spaCy/spacy/cli/train_from_config.py
Sofie Van Landeghem 569cc98982
Update spaCy for thinc 8.0.0 (#4920)
* Add load_from_config function

* Add train_from_config script

* Merge configs and expose via spacy.config

* Fix script

* Suggest create_evaluation_callback

* Hard-code for NER

* Fix errors

* Register command

* Add TODO

* Update train-from-config todos

* Fix imports

* Allow delayed setting of parser model nr_class

* Get train-from-config working

* Tidy up and fix scores and printing

* Hide traceback if cancelled

* Fix weighted score formatting

* Fix score formatting

* Make output_path optional

* Add Tok2Vec component

* Tidy up and add tok2vec_tensors

* Add option to copy docs in nlp.update

* Copy docs in nlp.update

* Adjust nlp.update() for set_annotations

* Don't shuffle pipes in nlp.update, decruft

* Support set_annotations arg in component update

* Support set_annotations in parser update

* Add get_gradients method

* Add get_gradients to parser

* Update errors.py

* Fix problems caused by merge

* Add _link_components method in nlp

* Add concept of 'listeners' and ControlledModel

* Support optional attributes arg in ControlledModel

* Try having tok2vec component in pipeline

* Fix tok2vec component

* Fix config

* Fix tok2vec

* Update for Example

* Update for Example

* Update config

* Add eg2doc util

* Update and add schemas/types

* Update schemas

* Fix nlp.update

* Fix tagger

* Remove hacks from train-from-config

* Remove hard-coded config str

* Calculate loss in tok2vec component

* Tidy up and use function signatures instead of models

* Support union types for registry models

* Minor cleaning in Language.update

* Make ControlledModel specifically Tok2VecListener

* Fix train_from_config

* Fix tok2vec

* Tidy up

* Add function for bilstm tok2vec

* Fix type

* Fix syntax

* Fix pytorch optimizer

* Add example configs

* Update for thinc describe changes

* Update for Thinc changes

* Update for dropout/sgd changes

* Update for dropout/sgd changes

* Unhack gradient update

* Work on refactoring _ml

* Remove _ml.py module

* WIP upgrade cli scripts for thinc

* Move some _ml stuff to util

* Import link_vectors from util

* Update train_from_config

* Import from util

* Import from util

* Temporarily add ml.component_models module

* Move ml methods

* Move typedefs

* Update load vectors

* Update gitignore

* Move imports

* Add PrecomputableAffine

* Fix imports

* Fix imports

* Fix imports

* Fix missing imports

* Update CLI scripts

* Update spacy.language

* Add stubs for building the models

* Update model definition

* Update create_default_optimizer

* Fix import

* Fix comment

* Update imports in tests

* Update imports in spacy.cli

* Fix import

* fix obsolete thinc imports

* update srsly pin

* from thinc to ml_datasets for example data such as imdb

* update ml_datasets pin

* using STATE.vectors

* small fix

* fix Sentencizer.pipe

* black formatting

* rename Affine to Linear as in thinc

* set validate explicitely to True

* rename with_square_sequences to with_list2padded

* rename with_flatten to with_list2array

* chaining layernorm

* small fixes

* revert Optimizer import

* build_nel_encoder with new thinc style

* fixes using model's get and set methods

* Tok2Vec in component models, various fixes

* fix up legacy tok2vec code

* add model initialize calls

* add in build_tagger_model

* small fixes

* setting model dims

* fixes for ParserModel

* various small fixes

* initialize thinc Models

* fixes

* consistent naming of window_size

* fixes, removing set_dropout

* work around Iterable issue

* remove legacy tok2vec

* util fix

* fix forward function of tok2vec listener

* more fixes

* trying to fix PrecomputableAffine (not succesful yet)

* alloc instead of allocate

* add morphologizer

* rename residual

* rename fixes

* Fix predict function

* Update parser and parser model

* fixing few more tests

* Fix precomputable affine

* Update component model

* Update parser model

* Move backprop padding to own function, for test

* Update test

* Fix p. affine

* Update NEL

* build_bow_text_classifier and extract_ngrams

* Fix parser init

* Fix test add label

* add build_simple_cnn_text_classifier

* Fix parser init

* Set gpu off by default in example

* Fix tok2vec listener

* Fix parser model

* Small fixes

* small fix for PyTorchLSTM parameters

* revert my_compounding hack (iterable fixed now)

* fix biLSTM

* Fix uniqued

* PyTorchRNNWrapper fix

* small fixes

* use helper function to calculate cosine loss

* small fixes for build_simple_cnn_text_classifier

* putting dropout default at 0.0 to ensure the layer gets built

* using thinc util's set_dropout_rate

* moving layer normalization inside of maxout definition to optimize dropout

* temp debugging in NEL

* fixed NEL model by using init defaults !

