project.yml

title: "Demo spancat in a new pipeline (Span Categorization)"
description: "A minimal demo spancat project for spaCy v3"

# Variables can be referenced across the project.yml using ${vars.var_name}
vars:
  name: "placing_holocaust"
  lang: "en"
  annotations_file: "annotated_data_spans.jsonl"
  train: "train"
  dev: "dev"
  test: "test"
  version: "0.0.1"
  # Set a random seed
  seed: 0
  # Set your GPU ID, -1 is CPU
  gpu_id: -1
  vectors_model_md: "en_core_web_md"
  vectors_model_lg: "en_core_web_lg"

# These are the directories that the project needs. The project CLI will make
# sure that they always exist.
directories: ["assets", "corpus", "configs", "training", "scripts", "packages"]

# Assets that should be downloaded or available in the directory. We're shipping
# them with the project, so they won't have to be downloaded.
assets:
  - dest: "assets/train.jsonl"
    description: "Training data. For this project, they were chunked into sentences."
  - dest: "assets/dev.jsonl"
    description: "Validation data. For this project, they were chunked into sentences."
  - dest: "assets/test.jsonl"
    description: "Testing data. For this project, they were chunked into sentences."

  - dest: "assets/annotated_data.json/"
    description: "All data, including those without annotations because they are negative examples."

  - dest: "assets/annotated_data_spans.jsonl"
    description: "This is just the data that contained examples of span annotations."

  - dest: "corpus/train.spacy"
    description: "Training data in serialized format."
  - dest: "corpus/dev.spacy"
    description: "Validation data in serialized format."
  - dest: "corpus/test.spacy"
    description: "Testing data in serialized format."

  - dest: "gold-training-data/*"
    description: "The original outputs from Prodigy, the annotation software used."

  - dest: "notebooks/*"
    description: "A collection of notebooks for testing different features of the project."

  - dest: "configs/*"
    description: "A collection of config files used for training the spaCy models." 
# Workflows are sequences of commands (see below) executed in order. You can
# run them via "spacy project run [workflow]". If a commands's inputs/outputs
# haven't changed, it won't be re-run.
workflows:
  all-sm-sents:
    - convert-sents
    - split
    - create-config-sm
    - train-sm
    - evaluate-sm
  # all-trf:
  #   - download
  #   - convert
  #   - create-config
  #   - train-with-vectors
  #   - evaluate

# Project commands, specified in a style similar to CI config files (e.g. Azure
# pipelines). The name is the command name that lets you trigger the command
# via "spacy project run [command] [path]". The help message is optional and
# shown when executing "spacy project run [optional command] [path] --help".
commands:

#### DOWNLOADING VECTORS #####
  - name: "download-lg"
    help: "Download a spaCy model with pretrained vectors"
    script:
      - "python -m spacy download ${vars.vectors_model_lg}"

  - name: "download-md"
    help: "Download a spaCy model with pretrained vectors"
    script:
      - "python -m spacy download ${vars.vectors_model_md}"

#### PREPROCESSING #####
  - name: "convert"
    help: "Convert the data to spaCy's binary format"
    script:
      - "python scripts/convert.py ${vars.lang} assets/${vars.train}.jsonl corpus"
      - "python scripts/convert.py ${vars.lang} assets/${vars.dev}.jsonl corpus"
      - "python scripts/convert.py ${vars.lang} assets/${vars.test}.jsonl corpus"
    deps:
      - "assets/${vars.train}.jsonl"
      - "assets/${vars.dev}.jsonl"
      - "assets/${vars.test}.jsonl"
      - "scripts/convert.py"
    outputs:
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
      - "corpus/test.spacy"

  - name: "convert-sents"
    help: "Convert the data to to sentences before converting to spaCy's binary format"
    script:
      - "python scripts/convert_sents.py ${vars.lang} assets/${vars.train}.jsonl corpus"
      - "python scripts/convert_sents.py ${vars.lang} assets/${vars.dev}.jsonl corpus"
      - "python scripts/convert_sents.py ${vars.lang} assets/${vars.test}.jsonl corpus"
    deps:
      - "assets/${vars.train}.jsonl"
      - "assets/${vars.dev}.jsonl"
      - "assets/${vars.test}.jsonl"
      - "scripts/convert.py"
    outputs:
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
      - "corpus/test.spacy"

  - name: "split"
    help: "Split data into train/dev/test sets"
    script:
      - "python scripts/split.py assets/${vars.annotations_file}"
    deps:
      - "scripts/split.py"
    outputs:
      - "assets/train.jsonl"
      - "assets/dev.jsonl"
      - "assets/test.jsonl"



