Skip to main content

Small-vocabulary neural sequence-to-sequence models

Project description

Yoyodyne 🪀

PyPI version Supported Python versions CircleCI

Yoyodyne provides neural models for small-vocabulary sequence-to-sequence generation with and without feature conditioning.

These models are implemented using PyTorch and Lightning.

While we provide classic LSTM and transformer models, some of the provided models are particularly well-suited for problems where the source-target alignments are roughly monotonic (e.g., transducer) and/or where source and target vocabularies have substantial overlap (e.g., pointer_generator_lstm).

Philosophy

Yoyodyne is inspired by FairSeq (Ott et al. 2019) but differs on several key points of design:

  • It is for small-vocabulary sequence-to-sequence generation, and therefore includes no affordances for machine translation or language modeling. Because of this:
    • It has no plugin interface and the architectures provided are intended to be reasonably exhaustive.
    • There is little need for data preprocessing; it works with TSV files.
  • It has support for using features to condition decoding, with architecture-specific code to handle feature information.
  • 🚧 UNDER CONSTRUCTION 🚧: It has exhaustive test suites.
  • 🚧 UNDER CONSTRUCTION 🚧: It has performance benchmarks.
  • 🚧 UNDER CONSTRUCTION 🚧: Releases are made regularly.
  • It uses validation accuracy (not loss) for model selection and early stoppping.

Install

First install dependencies:

pip install -r requirements.txt

Then install:

pip install .

It can then be imported like a regular Python module:

import yoyodyne

Usage

Training

Training is performed by the yoyodyne-train script. One must specify the following required arguments:

  • --model_dir: path for model metadata and checkpoints
  • --experiment: name of experiment (pick something unique)
  • --train: path to TSV file containing training data
  • --val: path to TSV file containing validation data

The user can also specify various optional training and architectural arguments. See below or run yoyodyne-train --help for more information.

Prediction

Prediction is performed by the yoyodyne-predict script. One must specify the following required arguments:

  • --model_dir: path for model metadata
  • --experiment: name of experiment
  • --checkpoint: path to checkpoint
  • --predict: path to TSV file containing data to be predicted
  • --output: path for predictions

Run yoyodyne-predict --help for more information.

Data format

The default data format is a two-column TSV file in which the first column is the source string and the second the target string.

source   target

To enable the use of a feature column, one specifies a (non-zero) argument to --features-col. For instance in the SIGMORPHON 2017 shared task, the first column is the source (a lemma), the second is the target (the inflection), and the third contains semi-colon delimited feature strings:

source   target    feat1;feat2;...

this format is specified by --features-col 3.

Alternatively, for the SIGMORPHON 2016 shared task data:

source   feat1,feat2,...    target

this format is specified by --features_col 2 --features_sep , --target_col 3.

In order to ensure that targets are ignored during prediction, one can specify --target_col 0.

Model checkpointing

Checkpointing is handled by Lightning. The path for model information, including checkpoints, is specified by a combination of --model_dir and --experiment, such that we build the path model_dir/experiment/version_n, where each run of an experiment with the same model_dir and experiment is namespaced with a new version number. A version stores everything needed to reload the model, including the hyperparameters (model_dir/experiment_name/version_n/hparams.yaml) and the checkpoints directory (model_dir/experiment_name/version_n/checkpoints).

By default, each run initializes a new model from scratch, unless the --train_from argument is specified. To continue training from a specific checkpoint, the full path to the checkpoint should be specified with for the --train_from argument. This creates a new version, but starts training from the provided model checkpoint.

During training, we save the best --save_top_k checkpoints (by default, 1) ranked according to accuracy on the --val set. For example, --save_top_k 5 will save the top 5 most accurate models.

Reserved symbols

Yoyodyne reserves symbols of the form <...> for internal use. Feature-conditioned models also use [...] to avoid clashes between feature symbols and source and target symbols. Therefore, users should not provide any symbols of the form <...> or [...].

Models

The user specifies the overall architecture for the model using the --arch flag. The value of this flag specifies the decoder's architecture and whether or not an attention mechanism is present. This flag also specifies a default architecture for the encoder(s), but it is possible to override this with additional flags. Supported values for --arch are:

  • "attentive_lstm": This is an LSTM decoder with LSTM encoders (by default) and an attention mechanism. The initial hidden state is treated as a learned parameter.
  • "lstm": This is an LSTM decoder with LSTM encoders (by default); in lieu of an attention mechanism, the last non-padding hidden state of the encoder is concatenated with the decoder hidden state.
  • pointer_generator_lstm: This is an LSTM decoder with LSTM encoders (by default) and a pointer-generator mechanism. Since this model contains a copy mechanism, it may be superior to the lstm when the input and output vocabularies overlap significantly.
  • "transducer": This is an LSTM decoder with LSTM encoders (by default) and a neural transducer mechanism. On model creation, expectation maximization is used to learn a sequence of edit operations, and imitation learning is used to train the model to implement the oracle policy, with roll-in controlled by the --oracle_factor flag (default: 1). Since this model assumes monotonic alignment, it may be superior to attentive models when the alignment between input and output is roughly monotonic and when input and output vocabularies overlap significantly.
  • "transformer": This is a transformer decoder with transformer encoders (by default) and an attention mechanism. Sinusodial positional encodings and layer normalization are used. The user may wish to specify the number of attention heads (with --attention_heads; default: 4).

