Kyle Sayers

Made a PR [here](https://github.com/onnx/onnx/pull/4324) to help reduce run time. I agree, the current bfs implementation and data structure is not optimized for speed

I also encountered this message and failed to run my transformers model with tflite ArmNN version: `ArmNN v29.0.0`, prebuilt binaries for x86 ACL version: Not downloaded TfLite version: `tensorflow ==...

I recently found that [ArmNN does not support layer norm](https://github.com/ARM-software/armnn/issues/461). If this is the issue, it'd be nice if the error reflected that

Thanks for the responses. I doubt this directly the issue, since I've seen this "backend missing" warning before with resnets with no issues running. I've downloaded the prebuilt binaries, is...

Thanks @ArmRyan, those resources are very helpful. I've added the libraries to my LD_LIBRARY_PATH and no longer get the warning, but outside of that nothing has changed ``` Info: ArmNN...

On Raspberry Pi I got the error `Warning: GetConnectedConstantAsInputTensors() called on Layer with no connected Constants as Input Tensors.`

The evalcallback already specially constructs an environment for evaluation, and this environment should act slightly differently than typical VecEnvs (for example, stop stepping after reaching a terminal state and do...

Side note, there's another discussion to be had about parallelizing the eval environment, aka passing a `n_envs` argument to `evaluate_policy`

> The two are related, evaluate_policy() already supports multiple envs, and that's because of VecEnv precisely. `evaluate_policy` has the capacity to support multiple envs, but currently hard codes just one....

> What makes you think that? common/evaluation.py:62 `env = DummyVecEnv([lambda: env])` common/envs/dummy_vec_env.py:42 `super().__init__(len(env_fns), env.observation_space, env.action_space)` common/envs/base_vec_env.py:63 `self.num_envs = num_envs` Therefore `env.num_envs` is always `1`. Is it possible I'm misreading something?