import hashlib
import os
import uuid
from collections import defaultdict
from copy import copy
from pathlib import Path
import tensorflow as tf
import tensorflow_ranking as tfr
import numpy as np
from tensorflow.python.keras import backend as K
from tqdm import tqdm
from capreolus.searcher import Searcher
from capreolus import ConfigOption, evaluator
from capreolus.trainer import Trainer
from capreolus.utils.loginit import get_logger
from capreolus.reranker.common import (
TFPairwiseHingeLoss,
TFCategoricalCrossEntropyLoss,
KerasPairModel,
KerasTripletModel,
KerasLCEModel,
TFLCELoss,
)
import tensorflow.keras.mixed_precision as mixed_precision
[docs]logger = get_logger(__name__)
from tensorflow.python.client import device_lib
[docs]def get_available_gpus():
local_device_protos = device_lib.list_local_devices()
return [x.name for x in local_device_protos if x.device_type == "GPU"]
@Trainer.register
[docs]class TensorflowTrainer(Trainer):
"""
Trains (optionally) on the TPU.
Uses two optimizers with different learning rates - one for the BERT layers and another for the classifier layers.
Configurable warmup and decay for bertlr.
WARNING: The optimizers depend on specific layer names (see train()) - if your reranker does not have layers with
'bert' in the name, the normal learning rate will be applied to it instead of the value supplied through the
bertlr ConfigOption
"""
[docs] module_name = "tensorflow"
[docs] config_spec = [
ConfigOption("batch", 32, "batch size"),
ConfigOption("evalbatch", 0, "batch size at inference time (or 0 to use training batch size)"),
ConfigOption("niters", 20, "number of iterations to train for"),
ConfigOption("itersize", 512, "number of training instances in one iteration"),
ConfigOption("bertlr", 2e-5, "learning rate for bert parameters"),
ConfigOption("lr", 0.001, "learning rate"),
ConfigOption("warmupiters", 0),
ConfigOption("loss", "pairwise_hinge_loss", "must be one of tfr.losses.RankingLossKey"),
ConfigOption("fastforward", False),
ConfigOption("validatefreq", 1),
ConfigOption("boardname", "default"),
ConfigOption("usecache", False),
ConfigOption("tpuname", None),
ConfigOption("tpuzone", None),
ConfigOption("storage", None),
ConfigOption("eager", False),
ConfigOption("decay", 0.0, "learning rate decay"),
ConfigOption("decayiters", 3),
ConfigOption("decaytype", None),
ConfigOption("amp", False, "use automatic mixed precision"),
]
[docs] config_keys_not_in_path = ["fastforward", "boardname", "usecache", "tpuname", "tpuzone", "storage"]
[docs] def build(self):
tf.random.set_seed(self.config["seed"])
self.evalbatch = self.config["evalbatch"] if self.config["evalbatch"] > 0 else self.config["batch"]
if os.environ.get("CAPR_QUEUE_SKIP_TPU") == "YES":
return
# Use TPU if available, otherwise resort to GPU/CPU
if self.config["tpuname"]:
if self.config["tpuname"] == "LOCAL":
logger.debug("using TPU VM")
self.tpu = tf.distribute.cluster_resolver.TPUClusterResolver(tpu="local")
else:
logger.debug("using Cloud TPU")
self.tpu = tf.distribute.cluster_resolver.TPUClusterResolver(
tpu=self.config["tpuname"], zone=self.config["tpuzone"]
)
else:
self.tpu = None
# TPUStrategy for distributed training
if self.tpu:
logger.info("Utilizing TPUs")
tf.config.experimental_connect_to_cluster(self.tpu)
tf.tpu.experimental.initialize_tpu_system(self.tpu)
self.strategy = tf.distribute.experimental.TPUStrategy(self.tpu)
elif len(get_available_gpus()) > 1:
self.strategy = tf.distribute.MirroredStrategy()
else: # default strategy that works on CPU and single GPU
self.strategy = tf.distribute.get_strategy()
self.amp = self.config["amp"]
if self.amp:
policy = mixed_precision.Policy("mixed_bfloat16" if self.tpu else "mixed_float16")
mixed_precision.set_policy(policy)
# Defining some props that we will later initialize
self.validate()
[docs] def validate(self):
if self.tpu and not all([self.config["storage"], self.config["tpuname"], self.config["tpuzone"]]):
raise ValueError("storage, tpuname and tpuzone configs must be provided when training on TPU")
if self.tpu and self.config["tpuname"] != "LOCAL" and not self.config["storage"].startswith("gs://"):
raise ValueError("For TPU utilization, the storage config should start with 'gs://'")
[docs] def train(self, reranker, train_dataset, train_output_path, dev_data, dev_output_path, qrels, metric, relevance_level=1):