The user can override the default encoder architectures. One can override the source encoder using the --source_encoder flag:

  • "feature_invariant_transformer": This is a variant of the transformer encoder for use with features; it concatenates source and features and uses a learned embedding to distinguish between source and features symbols.
  • "linear": This is a linear encoder.
  • "lstm": This is a LSTM encoder.
  • "transformer": This is a transformer encoder.

When using features, the user can also specify a non-default features encoder using the --features_encoder flag ("linear", "lstm", "transformer").

For all models, the user may also wish to specify:

  • --decoder_layers (default: 1): number of decoder layers
  • --embedding (default: 128): embedding size
  • --encoder_layers (default: 1): number of encoder layers
  • --hidden_size (default: 512): hidden layer size

By default, LSTM encoders are bidirectional. One can disable this with the --no_bidirectional flag.

Training options

A non-exhaustive list includes:

  • Batch size:
    • --batch_size (default: 32)
  • Regularization:
    • --dropout (default: 0.2)
    • --label_smoothing (default: 0.0)
    • --gradient_clip_val (default: not enabled)
  • Optimizer:
    • --learning_rate (default: 0.001)
    • --optimizer (default: "adam")
    • --beta1 (default: 0.9): $\beta_1$ hyperparameter for the Adam optimizer (--optimizer adam)
    • --beta2 (default: 0.99): $\beta_2$ hyperparameter for the Adam optimizer (--optimizer adam)
    • --scheduler (default: not enabled)
  • Duration:
    • --max_epochs
    • --min_epochs
    • --max_steps
    • --min_steps
    • --max_time
    • --patience
  • Seeding:
    • --seed
  • Weights & Biases:
    • --log_wandb (default: False): enables Weights & Biases tracking

Hyperparameter tuning

No neural model should be deployed without proper hyperparameter tuning. However, the default options give a reasonable initial settings for an attentive biLSTM. For transformer-based architectures, experiment with multiple encoder and decoder layers, much larger batches, and the warmup-plus-inverse square root decay scheduler.

Automatic tuning

yododyne-train --auto_lr_find uses a heuristic (see Smith 2017) to propose an initial learning rate. Batch auto-scaling is not supported.

Weights & Biases tuning

wandb_sweeps shows how to use Weights & Biases to run hyperparameter sweeps.

Accelerators

Hardware accelerators can be used during training or prediction. In addition to CPU (the default) and GPU (--accelerator gpu), other accelerators may also be supported but not all have been tested yet.

Precision

By default, training uses 32-bit precision. However, the --precision flag allows the user to perform training with half precision (16) or with the bfloat16 half precision format if supported by the accelerator. This may reduce the size of the model and batches in memory, allowing one to use larger batches.

Examples

The examples directory contains interesting examples, including:

For developers

Developers, developers, developers! - Steve Ballmer

This section contains instructions for the Yoyodyne maintainers.

Releasing

  1. Create a new branch. E.g., if you want to call this branch "release": git checkout -b release
  2. Sync your fork's branch to the upstream master branch. E.g., if the upstream remote is called "upstream": git pull upstream master
  3. Increment the version field in pyproject.toml.
  4. Stage your changes: git add pyproject.toml.
  5. Commit your changes: git commit -m "your commit message here"
  6. Push your changes. E.g., if your branch is called "release": git push origin release
  7. Tag the master branch's last commit. The tag should begin with v; e.g., if the new version is 3.1.4, the tag should be v3.1.4. This can be done:
    • on GitHub itself: click the "Releases" or "Create a new release" link on the right-hand side of the Yoyodyne GitHub page) and follow the dialogues.
    • from the command-line using git tag.
  8. Build the new release: python -m build
  9. Upload the result to PyPI: twine upload dist/*

References

Ott, M., Edunov, S., Baevski, A., Fan, A., Gross, S., Ng, N., Grangier, D., and Auli, M. 2019. fairseq: a fast, extensible toolkit for sequence modeling. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics (Demonstrations), pages 48-53.

Smith, L. N. 2017. Cyclical learning rates for training neural networks. In 2017 IEEE Winter Conference on Applications of Computer Vision, pages 464-472.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

yoyodyne-0.2.4.tar.gz (55.2 kB view hashes)

Uploaded Source

Built Distribution

yoyodyne-0.2.4-py3-none-any.whl (64.1 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page