Shulou(Shulou.com)06/03 Report--

This article mainly introduces "what is the role of various files generated when saving models in Tensorflow". In daily operation, I believe that many people have doubts about the role of various files generated when saving models in Tensorflow. Xiaobian consulted all kinds of materials and sorted out simple and easy-to-use methods of operation. I hope it will be helpful to answer the questions of "what is the role of various files generated when saving models in Tensorflow?" Next, please follow the editor to study!

Tensorboard file

What events.out.tfevents.*...: holds is the value of your accuracy or loss at different times.

Graph.pbtxt: this is actually a text file that stores the structure information of the model, some of which are as follows:

Node_def {name: "FixedLengthRecordDataset/input_pipeline_task10/buffer_size" op: "Const" device: "/ job:worker/task:10/device:CPU:0" attr {key: "dtype" value {type: DT_INT64}} attr {key: "value" value { Tensor {dtype: DT_INT64 tensor_shape {} int64_val: 262144}} node_def {name: "FixedLengthRecordDataset" op: "FixedLengthRecordDataset" input: "FixedLengthRecordDataset/input_pipeline_task10/filenames:output:0" input: "FixedLengthRecordDataset/input_pipeline_task10/header_ Bytes:output:0 "input:" FixedLengthRecordDataset/input_pipeline_task10/record_bytes:output:0 "input:" FixedLengthRecordDataset/input_pipeline_task10/footer_bytes:output:0 "input:" FixedLengthRecordDataset/input_pipeline_task10/buffer_size:output:0 "device:" / job:worker/task:10/device:CPU:0 "}. The file generated when the model is saved

Checkpoint: it's actually a txt file that stores path information. We can take a look at what it contains:

Model_checkpoint_path: "model.ckpt-5000" all_model_checkpoint_paths: "model.ckpt-0" all_model_checkpoint_paths: "model.ckpt-5000"

You can see that the first line represents the most recent checkpoints path information, which means that your model training may have been interrupted for some reason. But it doesn't matter, the next time I retrain, I will automatically continue to train from the last breakpoint without having to retrain. The last two items represent all breakpoint paths that have been saved.

Model.ckpt-*.meta: in fact, the same as the above graph.pbtxt function to save the graph structure, but the meta file is binary, it includes GraphDef,SaverDef, etc., when there is meta file, we can not define the model in the file, but also can run, and if there is no meta file, we need to define the model, and then load data file to get the value of the variable.

Model.ckpt-*.index: this is a string-string table,table whose key value is tensor name and value is serialized BundleEntryProto. Each BundleEntryProto represents the metadata of the tensor, for example, the data file contains tensor, the offset in the file, some auxiliary data, and so on.

Model.ckpt-*.data-*: holds the values of all variables in the model, the TensorBundle collection.

At this point, the study on "what are the functions of the various files generated when saving the model in Tensorflow" is over. I hope to be able to solve your doubts. The collocation of theory and practice can better help you learn, go and try it! If you want to continue to learn more related knowledge, please continue to follow the website, the editor will continue to work hard to bring you more practical articles!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.