Performance issue in discussion/turnkey_inference_candidate/window_tune_nuts_sampling.py

[DLPerf]

Hello! Our static bug checker has found a performance issue in discussion/turnkey_inference_candidate/window_tune_nuts_sampling.py: Python type argument num_steps is passed to tf.function decorated function slow_adaptation_interval instead of tensors. As there is a for loop, every time the value changes, a new graph will be generated, and that can trigger tf.function retracing warning.

Here is the tensorflow document to support it.

Looking forward to your reply.

[DLPerf]

I mean it should be a problem but doesn't matter much if rectifying is hard or may make the code more complex.