(self):
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`train_file` should be a csv, a json or a txt file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json", "txt"], "`validation_file` should be a csv, a json or a txt file."
def add_chinese_references(dataset, ref_file):
with open(ref_file, "r", encoding="utf-8") as f:
refs = [json.loads(line) for line in f.read().splitlines() if (len(line) > 0 and not line.isspace())]
assert len(dataset) == len(refs)
dataset_dict = {c: dataset[c] for c in dataset.column_names}
dataset_dict["chinese_ref"] = refs
return Dataset.from_dict(dataset_dict)
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()
# 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:
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."
)
# 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)],
)
logger.setLevel(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 if no column called
# 'text' is found. 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.
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
config_kwargs = {
"cache_dir": model_args.cache_dir,
"revision": model_args.model_revision,
"use_auth_token": True if model_args.use_auth_token else None,
}
if model_args.config_name:
config = AutoConfig.from_pretrained(model_args.config_name, **config_kwargs)
elif model_args.model_name_or_path:
config = AutoConfig.from_pretrained(model_args.model_name_or_path, **config_kwargs)
else:
config = CONFIG_MAPPING[model_args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
if model_args.config_overrides is not None:
logger.info(f"Overriding config: {model_args.config_overrides}")
config.update_from_string(model_args.config_overrides)
logger.info(f"New config: {config}")
tokenizer_kwargs = {
"cache_dir": model_args.cache_dir,
"use_fast": model_args.use_fast_tokenizer,
"revision": model_args.model_revision,
"use_auth_token": True if model_args.use_auth_token else None,
}
if model_args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(model_args.tokenizer_name, **tokenizer_kwargs)
elif model_args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path, **tokenizer_kwargs)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. This is not supported by this script. "
"You can do it from another script, save it, and load it from here, using --tokenizer_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,
revision=model_args.model_revision,
token=True if model_args.use_auth_token else None,
)
else:
logger.info("Training new model from scratch")
model = AutoModelForMaskedLM.from_config(config)
model.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]
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)
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,
)
# Add the chinese references if provided
if data_args.train_ref_file is not None:
tokenized_datasets["train"] = add_chinese_references(tokenized_datasets["train"], data_args.train_ref_file)
if data_args.validation_ref_file is not None:
tokenized_datasets["validation"] = add_chinese_references(
tokenized_datasets["validation"], data_args.validation_ref_file
)
# If we have ref files, need to avoid it removed by trainer
has_ref = data_args.train_ref_file or data_args.validation_ref_file
if has_ref:
training_args.remove_unused_columns = False
# Data collator
# This one will take care of randomly masking the tokens.
data_collator = DataCollatorForWholeWordMask(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:
if last_checkpoint is not None:
checkpoint = last_checkpoint
elif model_args.model_name_or_path is not None and os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
output_train_file = os.path.join(training_args.output_dir, "train_results.txt")
if trainer.is_world_process_zero():
with open(output_train_file, "w") as writer:
logger.info("***** Train results *****")
for key, value in sorted(train_result.metrics.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
# Need to save the state, since Trainer.save_model saves only the tokenizer with the model
trainer.state.save_to_json(os.path.join(training_args.output_dir, "trainer_state.json"))
# 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_wwm.txt")
if trainer.is_world_process_zero():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key, value in sorted(results.items()):
logger.info(f" {key} = {value}")
writer.write(f"{key} = {value}\n")
return results
def _mp_fn(index):
# For xla_spawn (TPUs)
main()
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/mlm_wwm/requirements.txt
|
datasets >= 1.1.3
sentencepiece != 0.1.92
protobuf
ltp
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/information-gain-filtration/README.md
|
# Information Gain Filtration(IGF)
Authors @Tuko @mraunak
This folder contains the code how to implement IGF for finetuning on GPT-2.
## What is IGF?
Here we present a general fine-tuning method that we call information gain filtration for improving the overall training efficiency and final
performance of language model fine-tuning(see paper below). The method is an alternative fine-tuning method that trains
a secondary model (e.g., a simple convolutional network) to predict the amount of information
gained over a given pre-trained model. The secondary model is lightweight and trained to
predict the Information Gain measure. Information Gain is defined as the change in a loss
function for a model before and after an SGD update with a sample (Equation X in the paper).