#### CONFIG CREATIONS #####

  - name: "create-config-sm"
    help: "Create a new config with a spancat pipeline component"
    script:
      - "python -m spacy init fill-config configs/base_config_sm.cfg configs/config_sm.cfg"
    deps:
      - configs/base_config_sm.cfg
    outputs:
      - "configs/config.cfg"


#### TRAINING #####

### small ###
  - name: "train-sm"
    help: "Train the spancat model"
    script:
      - >-
        python -m spacy train configs/config_sm.cfg --output training/sm/ 
        --paths.train corpus/train.spacy --paths.dev corpus/dev.spacy
         --training.eval_frequency 50 
        --training.patience 0 
        --gpu-id ${vars.gpu_id}
        --system.seed ${vars.seed}
    deps:
      - "configs/config_lg.cfg"
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
    outputs:
      - "training/model-best"


### medium ###
  - name: "train-md"
    help: "Train the spancat model with vectors"
    script:
      - >-
        python -m spacy train configs/config_md.cfg --output training/md/ 
        --paths.train corpus/train.spacy --paths.dev corpus/dev.spacy 
        --training.eval_frequency 50 
        --training.patience 0 
        --gpu-id ${vars.gpu_id} 
        --initialize.vectors ${vars.vectors_model_md} 
        --system.seed ${vars.seed}
        --components.tok2vec.model.embed.include_static_vectors true
    deps:
      - "configs/config_md.cfg"
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
    outputs:
      - "training/model-best"


### large ###
  - name: "train-lg"
    help: "Train the spancat model with vectors"
    script:
      - >-
        python -m spacy train configs/config_lg.cfg --output training/lg/ 
        --paths.train corpus/train.spacy --paths.dev corpus/dev.spacy 
        --training.eval_frequency 50 
        --training.patience 0 
        --gpu-id ${vars.gpu_id} 
        --initialize.vectors ${vars.vectors_model_lg} 
        --system.seed ${vars.seed}
        --components.tok2vec.model.embed.include_static_vectors true
    deps:
      - "configs/config_lg.cfg"
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
    outputs:
      - "training/model-best"


### transformer ###
  - name: "train-trf"
    help: "Train the spancat model"
    script:
      - >-
        python -m spacy train configs/config_trf.cfg --output training/trf/ 
        --paths.train corpus/train.spacy --paths.dev corpus/dev.spacy 
        --training.patience 100
        --gpu-id ${vars.gpu_id}
        --system.seed ${vars.seed}
    deps:
      - "configs/config.cfg"
      - "corpus/train.spacy"
      - "corpus/dev.spacy"
    outputs:
      - "training/model-best"


#### EVALUATION #####

### small ###
  - name: "evaluate-sm"
    help: "Evaluate the model and export metrics"
    script:
      - "python -m spacy evaluate training/sm/model-best corpus/test.spacy --output training/sm/metrics.json"
    deps:
      - "corpus/test.spacy"
      - "training/sm/model-best"
    outputs:
      - "training/sm/metrics.json"

### medium ###

  - name: "evaluate-md"
    help: "Evaluate the model and export metrics"
    script:
      - "python -m spacy evaluate training/md/model-best corpus/test.spacy --output training/md/metrics.json"
    deps:
      - "corpus/test.spacy"
      - "training/md/model-best"
    outputs:
      - "training/md/metrics.json"

### large ###
  - name: "evaluate-lg"
    help: "Evaluate the model and export metrics"
    script:
      - "python -m spacy evaluate training/lg/model-best corpus/test.spacy --output training/lg/metrics.json"
    deps:
      - "corpus/test.spacy"
      - "training/lg/model-best"
    outputs:
      - "training/lg/metrics.json"


#### PACKAGING #####

  - name: "build-table"
    help: "builds a nice table from the metrics for README.md"
    script:
      - "python scripts/build-table.py"

  - name: "readme"
    help: "builds a nice table from the metrics for README.md"
    script:
      - "python scripts/readme.py"

  - name: package
    help: "Package the trained model as a pip package"
    script:
      - "python -m spacy package training/model-best packages --name ${vars.name} --version ${vars.version} --force"
    deps:
      - "training/model-best"
    outputs_no_cache:
      - "packages/${vars.lang}_${vars.name}-${vars.version}/dist/${vars.lang}_${vars.name}-${vars.version}.tar.gz"

  - name: clean
    help: "Remove intermediary directories"
    script:
      - "rm -rf corpus/*"
      - "rm -rf training/*"
      - "rm -rf metrics/*"