output is re-assigned as the variable embeddings

bark-simulator / bark

Open-Source Framework for Development, Simulation and Benchmarking of Behavior Planning Algorithms for Autonomous Driving

https://bark-simulator.github.io/

MIT License

287 stars 69 forks source link

output is re-assigned as the variable embeddings #505

Closed tinmodeHuang closed 3 years ago

tinmodeHuang commented 3 years ago

I wonder what the 2nd statement in the snippet does, which have confused me for long.

    output, network_state = self._encoder(embeddings, training=training)  # its call func be called
    output = embeddings

that's from gnn_actor_network.pyin /library_wrappers/lib_tf_agents/networks. it is infered from the comparison with ActorDistributionNetwork that the encoder here doen't take effect, isn't it?

patrickhart commented 3 years ago

As this is BARK-ML specific please consider posting it there. Also, feel free to create a PR for changing it. Best, Patrick

tinmodeHuang commented 3 years ago

sorry for no additon to detail it, read the complement, please @patrickhart . and another question, is the PR you said short for process?

patrickhart commented 3 years ago

BARK-ML can be found here: https://github.com/bark-simulator/bark-ml/tree/master/utils