if self.tpu:
# WARNING: not sure if pathlib is compatible with gs://
train_output_path = Path(
"{0}/{1}/{2}".format(
self.config["storage"], "train_output", hashlib.md5(str(train_output_path).encode("utf-8")).hexdigest()
)
)
dev_best_weight_fn, weights_output_path, info_output_path, loss_fn, metric_fn = self.get_paths_for_early_stopping(
train_output_path, dev_output_path
)
train_records = self.get_tf_train_records(reranker, train_dataset)
dev_records = self.get_tf_dev_records(reranker, dev_data)
dev_dist_dataset = self.strategy.experimental_distribute_dataset(dev_records)
# Does not very much from https://www.tensorflow.org/tutorials/distribute/custom_training
strategy_scope = self.strategy.scope()
with strategy_scope:
reranker.build_model()
wrapped_model = self.get_wrapped_model(reranker.model)
loss_object = self.get_loss(self.config["loss"])
optimizer_1 = tf.keras.optimizers.Adam(learning_rate=self.config["lr"])
optimizer_2 = tf.keras.optimizers.Adam(learning_rate=self.config["bertlr"])
# "You should remove the use of the LossScaleOptimizer when TPUs are used."
if self.amp and not self.tpu:
optimizer_2 = mixed_precision.LossScaleOptimizer(optimizer_2, loss_scale="dynamic")
def compute_loss(labels, predictions):
per_example_loss = loss_object(labels, predictions)
return tf.nn.compute_average_loss(per_example_loss, global_batch_size=self.config["batch"])
def is_bert_variable(name):
if "bert" in name:
return True
if "electra" in name:
return True
return False
def train_step(inputs):
data, labels = inputs
with tf.GradientTape() as tape:
train_predictions = wrapped_model(data, training=True)
loss = compute_loss(labels, train_predictions)
if self.amp and not self.tpu:
loss = optimizer_2.get_scaled_loss(loss)
gradients = tape.gradient(loss, wrapped_model.trainable_variables)
if self.amp and not self.tpu:
optimizer_2.get_unscaled_gradients(gradients)
bert_variables = [
(gradients[i], variable)
for i, variable in enumerate(wrapped_model.trainable_variables)
if is_bert_variable(variable.name) and "classifier" not in variable.name
]
classifier_vars = [
(gradients[i], variable)
for i, variable in enumerate(wrapped_model.trainable_variables)
if "classifier" in variable.name
]
other_vars = [
(gradients[i], variable)
for i, variable in enumerate(wrapped_model.trainable_variables)
if not is_bert_variable(variable.name) and "classifier" not in variable.name
]
assert len(bert_variables) + len(classifier_vars) + len(other_vars) == len(wrapped_model.trainable_variables)
# TODO: Clean this up for general use
# Making sure that we did not miss any variables
optimizer_1.apply_gradients(classifier_vars)
optimizer_2.apply_gradients(bert_variables)
if other_vars:
optimizer_1.apply_gradients(other_vars)
return loss
def test_step(inputs):
data, labels = inputs
predictions = wrapped_model.predict_step(data)
return predictions
@tf.function
def distributed_train_step(dataset_inputs):
per_replica_losses = self.strategy.run(train_step, args=(dataset_inputs,))
return self.strategy.reduce(tf.distribute.ReduceOp.SUM, per_replica_losses, axis=None)
@tf.function
def distributed_test_step(dataset_inputs):
return self.strategy.run(test_step, args=(dataset_inputs,))
train_records = train_records.shuffle(100_000)
train_dist_dataset = self.strategy.experimental_distribute_dataset(train_records)
initial_iter, metrics = (
self.fastforward_training(wrapped_model, weights_output_path, loss_fn, metric_fn)
if self.config["fastforward"]
else (0, {})
)
dev_best_metric = metrics.get(metric, -np.inf)
logger.info("starting training from iteration %s/%s", initial_iter + 1, self.config["niters"])
logger.info(f"Best metric loaded: {metric}={dev_best_metric}")
best_trec_preds = {}
cur_step = initial_iter * self.n_batch_per_iter
initial_lr = self.change_lr(step=cur_step, lr=self.config["bertlr"])
K.set_value(optimizer_2.lr, K.get_value(initial_lr))
train_loss = self.load_loss_file(loss_fn) if initial_iter > 0 else []
if 0 < initial_iter < self.config["niters"]:
self.exhaust_used_train_data(train_dist_dataset, n_batch_to_exhaust=initial_iter * self.n_batch_per_iter)
niter = initial_iter
total_loss = 0
trec_preds = {}
iter_bar = tqdm(desc="Training iteration", total=self.n_batch_per_iter)
# Goes through the dataset ONCE (i.e niters * itersize).
# However, the dataset may already contain multiple instances of the same sample,
# depending upon what Sampler was used.
# If you want multiple epochs, achieve it by tweaking the niters and itersize values.
for x in train_dist_dataset:
total_loss += distributed_train_step(x)
cur_step += 1
iter_bar.update(1)
# Do warmup and decay
new_lr = self.change_lr(step=cur_step, lr=self.config["bertlr"])
K.set_value(optimizer_2.lr, K.get_value(new_lr))
if cur_step % self.n_batch_per_iter == 0:
niter += 1
iter_bar.close()
iter_bar = tqdm(total=self.n_batch_per_iter)
train_loss.append(total_loss / self.n_batch_per_iter)
logger.info("iter={} loss = {}".format(niter, train_loss[-1]))
self.write_to_loss_file(loss_fn, train_loss)
total_loss = 0
if self.config["fastforward"]:
wrapped_model.save_weights(f"{weights_output_path}/{niter}")
if niter % self.config["validatefreq"] == 0:
dev_predictions = []
for x in tqdm(dev_dist_dataset, desc="validation"):
pred_batch = (
distributed_test_step(x).values
if self.strategy.num_replicas_in_sync > 1
else [distributed_test_step(x)]
)
for p in pred_batch:
dev_predictions.extend(p)
trec_preds = self.get_preds_in_trec_format(dev_predictions, dev_data)
metrics = evaluator.eval_runs(trec_preds, dict(qrels), evaluator.DEFAULT_METRICS, relevance_level)
logger.info("dev metrics: %s", " ".join([f"{metric}={v:0.3f}" for metric, v in sorted(metrics.items())]))
if metrics[metric] > dev_best_metric:
dev_best_metric = metrics[metric]
logger.info("new best dev metric: %0.4f", dev_best_metric)
self.write_to_metric_file(metric_fn, metrics)
wrapped_model.save_weights(dev_best_weight_fn)
Searcher.write_trec_run(trec_preds, outfn=(dev_output_path / "best").as_posix())
best_trec_preds = trec_preds
if cur_step >= self.config["niters"] * self.n_batch_per_iter:
break
return best_trec_preds
[docs] def predict(self, reranker, pred_data, pred_fn):
pred_records = self.get_tf_dev_records(reranker, pred_data)
pred_dist_dataset = self.strategy.experimental_distribute_dataset(pred_records)
strategy_scope = self.strategy.scope()
with strategy_scope:
wrapped_model = self.get_wrapped_model(reranker.model)
def test_step(inputs):
data, labels = inputs
predictions = wrapped_model.predict_step(data)
return predictions
@tf.function
def distributed_test_step(dataset_inputs):
return self.strategy.run(test_step, args=(dataset_inputs,))
predictions = []
for x in tqdm(pred_dist_dataset, desc="validation"):
pred_batch = distributed_test_step(x).values if self.strategy.num_replicas_in_sync > 1 else [distributed_test_step(x)]
for p in pred_batch:
predictions.extend(p)
trec_preds = self.get_preds_in_trec_format(predictions, pred_data)
os.makedirs(os.path.dirname(pred_fn), exist_ok=True)
Searcher.write_trec_run(trec_preds, pred_fn)
return trec_preds
[docs] def find_cached_tf_records(self, dataset, required_sample_count):
"""
Looks for a tf record for the passed dataset that has at least the specified number of samples
"""
parent_dir = (
"{0}/capreolus_tfrecords/".format(self.config["storage"]) if self.tpu else "{0}".format(self.get_cache_path())
)
if not tf.io.gfile.exists(parent_dir):
return None
else:
child_dirs = tf.io.gfile.listdir(parent_dir)
required_prefix = dataset.get_hash()
for child_dir in child_dirs:
child_dir_ending = child_dir.split("_")[-1][-1]
