'TF-Agents BoundedArray minimum value
I'm looking for information on how TF-agents handle Observation values if they are lower that the BoundedArraySpec minimum. My observation spec contains 3 array specs, each defined like:
'url1_m1':array_spec.BoundedArraySpec(shape=self.p_data_url1_m1.shape, dtype=np.float32, minimum = 0, name='url1_m1')
During training, and publishing the model negative values may be passed in (I have now changed the minimum value). I'm wondering how do TF-agents handle the negative values. Does it treat them as zero as the minimum allowed value?
Both training and published models function from previous version, and no error notified.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|
