This is a simple example of how to use Tensorflow's Estimator to train a model
on MNIST using the TFRecords data format. There are two scripts here:
convert_to_records.py, which will get the MNIST data and convert it to
TFRecords format; and trainer.py, which will train on MNIST.
Example usage:
convert_to_records.py --data_dir=$HOME/data/mnist
trainer.py \
--batch_size=128 \
--train_file_pattern=$HOME/data/mnist/train.tfrecords \
--eval_file_pattern=$HOME/data/mnist/test.tfrecords \
--eval_freq=100 \
--model_dir=/tmp/mnist \
--train_steps=10000 \
--save_summary_steps=10
--shuffle_buffer_size=60000 \
--learning_rate=0.0001 \
--mode=train_and_eval
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent