Inference¶

In this tutorial, we will learn how to run inference on a test data.

Workflow¶

For inference, we have 2 steps:

Specify LOG_DIR (LOG_DIR is a directory where parameters are saved)
Run predict method

(Step 1) Pre-trained Model Restoration¶

When training starts, a log directory where the results are saved will be made. To restore the results, specify the log directory path in a config file and pass it when initialization.

# infer.yml
LOGGER:
  # For default, log_dir name is randomly generated
  LOG_DIR: "bfb5118b-7687-453d-a8d8-6100df7d36d4"

from tsts.solvers import TimeSeriesForecaster

forecaster = TimeSeriesForecaster("infer.yml")

To specify the name of log_dir, pass a config file when starting training.

# custom-log-dir.yml
LOGGER:
  LOG_DIR: "mymodel"

import torch
from tsts.solvers import TimeSeriesForecaster

sin_dataset = torch.sin(torch.arange(0, 100, 0.1))
sin_dataset = sin_dataset.unsqueeze(-1)
forecaster = TimeSeriesForecaster("custom-log-dir.yml")
forecaster.fit([sin_dataset])

(Step 2) Running Inference¶

Run predict method to infer on test data.

import torch
from tsts.solvers import TimeSeriesForecaster

test_data = torch.arange(0, 10, 0.1)
test_data = test_data.unsqueeze(-1)
forecaster = TimeSeriesForecaster("custom-log-dir.yml")
print(forecaster.predict(test_data))

"""
Output:
tensor([[0.1068],
    [0.2669],
    [0.3835],
    [0.4387],
    [0.4649],
    [0.4782],
    [0.4856],
    [0.4902]])
"""