# The child dir will end with '/' if it's on gcloud, but not on local disk.
if child_dir_ending == "/":
sample_count = int(child_dir.split("_")[-1][:-1])
else:
sample_count = int(child_dir.split("_")[-1])
prefix = "_".join(child_dir.split("_")[:-1])
# TODO: Add checks to make sure that the child dir is not empty
if prefix == required_prefix and sample_count >= required_sample_count:
return "{0}/{1}".format(parent_dir.rstrip("/"), child_dir)
return None
[docs] def get_tf_train_records(self, reranker, dataset):
"""
1. Returns tf records from cache (disk) if applicable
2. Else, converts the dataset into tf records, writes them to disk, and returns them
"""
required_samples = self.config["niters"] * self.config["itersize"]
cached_tf_record_dir = self.find_cached_tf_records(dataset, required_samples)
if self.config["usecache"] and cached_tf_record_dir is not None:
filenames = tf.io.gfile.listdir(cached_tf_record_dir)
filenames = ["{0}/{1}".format(cached_tf_record_dir.rstrip("/"), name) for name in filenames]
else:
filenames = self.convert_to_tf_train_record(reranker, dataset)
return self.load_tf_train_records_from_file(reranker, filenames, self.config["batch"])
[docs] def load_tf_train_records_from_file(self, reranker, filenames, batch_size):
raw_dataset = tf.data.TFRecordDataset(filenames)
tf_records_dataset = raw_dataset.batch(batch_size, drop_remainder=True).map(
reranker.extractor.parse_tf_train_example, num_parallel_calls=tf.data.experimental.AUTOTUNE
)
return tf_records_dataset
[docs] def convert_to_tf_train_record(self, reranker, dataset):
"""
Tensorflow works better if the input data is fed in as tfrecords
Takes in a dataset, iterates through it, and creates multiple tf records from it.
Creates exactly niters * itersize samples.
The exact structure of the tfrecords is defined by reranker.extractor. For example, see BertPassage.get_tf_train_feature()
params:
reranker - A capreolus.reranker.Reranker instance
dataset - A capreolus.sampler.Sampler instance
"""
dir_name = self.form_tf_record_cache_path(dataset)
tf_features = []
tf_record_filenames = []
required_sample_count = self.config["niters"] * self.config["itersize"]
sample_count = 0
iter_bar = tqdm(total=required_sample_count)
for sample in dataset:
tf_features.extend(reranker.extractor.create_tf_train_feature(sample))
if len(tf_features) > 20000:
tf_record_filenames.append(self.write_tf_record_to_file(dir_name, tf_features))
tf_features = []
iter_bar.update(1)
sample_count += 1
if sample_count >= required_sample_count:
break
iter_bar.close()
assert sample_count == required_sample_count, "dataset generator ran out before generating enough samples"
if len(tf_features):
tf_record_filenames.append(self.write_tf_record_to_file(dir_name, tf_features))
return tf_record_filenames
[docs] def get_tf_dev_records(self, reranker, dataset):
"""
1. Returns tf records from cache (disk) if applicable
2. Else, converts the dataset into tf records, writes them to disk, and returns them
"""
cached_tf_record_dir = self.form_tf_record_cache_path(dataset)
if self.config["usecache"] and tf.io.gfile.exists(cached_tf_record_dir):
filenames = sorted(tf.io.gfile.listdir(cached_tf_record_dir), key=lambda x: int(x.replace(".tfrecord", "")))
filenames = ["{0}/{1}".format(cached_tf_record_dir, name) for name in filenames]
else:
filenames = self.convert_to_tf_dev_record(reranker, dataset)
return self.load_tf_dev_records_from_file(reranker, filenames, self.evalbatch)
[docs] def load_tf_dev_records_from_file(self, reranker, filenames, batch_size):
raw_dataset = tf.data.TFRecordDataset(filenames)
tf_records_dataset = raw_dataset.batch(batch_size, drop_remainder=True).map(
reranker.extractor.parse_tf_dev_example, num_parallel_calls=tf.data.experimental.AUTOTUNE
)
return tf_records_dataset
[docs] def convert_to_tf_dev_record(self, reranker, dataset):
evalbatch = self.evalbatch
dir_name = self.form_tf_record_cache_path(dataset)
tf_features = []
tf_record_filenames = []
tf_file_id = 0
element_to_copy = None
for sample in dataset:
tf_features.extend(reranker.extractor.create_tf_dev_feature(sample))
if element_to_copy is None:
element_to_copy = tf_features[0]
if len(tf_features) > 20000:
tf_record_filenames.append(self.write_tf_record_to_file(dir_name, tf_features, file_name=str(tf_file_id)))
tf_features = []
tf_file_id += 1
# TPU's require drop_remainder = True. But we cannot drop things from validation dataset
# As a workaroud, we pad the dataset with the last sample until it reaches the batch size.
for i in range(evalbatch):
tf_features.append(copy(element_to_copy))
tf_record_filenames.append(self.write_tf_record_to_file(dir_name, tf_features, file_name=str(tf_file_id)))
return tf_record_filenames
[docs] def write_tf_record_to_file(self, dir_name, tf_features, file_name=None):
"""
Actually write the tf record to file. The destination can also be a gcs bucket.
