

Model¶

The following section describe the commands related to Models.

Defined in animius\Console.py

createModel¶

Create a model.

1	createModel -n 'model name' -t 'ModelType' -c 'model_config name' -d 'data name'

Keyword Arguments:

-n, --name (str) -- Name of model
-t, --type (str) -- Type of model
-c, --model_config (str) -- Name of model config to use
-d, --data (str) -- Name of data to use
-i, --intent_ner_model* (str) -- (Optional) Name of IntentNER Model (Only required for creating CombinedChatbot Model)

Here's a list of model types.

Chatbot: am.Chatbot.ChatbotModel()

CombinedChatbot: animius.Chatbot.CombinedChatbotModel()

IntentNER: animius.IntentNER.IntentNERModel()

SpeakerVerification: animius.SpeakerVerification.SpeakerVerificationModel()

deleteModel¶

Delete a model.

1	deleteModel -n 'model name'

Keyword Arguments:

-n, --name (str) -- Name of model to delete

saveModel¶

Save a model.

The graph is saved in '\resource\model_name\model_name_graph.pb'

1	saveModel -n 'model name' -g True

Keyword Arguments:

-n, --name (str) -- Name of model to save
-g, --graph (bool) -- Whether to save the graph

loadModel¶

Load a model.

1	loadModel -n 'model name' -d 'data name'

Keyword Arguments:

-n, --name (str) -- Name of model to load
-d, --data (str) -- Name of data to set to model

exportModel¶

Export a model to zip file.

1	exportModel -n 'model name' -p 'some\path\to\export\'

Keyword Arguments:

-n, --name (str) -- Name of model to export
-p, --path (str) -- Path to export file

importModel¶

Import a model from zip file.

1	importModel -n 'model name' -p 'some\path\to\export\model_name.zip'

Keyword Arguments:

-n, --name (str) -- Name of model to export
-p, --path (str) -- Path to import file

getModels¶

Get a list of existing models.

getModels

No argument required.

This command returns a dictionary of which the keys are the name of models and the values are the details.

The details will be empty if the model is not loaded.

{
    "model_name": {
        "name": "model_name",
        "type": "<class 'model_class'>"
    }
}

getModelDetails¶

Get the details of a model.

1	getModelDetails -n 'model name'

Keyword Arguments:

-n, --name (str) -- Name of model

This command returns a dictionary of details of a model, which contains configs, hyperparameters, structures, saved name and saved directory of the model.

{
    'config': {
        'device': '/gpu:0',
        'class': 'IntentNER',
        'epoch': 0,
        'cost': None,
        'display_step': 1,
        'tensorboard': None,
        'hyperdash': None,
        'graph': 'resources\\models\\model_name\\model_name_graph.pb',
        'frozen_graph': 'resources\\models\\model_name\\frozen_model.pb'
    },
    'model_structure': {
        'max_sequence': 20,
        'n_hidden': 128,
        'gradient_clip': 5.0,
        'node': 'gru',
        'n_intent_output': 15,
        'n_ner_output': 8,
        'n_vector': 303,
        'word_count': 100000
    },
    'hyperparamter': {
        'learning_rate': 0.003,
        'batch_size': 1024,
        'optimizer': 'adam'
    },
    'saved_directory': 'resources\\models\\model_name',
    'saved_name': 'model_name'
}

setData¶

Set model data.

1	setData -n 'model name' -d 'data name'

Keyword Arguments:

-n, --name (str) -- Name of model
-d, --data (str) -- Name of data

train¶

Train a model.

The training process will be held in another thread.

The training device is defined in the model config.

1	train -n 'model name' -e 10

Keyword Arguments:

-n, --name (str) -- Name of model to train
-e, --epoch (int) -- Number of epochs to train for

stopTraining¶

Cancel training a model.

The model will stop once it finishes the current epoch.

1	stopTraining -n 'model name'

Keyword Arguments:

-n, --name (str) -- Name of model to stop

predict¶

Make predictions with a model.

1	predict -n 'model name' -i 'name of input data' -s '\some\path.txt'

Keyword Arguments:

-n, --name (str) -- Name of model
-i, --input_data (str) -- Name of input data
-s, --save_path (str) -- Path to save result (Optional)

freezeGraph¶

Freeze Tensorflow graph and latest checkpoint to 'resource\model_name\frozen_model.pb'.

1	freezeGraph -n 'model name'

Keyword Arguments:

-n, --name (str) -- Name of model

optimize¶

Optimize a frozen model (see FreezeGraph) for inference.

1	optimize -n 'model name'

Keyword Arguments:

-n, --name (str) -- Name of model