Created
October 14, 2019 12:51
-
-
Save tlkh/d252abcb3a5b59a7b8c47660997fd390 to your computer and use it in GitHub Desktop.
tf_transformers
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import os | |
import tensorflow as tf | |
gpus = tf.config.experimental.list_physical_devices('GPU') | |
if gpus: | |
for gpu in gpus: | |
tf.config.experimental.set_memory_growth(gpu, True) | |
import tensorflow_datasets | |
from transformers import BertTokenizer, TFBertForSequenceClassification, glue_convert_examples_to_features#, BertForSequenceClassification | |
# script parameters | |
BATCH_SIZE = 8 | |
EVAL_BATCH_SIZE = BATCH_SIZE | |
USE_XLA = False | |
USE_AMP = False | |
tf.config.optimizer.set_jit(USE_XLA) | |
tf.config.optimizer.set_experimental_options({"auto_mixed_precision": USE_AMP}) | |
# Load tokenizer and model from pretrained model/vocabulary | |
tokenizer = BertTokenizer.from_pretrained('bert-base-cased') | |
model = TFBertForSequenceClassification.from_pretrained('bert-base-cased') | |
# Load dataset via TensorFlow Datasets | |
data, info = tensorflow_datasets.load('glue/mrpc', with_info=True) | |
train_examples = info.splits['train'].num_examples | |
valid_examples = info.splits['validation'].num_examples | |
# Prepare dataset for GLUE as a tf.data.Dataset instance | |
train_dataset = glue_convert_examples_to_features(data['train'], tokenizer, 512, 'mrpc') | |
valid_dataset = glue_convert_examples_to_features(data['validation'], tokenizer, 512, 'mrpc') | |
train_dataset = train_dataset.shuffle(128).batch(BATCH_SIZE).repeat(-1) | |
valid_dataset = valid_dataset.batch(EVAL_BATCH_SIZE) | |
# Prepare training: Compile tf.keras model with optimizer, loss and learning rate schedule | |
opt = tf.keras.optimizers.Adam(learning_rate=3e-5, epsilon=1e-08) | |
if USE_AMP: | |
# loss scaling is currently required when using mixed precision | |
opt = tf.keras.mixed_precision.experimental.LossScaleOptimizer(opt, 'dynamic') | |
loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True) | |
metric = tf.keras.metrics.SparseCategoricalAccuracy('accuracy') | |
model.compile(optimizer=opt, loss=loss, metrics=[metric]) | |
# Train and evaluate using tf.keras.Model.fit() | |
train_steps = train_examples//BATCH_SIZE | |
valid_steps = valid_examples//EVAL_BATCH_SIZE | |
history = model.fit(train_dataset, epochs=2, steps_per_epoch=train_steps, | |
validation_data=valid_dataset, validation_steps=valid_steps) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I just made a working example by loading dataset as
tf.data.Dataset
It seems
tf.data.Dataset.from_generator()
is more optimized compared to Keras generators....