TensorFlow Serving API设置warmup

• 运行后附代码文件生成tf_serving_warmup_requests文件；
• 置其于模型保存文件夹内的assets.extra文件夹下；
• 重启docker API即可，可观察到屏幕输出显示warmp启动成功；

保存模型的文件结构于是乎成为：

.
├── assets
├── assets.extra
│   └── tf_serving_warmup_requests
├── saved_model.pb
└── variables
    ├── variables.data-00000-of-00001
    └── variables.index

生成文件的代码如下：

#  -*- coding: utf-8 -*-

"""
file: warmup.py
date: 2020-02-19 
---------------------------
注意用saved_model_cli show --dir /model/saved/path --all
查看模型的signatures信息
生成的tf_serving_warmup_requests 文件存放在模型保持文件的assets.extra文件夹下，重启docker API即可
"""
import tensorflow as tf
from tensorflow_serving.apis import predict_pb2
from tensorflow_serving.apis import prediction_log_pb2

IMAGE_PATH = './warmup.jpg'
NUM_RECORDS = 100

X_new = tf.io.read_file(IMAGE_PATH)
X_new = tf.image.decode_image(X_new)
X_new = tf.image.resize(X_new, (224,224))
X_new = tf.cast(X_new, tf.float32)/255.0

def main():
    """Generate TFRecords for warming up."""

    with tf.io.TFRecordWriter("tf_serving_warmup_requests") as writer:
        predict_request = predict_pb2.PredictRequest()
        predict_request.model_spec.name = 'distracted_driver_model'
        predict_request.model_spec.signature_name = 'serving_default'  
        predict_request.inputs["input_1"].CopyFrom(tf.make_tensor_proto(tf.reshape(X_new,(-1,224,224,3))))     
        log = prediction_log_pb2.PredictionLog(
            predict_log=prediction_log_pb2.PredictLog(request=predict_request))
        for r in range(NUM_RECORDS):
            writer.write(log.SerializeToString())    

if __name__ == "__main__":
    main()