Hi everyone,
I’m currently trying to run inference with a model trained using ML-Agents in an environment outside of Unity. To do this, I need to input tensors correctly converted from observation into the model.
I understand that numerical values like the x-coordinate and velocity specified in CollectObservation are directly incorporated into an array, and RayPerception returns the presence of detected tags along with the distance to the target. However, I’m unsure how these observations are transformed into tensors before being inputted into the model.
Could anyone advise on how to directly view these input tensors on the Console? I’m looking to understand the exact transformations applied during this process.
Thanks in advance for your help!