Welcome to gobbli’s documentation¶
gobbli is a library designed to make experimentation and analysis using deep learning easier. It provides a simple, uniform interface to deep learning models that abstracts away most of the complexity in terms of different input/output formats, library versions, etc. It attempts to implement a set of common use cases with an emphasis on usability rather than performance.
gobbli is not designed to provide deep learning models in a production context. Each task generally involves running a Docker container in the background and transferring a large amount of data to and from disk, which creates significant overhead. Additionally, gobbli does not support fine-grained model-specific tuning, such as custom loss functions. Our goal is to take the user 80% of the way to their deep learning solution as quickly as possible so they can decide whether it’s worth the effort to resolve the remaining 20%.
- Prerequisites
- Quickstart
- Interactive Apps
- Troubleshooting
- Advanced Usage
- API Reference
- gobbli package
- Subpackages
- gobbli.augment package
- gobbli.dataset package
- gobbli.experiment package
- gobbli.inspect package
- gobbli.model package
- gobbli.test package
- Subpackages
- gobbli.test.augment package
- gobbli.test.classification package
- gobbli.test.dataset package
- gobbli.test.experiment package
- gobbli.test.interactive package
- gobbli.test.model package
- Submodules
- gobbli.test.model.test_base_model module
- gobbli.test.model.test_bert module
- gobbli.test.model.test_fasttext module
- gobbli.test.model.test_mtdnn module
- gobbli.test.model.test_sklearn module
- gobbli.test.model.test_spacy module
- gobbli.test.model.test_transformer module
- gobbli.test.model.test_use module
- Module contents
- Submodules
- Submodules
- Module contents
- Subpackages
- Submodules
- Module contents
- Subpackages
- gobbli package