easy_rec.python.main

easy_rec.python.main.train_and_evaluate(pipeline_config_path, continue_train=False)[source]

Train and evaluate a EasyRec model defined in pipeline_config_path.

Build an EasyRecEstimator, and then train and evaluate the estimator.

Parameters:
  • pipeline_config_path – a path to EasyRecConfig object, specifies

  • train_config – model_config, data_config and eval_config

  • continue_train – whether to restart train from an existing checkpoint

Returns:

None, the model will be saved into pipeline_config.model_dir

easy_rec.python.main.evaluate(pipeline_config, eval_checkpoint_path='', eval_data_path=None, eval_result_filename='eval_result.txt')[source]

Evaluate a EasyRec model defined in pipeline_config_path.

Evaluate the model defined in pipeline_config_path on the eval data, the metrics will be displayed on tensorboard and saved into eval_result.txt.

Parameters:
  • pipeline_config – either EasyRecConfig path or its instance

  • eval_checkpoint_path – if specified, will use this model instead of model specified by model_dir in pipeline_config_path

  • eval_data_path – eval data path, default use eval data in pipeline_config could be a path or a list of paths

  • eval_result_filename – evaluation result metrics save path.

Returns:

the metrics are specified in

pipeline_config_path

global_step: the global step for which this evaluation was performed.

Return type:

A dict of evaluation metrics

Raises:

AssertionError, if

  • pipeline_config_path does not exist

easy_rec.python.main.predict(pipeline_config, checkpoint_path='', data_path=None)[source]

Predict a EasyRec model defined in pipeline_config_path.

Predict the model defined in pipeline_config_path on the eval data.

Parameters:
  • pipeline_config – either EasyRecConfig path or its instance

  • checkpoint_path – if specified, will use this model instead of model specified by model_dir in pipeline_config_path

  • data_path – data path, default use eval data in pipeline_config could be a path or a list of paths

Returns:

A list of dict of predict results

Raises:

AssertionError, if

  • pipeline_config_path does not exist

easy_rec.python.main.export(export_dir, pipeline_config, checkpoint_path='', asset_files=None, verbose=False, **extra_params)[source]

Export model defined in pipeline_config_path.

Parameters:
  • export_dir – base directory where the model should be exported

  • pipeline_config – proto.EasyRecConfig instance or file path specify proto.EasyRecConfig

  • checkpoint_path – if specified, will use this model instead of model in model_dir in pipeline_config_path

  • asset_files – extra files to add to assets, comma separated; if asset file variable in graph need to be renamed, specify by new_file_name:file_path

  • version – if version is defined, then will skip writing embedding to redis, assume that embedding is already write into redis

  • verbose – dumps debug information

  • extra_params – keys related to write embedding to redis/oss redis_url, redis_passwd, redis_threads, redis_batch_size, redis_timeout, redis_expire if export embedding to redis; redis_embedding_version: if specified, will kill export to redis – oss_path, oss_endpoint, oss_ak, oss_sk, oss_timeout, oss_expire, oss_write_kv, oss_embedding_version

Returns:

the directory where model is exported

Raises:

AssertionError, if

  • pipeline_config_path does not exist