A small subset of the training set named the “objective” set, is used to measure information
gain on the pre-trained model, and consequently to train the secondary model. After
training, the model is used for filtering samples for the fine-tuning process. Therefore,
a high information gain value would suggest a sample is informative, whereas a low value
would suggest a non-informative sample that should be filtered out. Thus, a thresholding
strategy is defined to select informative samples. With such a strategy, samples are filtered
and once enough samples are selected to form a mini-batch and a usual fine-tuning/optimization
step is applied. The filtration process is repeated until the fine-tuning process is over.
Paper [Selecting Informative Contexts Improves Language Model Finetuning](https://arxiv.org/abs/2005.00175)
# Results
Several experiments were conducted to show the robustness of the IGF method versus the
standard fine-tuning process. For example, we achieve a median perplexity of 54.0 on the
Books dataset compared to 57.3 for standard fine-tuning on GPT-2 Small. The code was
implemented using the Transformers library and Pytorch. While the method may seem more
expensive, we saw enough evidence that it may lead to a performance benefit in the final models.
Figure 1: Comparing IGF to Standard Fine-tuning:
IGF with constant (p < 10−3 , t-test) and shifting(p < 10−6 , t-test) thresholding significantly outperform standard fine-tuning. The left-hand figure shows
test-set perplexity after each fine-tuning batch, averaged over 50 runs (error bars denote ± one standard error). The right-hand figure shows the perplexity of each
method after 60 batches. IGF with shifting thresholding (red) clearly improves over standard batched fine-tuning with Adam
## How to use this project?
To fine-tune a transformer model with IGF on a language modeling task, use the following script:
- `model_name_or_path`: Path to pretrained model or model identifier from huggingface.co/models
- `data_file`: A jbl file containing tokenized data which can be split as objective dataset,
train_dataset and test_dataset
- `igf_data_file`: A jbl file containing the context and information gain pairs to train secondary learner.
- `context_len`: The maximum total input sequence length after tokenization. Sequences longer
than this will be truncated, sequences shorter will be padded.
- `size_objective_set`: Number of articles that are long enough to be used as our objective set"
- `min_len`: The minimum length of the article to be used as objective set
- `trim`: Truncate the example if it exceeds context length
- `eval_freq`: Secondary model evaluation can be triggered at eval_freq
- `max_steps`: To calculate training epochs
- `number`: The number of examples split to be used as objective_set/test_data
- `secondary_learner_batch_size`: The batch size of training data for secondary learner
- `secondary_learner_max_epochs`: The number of epochs to train secondary learner
- `recopy_model`: Reset the model to the original pretrained GPT-2 weights after each iteration
- `eval_interval`: Decay the selectivity of our secondary learner filter from"
1 standard deviation above average to 1 below average after eval_interval(10) batches"
```python
python run_clm_igf.py\
--model_name_or_path "gpt2" \
--data_file="data/tokenized_stories_train_wikitext103" \
--igf_data_file="data/IGF_values" \
--context_len 32 \
--size_objective_set 100 \
--min_len 1026 \
--trim True \
--eval_freq 100 \
--max_steps 1000 \
--secondary_learner_batch_size 128 \
--secondary_learner_max_epochs 15 \
--number 100 \
--recopy_model \
--eval_interval 10 \
```
## Citation
If you find the resource useful, please cite the following paper
```
@inproceedings{antonello-etal-2021-selecting,
title = "Selecting Informative Contexts Improves Language Model Fine-tuning",
author = "Antonello, Richard and Beckage, Nicole and Turek, Javier and Huth, Alexander",
booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.acl-long.87",
doi = "10.18653/v1/2021.acl-long.87",
pages = "1072--1085",
}
```
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/information-gain-filtration/run_clm_igf.py
|
# Copyright 2022 - Intel Corp. All rights reserved.
# Authors: Mayank Kumar Raunak, Javier Turek, Nicole Beckage
"""
Implementation of a new method for fine-tuning transformer models that we call
Information Gain Filtration 'IGF' on WikiText data set and compared the results
with the standard fine-tuning method
Steps followed in the code:
1) Generate a objective dataset of pairs (X, IG(X)). IG(X)--Informativeness of context 'X'.
