Скачать с ютуб The Sensory Neuron as a Transformer in PyTorch в хорошем качестве

The Sensory Neuron as a Transformer in PyTorch

In this video, we implement a paper called "The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning" in PyTorch. It proposes a permutation invariant module called the Attention Neuron. Its goal is to independently process local information from the features and then combine the local knowledge into a global picture. Paper: https://arxiv.org/abs/2109.02869 Official website with visualizations: https://attentionneuron.github.io/ Official code: https://github.com/google/brain-tokyo... Code from this video: https://github.com/jankrepl/mildlyove... 00:00 Intro + credits + disclaimer 01:10 Paper overview: Goals [slides] 02:10 Paper overview: Inputs and outputs [slides] 02:54 Paper overview: Pipeline [slides] 03:42 Paper overview: Attention inputs [slides] 05:22 Paper overview: Attention [slides] 06:16 Paper overview: Linear module [slides] 06:44 Multi-layer perceptron [code] 09:50 Positional encoding [code] 11:12 AttentionMatrix module [code] 13:23 AttentionNeuron module [code] 18:02 AttentionNeuron testing it out [code] 20:31 PermutationInvariantNetwork module [code] 22:10 Solution abstract class [code] 26:22 MLPSolution [code] 27:36 PermutationInvariantSolution [code] 29:33 Custom CartPoleSwing up [no code] 30:19 Task constructor [code] 32:44 Task helper methods [code] 34:06 Task rollout [code] 36:17 Covariance matrix adaptation [no code] 37:02 Training script helpers [code] 38:14 Training script CLI [code] 39:43 Training script main logic [code] 44:02 Launch multiple experiments [code] 45:10 Tensorboard results [plot] 46:26 Feature shuffling evaluation [plot] 48:34 Noise feature injection [plot] 49:25 Final thoughts + outro If you have any video suggestions or you just wanna chat feel free to join the discord server:   / discord   Twitter:   / moverfitted   Credits logo animation Title: Conjungation · Author: Uncle Milk · Source:   / unclemilk   · License: https://creativecommons.org/licenses/... · Download (9MB): https://auboutdufil.com/?id=600