添加checkpoint warm_start_from 配置

[shfshf]

seq2annotation.trainer.train_model.py中添加warm_start_from=config.get("warm_start_dir", None)：
estimator = tf.estimator.Estimator(
model_fn, instance_model_dir, cfg, estimator_params,
warm_start_from=config.get("warm_start_dir", None)
)
在configure.yaml配置中添加：warm_start_dir:

经过非专业测试，已无问题

[xiaomihao]

shf could you please tell us what the so_called warm_start_dir is meaning. Why should add this param to the func. What's the point please.

[shfshf]

warm_start_from:
Optional, string, checkpoint file path, used to indicate where to start the warm boot. Or the tf.estimator.WarmStartSettings class to configure all hot starts. If it is a serial path, all variables are hot-started, and the names of the Tensor and vocabulary are not changed.
why we need this param: Change a small amount of corpus later without restarting training