Our IG (information gain) model is learning to predict the ‘informativeness’ of a particular
context. Informativeness is the change in metric between the model’s accuracy on an
objective set before and after seeing that context. For casual language modeling, the
metric is perplexity.
2) A secondary learner is trained to infer a function approximation for IG using the dataset
created in (1).
3) The learner created in (2) is used to inform the fine-tuning process and filter out low informative samples.
Last, a plot is generated to compare the performance of IGF to standard fine-tuning without any filtering
"""
# Prerequisite libraries:
import argparse
import random
import joblib
import numpy as np
import torch
from igf.igf import (
SecondaryLearner,
collect_objective_set,
compute_perplexity,
generate_datasets,
load_gpt2,
recopy_gpt2,
set_seed,
train_secondary_learner,
)
from torch.utils.data import DataLoader, RandomSampler
from transformers import GPT2LMHeadModel
def generate_n_pairs(
context_len=32,
max_steps=10,
size_objective_set=100,
min_len=1026,
trim=True,
data_file="data/tokenized_stories_train_wikitext103.jbl",
igf_data_file="igf_context_pairs.jbl",
):
"""
Collecting *n* pairs for training the secondary learner
Args:
context_len: The maximum total input sequence length after tokenization. Sequences longer
than this will be truncated, sequences shorter will be padded
max_steps: To calculate training epochs of secondary learner
size_objective_set: size of objective data set used to create (X,IG(X)) pairs which is the training data for secondary learner
min_len: The minimum length of the article to be used as objective set
trim: If True truncate the context if it exceeds context length
data_file: Tokenized data set split for training and evaluation of model
igf_data_file: file to store (I,IG(X)) paired data set to train secondary learner
Returns:
Data stored in igf_data_file
"""
# generates same data everytime
set_seed(3)
# generate train_data and objective_set
train_data, objective_set = generate_datasets(
context_len, data_file, number=size_objective_set, min_len=1026, trim=True
)
# keeps model same across runs
set_seed(4)
# model, lm_optimizer, lm_scheduler = recopy_gpt2(model, device, max_steps) # store original model weights
# can we train on GPU?
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# load pretrained model
model = load_gpt2("gpt2").to(device)
print("computing perplexity on objective set")
orig_perp = compute_perplexity(model, objective_set, context_len).item()
print("perplexity on objective set:", orig_perp)
# collect igf pairs and save to file demo.jbl
collect_objective_set(model, orig_perp, context_len, train_data, objective_set, max_steps, device, igf_data_file)
# clean up, delete model and data we don't need anymore
del model, train_data, objective_set
torch.cuda.empty_cache()
def training_secondary_learner(
secondary_learner_train_data,
secondary_learner_max_epochs=15,
secondary_learner_batch_size=128,
eval_freq=100,
igf_model_path="igf_model.pt",
):
"""
Train the secondary learner
Args:
secondary_learner_train_data: Data set with (X,IG(X)) pairs to train secondary learner where IG(X) - measure of informativeness and X- context
secondary_learner_max_epochs: Number of epochs to train secondary learner
secondary_learner_batch_size: Batch size to train secondary learner
eval_freq (object): secondary model evaluation can be triggered at eval_freq
igf_model_path: path to store trained secondary learner
Returns:
Trained secondary learner
"""
set_seed(42)
# Load pre-trained model
model = GPT2LMHeadModel.from_pretrained("gpt2")
# Initialize secondary learner to use embedding weights of model
secondary_learner = SecondaryLearner(model)
# Train secondary learner
secondary_learner = train_secondary_learner(
secondary_learner,
secondary_learner_train_data,
max_epochs=secondary_learner_max_epochs,
batch_size=secondary_learner_batch_size,
eval_freq=100,
igf_model_path=igf_model_path,
)
del model, secondary_learner_train_data
torch.cuda.empty_cache()
return secondary_learner
def finetune(
model,
train_dataset,
test_dataset,
context_len=32,
max_steps=1000,
batch_size=16,
threshold=1.0,
recopy_model=recopy_gpt2,
secondary_learner=None,
eval_interval=10,
finetuned_model_name="gpt2_finetuned.pt",
):
"""
fine-tune with IGF if secondary_learner is not None, else standard fine-tuning
Args:
model: pre-trained GPT-2 model
train_dataset: Data set to train GPT-2 model
test_dataset: Evaluate GPT-2 model
context_len: The maximum total input sequence length after tokenization. Sequences longer
than this will be truncated, sequences shorter will be padded
max_steps: To calculate training epochs
batch_size: Batch size to train GPT-2 model
threshold: The threshold value used by secondary learner to filter the train_data and allow only"
informative data as input to the model
recopy_model: Reset the model to the original pretrained GPT-2 weights after each iteration
secondary_learner: Selection of IGF as fine-tuning method if not None
eval_interval: number of batches after which decay the selectivity of our secondary learner filter from
1 standard deviation above average to 1 below average
fine-tuned_model_name: name of the final final-tuned GPT-2 model
Returns:
Fine-tuned GPT-2 model
"""
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
train_sampler = RandomSampler(train_dataset)
train_dataloader = DataLoader(train_dataset, sampler=train_sampler)
num_train_epochs = max_steps // (len(train_dataset)) + 1
global_step = 0
context = torch.zeros((1, context_len), dtype=torch.long, device=device)
model, lm_optimizer, lm_scheduler = recopy_model(model, device, max_steps)
model.train()
if secondary_learner is not None:
secondary_learner.to(device)
secondary_learner.eval()
contexts = []
examples = 0
observed_qs = []
test_perps = []
# Compute the performance of the transformer model at the beginning
real_perp = compute_perplexity(model, test_dataset, context_len)
test_perps.append(real_perp)
print("Test perplexity, step", global_step, ":", real_perp)
for epoch in range(int(num_train_epochs)):
for step, example in enumerate(train_dataloader):
torch.cuda.empty_cache()
start = random.randint(0, example.size(2) - context_len - 1)
context[0, :] = example[0, 0, start : start + context_len]
lm_optimizer.zero_grad()
outputs = model(context, labels=context)
do_backprop = True
if secondary_learner is not None:
predicted_q = secondary_learner.forward(
torch.tensor(context, dtype=torch.long, device=device).unsqueeze(0)
)[0].item()
observed_qs.append(float(predicted_q))
# Here we implement the simple non-constant threshold for the predicted IG(X) value
# We will decay the selectivity of our secondary learner filter from
# 1 standard deviation above average to 1 below average after 10 batches.
if global_step == 10:
threshold = -1
if predicted_q < threshold:
do_backprop = False
# If we passed the filter, add the context to the batch!
if do_backprop:
contexts.append(np.array(context.cpu()))
lm_loss = outputs[0]
lm_loss.backward()
examples += 1
del outputs
# Once the batch is filled with enough contexts, backprop on the batch.
if examples == batch_size:
torch.cuda.empty_cache()
examples = 0
# Do LM backprop
torch.nn.utils.clip_grad_norm_(model.parameters(), 3.0)
lm_optimizer.step()
lm_scheduler.step() # Update learning rate schedule
global_step += 1
# Compute the performance of the transformer model at this batch
if global_step % eval_interval == 0:
real_perp = compute_perplexity(model, test_dataset, context_len)
test_perps.append(real_perp)
print("Test perplexity, step", global_step, ":", real_perp)
# Break out of the loop after 60 batches
if max_steps > 0 and global_step > 60:
break
if max_steps > 0 and global_step > 60:
break
# save finetuned transformer model
torch.save(model.state_dict(), finetuned_model_name)
torch.cuda.empty_cache()
# Do some cleaning up so we can reinitialize for the next run of this function
del lm_optimizer
del lm_scheduler
return model
def main():
parser = argparse.ArgumentParser(description="Fine-tune a transformer model with IGF on a language modeling task")
# Required parameters
parser.add_argument(
"--data_dir",
default=None,
type=str,
required=True,
help="The input data dir. Should contain data files for WikiText.",
)
parser.add_argument(
"--model_name_or_path",
default=None,
type=str,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models",
)
parser.add_argument(
"--data_file",
type=str,
default=None,
help=(
"A jbl file containing tokenized data which can be split as objective dataset, "
"train_dataset and test_dataset."
