Hi, I’m doing a project to train my RL Agent with external lib (pyTorch).
After training i convert and save my model as onnx, and as i studied, the format(structure) of onnx is important for ml-agents, so i tried to make my onnx has same structure with example.
The left figure is my onnx structure, and the right one is for the example(Ball 3D).
Even the structures are same, in my project, a warning says “The model uses deterministic inference but does not contain Deterministic Continuous Action Output Tensor.” or “The model uses stochastic inference but does not contain Continuous Action Output Tensor.”, but my onnx has both continuous_actions and deterministic_continuous_actions.
Also, if i play the game ignoring the warning, i stuck with an error which says “AssertionException: Cannot peek output tensor recurrent_out as model does not contain an output with that name”.
Please help me, and sorry for my english.
Thanks