TODO: Use generators to optimize memory usage
"""
file_name = str(uuid.uuid4()) if not file_name else file_name
filename = "{0}/{1}.tfrecord".format(dir_name, file_name)
examples = [tf.train.Example(features=tf.train.Features(feature=feature)) for feature in tf_features]
if not os.path.isdir(dir_name):
os.makedirs(dir_name, exist_ok=True)
examples = [example.SerializeToString() for example in examples]
with tf.io.TFRecordWriter(filename) as writer:
for example in examples:
writer.write(example)
logger.info("Wrote tf record file: {}".format(filename))
return str(filename)
@staticmethod
[docs] def get_loss(self, loss_name):
try:
if loss_name == "pairwise_hinge_loss":
loss = TFPairwiseHingeLoss(reduction=tf.keras.losses.Reduction.NONE)
elif loss_name == "crossentropy":
loss = TFCategoricalCrossEntropyLoss(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)
elif loss_name == "lce":
loss = TFLCELoss(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)
else:
loss = tfr.keras.losses.get(loss_name)
except ValueError:
loss = tf.keras.losses.get(loss_name)
return loss
[docs] def get_wrapped_model(self, model):
"""
We need a wrapped model because the logic changes slightly depending on whether the input is pointwise or pairwise:
1. In case of pointwise input, there's no "negative document" - so in this case we just have to execute the model's call() method
2. In case of pairwise input, we need to execute the model's call() method twice (one for positive doc and then again for negative doc), and then
stack the results together before passing to a pairwise loss function.
The alternative was to let the user manually configure everything, for example:
`loss=crossentropy reranker.trainer.input=pairwise ...` - we already have too many ConfigOptions :shrug:
"""
if self.config["loss"] == "crossentropy":
return KerasPairModel(model)
elif self.config["loss"] == "lce":
return KerasLCEModel(model)
return KerasTripletModel(model)
[docs] def fastforward_training(self, model, weights_path, loss_fn, best_metric_fn):
"""Skip to the last training iteration whose weights were saved.
If saved model and optimizer weights are available, this method will load those weights into model
and optimizer, and then return the next iteration to be run. For example, if weights are available for
iterations 0-10 (11 zero-indexed iterations), the weights from iteration index 10 will be loaded, and
this method will return 11.
If an error or inconsistency is encountered when checking for weights, this method returns 0.
This method checks several files to determine if weights "are available". First, loss_fn is read to
determine the last recorded iteration. (If a path is missing or loss_fn is malformed, 0 is returned.)
Second, the weights from the last recorded iteration in loss_fn are loaded into the model and optimizer.
If this is successful, the method returns `1 + last recorded iteration`. If not, it returns 0.
(We consider loss_fn because it is written at the end of every training iteration.)
Args:
model (Reranker): a PyTorch Reranker whose state should be loaded
weights_path (Path): directory containing model and optimizer weights
loss_fn (Path): file containing loss history
Returns:
int: the next training iteration after fastforwarding. If successful, this is > 0.
If no weights are available or they cannot be loaded, 0 is returned.
"""
default_return_values = (0, {})
if not (weights_path.exists() and loss_fn.exists()):
return default_return_values
try:
loss = self.load_loss_file(loss_fn)
metrics = self.load_metric(best_metric_fn)
except IOError:
return default_return_values
last_loss_iteration = len(loss) - 1
weights_fn = weights_path / f"{last_loss_iteration}"
try:
model.load_weights(weights_fn)
return last_loss_iteration + 1, metrics
except: # lgtm [py/catch-base-exception]
logger.info("attempted to load weights from %s but failed, starting at iteration 0", weights_fn)
return default_return_values
[docs] def load_best_model(self, reranker, train_output_path):
# TODO: Do the train_output_path modification at one place?
if self.tpu:
train_output_path = "{0}/{1}/{2}".format(
self.config["storage"], "train_output", hashlib.md5(str(train_output_path).encode("utf-8")).hexdigest()
)
reranker.build_model()
# Because the saved weights are that of a wrapped model.
wrapped_model = self.get_wrapped_model(reranker.model)
wrapped_model.load_weights("{0}/dev.best".format(train_output_path))
return wrapped_model.model