softmax_temperature
Experimental
Experimental: parameter and/or its behaviour may change or be removed without notification. Please contact for more information.
A float value that influences the confidence of predictions. Increasing the value reduces confidence, while decreasing it increases confidence.
The optimal value will give the best calibration of the model. A model that is correctly calibrated provides probabilities matching the real-world accuracy of predictions. The optimal value is dependent on which specialized model is used.
Example use
Requires curl and jq.
Requires the Python requests library. Available with pip install requests.