),
)
parser.add_argument(
"--igf_data_file",
type=str,
default=None,
help="A jbl file containing the context and information gain pairs to train secondary learner.",
)
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help="The output directory where the final fine-tuned model is stored.",
)
parser.add_argument(
"--tokenizer_name",
default=None,
type=str,
help="Pretrained tokenizer name or path if not the same as model_name",
)
parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
parser.add_argument(
"--context_len",
default=32,
type=int,
help=(
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
),
)
parser.add_argument(
"--size_objective_set",
default=100,
type=int,
help="number of articles that are long enough to be used as our objective set",
)
parser.add_argument(
"--eval_freq", default=100, type=int, help="secondary model evaluation is triggered at eval_freq"
)
parser.add_argument("--max_steps", default=1000, type=int, help="To calculate training epochs")
parser.add_argument(
"--secondary_learner_batch_size",
default=128,
type=int,
help="batch size of training data for secondary learner",
)
parser.add_argument(
"--batch_size", default=16, type=int, help="batch size of training data of language model(gpt2) "
)
parser.add_argument(
"--eval_interval",
default=10,
type=int,
help=(
"decay the selectivity of our secondary learner filter from "
"1 standard deviation above average to 1 below average after 10 batches"
),
)
parser.add_argument(
"--number", default=100, type=int, help="The number of examples split to be used as objective_set/test_data"
)
parser.add_argument(
"--min_len", default=1026, type=int, help="The minimum length of the article to be used as objective set"
)
parser.add_argument(
"--secondary_learner_max_epochs", default=15, type=int, help="number of epochs to train secondary learner"
)
parser.add_argument("--trim", default=True, type=bool, help="truncate the example if it exceeds context length")
parser.add_argument(
"--threshold",
default=1.0,
type=float,
help=(
"The threshold value used by secondary learner to filter the train_data and allow only"
" informative data as input to the model"
),
)
parser.add_argument("--finetuned_model_name", default="gpt2_finetuned.pt", type=str, help="finetuned_model_name")
parser.add_argument(
"--recopy_model",
default=recopy_gpt2,
type=str,
help="Reset the model to the original pretrained GPT-2 weights after each iteration",
)
# function calls
# Collecting *n* pairs of context and information gain(X, IG(X)) for training the secondary learner
generate_n_pairs(
context_len=32,
max_steps=10,
size_objective_set=100,
min_len=1026,
trim=True,
data_file="data/tokenized_stories_train_wikitext103.jbl",
igf_data_file="igf_context_pairs.jbl",
)
# Load train data for secondary learner
secondary_learner_train_data = joblib.load("data/IGF_values.jbl")
# Train secondary learner
secondary_learner = training_secondary_learner(
secondary_learner_train_data,
secondary_learner_max_epochs=15,
secondary_learner_batch_size=128,
eval_freq=100,
igf_model_path="igf_model.pt",
)
# load pretrained gpt2 model
model = GPT2LMHeadModel.from_pretrained("gpt2")
set_seed(42)
# Generate train and test data to train and evaluate gpt2 model
train_dataset, test_dataset = generate_datasets(
context_len=32, file="data/tokenized_stories_train_wikitext103.jbl", number=100, min_len=1026, trim=True
)
# fine-tuning of the gpt2 model using igf (Information Gain Filtration)
finetune(
model,
train_dataset,
test_dataset,
context_len=32,
max_steps=1000,
batch_size=16,
threshold=1.0,
recopy_model=recopy_gpt2,
secondary_learner=secondary_learner,
eval_interval=10,
finetuned_model_name="gpt2_finetuned.pt",
)
if __name__ == "__main__":
main()
| 0
|
hf_public_repos/transformers/examples/research_projects
|
hf_public_repos/transformers/examples/research_projects/information-gain-filtration/requirements.txt
|
matplotlib
numpy>=1.17.2
joblib>=0.13.2
scipy
torch>=1.10.1
transformers>=3.5
| 0
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.