Project structure of: google-research/robotics_transformer

• robotics_transformer Efficient Robotics with Diverse Transformers
  • __init__.py Robotics Transformer library version 0.1.0, Apache License 2.0, new PyPI release on version change.
  • configs
    • transformer_mixin.gin Configures TransformerNetwork for robotics actor-critic model with Adam optimizer.
  • film_efficientnet Tests EfficientNet film cat detection.
    • __init__.py Apache License, Version 2.0 copyright holder notice.
    • film_conditioning_layer.py FilmConditioning layer: deep learning FiLM technique for image processing
    • film_conditioning_layer_test.py Test: FilmConditioningLayer class, output rank matching.
    • film_efficientnet_encoder.py EfficientNet models for Keras image classification
    • film_efficientnet_encoder_test.py FilmEfficientNet cat detection test. Keras vs TF, EfficientNet encoder.
    • preprocessors.py Image preprocessing for deep learning models
    • preprocessors_test.py Testing image preprocessor functionality.
    • pretrained_efficientnet_encoder.py EfficientNetEncoder: Pre-trained model for feature extraction.
    • pretrained_efficientnet_encoder_test.py Tests EfficientNet encoder's image classification and encoding.
  • README.md Efficient Robotics Transformer Library for Control
  • requirements.txt Robotics transformer: Dependencies, absl-py, numpy, tensorflow.
  • sequence_agent.py RL agent class with SequenceAgent output and training functionality.
  • sequence_agent_test.py Test SequenceAgent using tf_agents library.
  • sequence_agent_test_set_up.py SequenceAgent test setup and tracking.
  • tokenizers Efficient transformer-based tokenizers for robotics.
    • __init__.py Apache License 2.0, copyright notice for code file.
    • action_tokenizer.py Action tokenizer for robotics transformer library.
    • action_tokenizer_test.py ActionTokenizer test: accuracy, int32 handling, episode termination.
    • image_tokenizer.py Image tokenizer using EfficientNet with optional token learner
    • image_tokenizer_test.py Image tokenizer test: checks shape of generated tokens.
    • token_learner.py Token Learner TF implementation for robotics transformer
    • token_learner_test.py TokenLearnerModule TensorFlow test. Batch * seq, num_tokens, embedding_dim output.
  • transformer.py Transformer layer with configurable parameters
  • transformer_network.py Transformer network for robotics tasks, with training/inference functions.
  • transformer_network_test.py Test transformer network with parameterized tests and masks.
  • transformer_network_test_set_up.py Configures testing parameters for robotics transformer network setup.
  • transformer_test.py Transformer model testing in robotics_transformer library.