* fixing after set_dropout_rate refactor

* proper fix

* fix test_update_doc after refactoring optimizers in thinc

* Add CharacterEmbed layer

* Construct tagger Model

* Add missing import

* Remove unused stuff

* Work on textcat

* fix test (again :)) after optimizer refactor

* fixes to allow reading Tagger from_disk without overwriting dimensions

* don't build the tok2vec prematuraly

* fix CharachterEmbed init

* CharacterEmbed fixes

* Fix CharacterEmbed architecture

* fix imports

* renames from latest thinc update

* one more rename

* add initialize calls where appropriate

* fix parser initialization

* Update Thinc version

* Fix errors, auto-format and tidy up imports

* Fix validation

* fix if bias is cupy array

* revert for now

* ensure it's a numpy array before running bp in ParserStepModel

* no reason to call require_gpu twice

* use CupyOps.to_numpy instead of cupy directly

* fix initialize of ParserModel

* remove unnecessary import

* fixes for CosineDistance

* fix device renaming

* use refactored loss functions (Thinc PR 251)

* overfitting test for tagger

* experimental settings for the tagger: avoid zero-init and subword normalization

* clean up tagger overfitting test

* use previous default value for nP

* remove toy config

* bringing layernorm back (had a bug - fixed in thinc)

* revert setting nP explicitly

* remove setting default in constructor

* restore values as they used to be

* add overfitting test for NER

* add overfitting test for dep parser

* add overfitting test for textcat

* fixing init for linear (previously affine)

* larger eps window for textcat

* ensure doc is not None

* Require newer thinc

* Make float check vaguer

* Slop the textcat overfit test more

* Fix textcat test

* Fix exclusive classes for textcat

* fix after renaming of alloc methods

* fixing renames and mandatory arguments (staticvectors WIP)

* upgrade to thinc==8.0.0.dev3

* refer to vocab.vectors directly instead of its name

* rename alpha to learn_rate

* adding hashembed and staticvectors dropout

* upgrade to thinc 8.0.0.dev4

* add name back to avoid warning W020

* thinc dev4

* update srsly

* using thinc 8.0.0a0 !

Co-authored-by: Matthew Honnibal <honnibal+gh@gmail.com>
Co-authored-by: Ines Montani <ines@ines.io>
2020-01-29 17:06:46 +01:00

446 lines
14 KiB
Python

import plac
from thinc.util import require_gpu
from wasabi import msg
from pathlib import Path
import thinc
import thinc.schedules
from thinc.model import Model
from spacy.gold import GoldCorpus
import spacy
from spacy.pipeline.tok2vec import Tok2VecListener
from typing import Optional, Dict, List, Union, Sequence
from pydantic import BaseModel, FilePath, StrictInt
import tqdm
from ..ml import component_models
from .. import util
registry = util.registry
CONFIG_STR = """
[training]
patience = 10
eval_frequency = 10
dropout = 0.2
init_tok2vec = null
vectors = null
max_epochs = 100
orth_variant_level = 0.0
gold_preproc = false
max_length = 0
use_gpu = 0
scores = ["ents_p", "ents_r", "ents_f"]
score_weights = {"ents_f": 1.0}
limit = 0
[training.batch_size]
@schedules = "compounding.v1"
start = 100
stop = 1000
compound = 1.001
[optimizer]
@optimizers = "Adam.v1"
learn_rate = 0.001
beta1 = 0.9
beta2 = 0.999
[nlp]
lang = "en"
vectors = ${training:vectors}
[nlp.pipeline.tok2vec]
factory = "tok2vec"
[nlp.pipeline.ner]
factory = "ner"
[nlp.pipeline.ner.model]
@architectures = "transition_based_ner.v1"
nr_feature_tokens = 3
hidden_width = 64
maxout_pieces = 3
[nlp.pipeline.ner.model.tok2vec]
@architectures = "tok2vec_tensors.v1"
width = ${nlp.pipeline.tok2vec.model:width}
[nlp.pipeline.tok2vec.model]
@architectures = "hash_embed_cnn.v1"
pretrained_vectors = ${nlp:vectors}
width = 128
depth = 4
window_size = 1
embed_size = 10000
maxout_pieces = 3
"""
class PipelineComponent(BaseModel):
factory: str
model: Model
class Config:
arbitrary_types_allowed = True
class ConfigSchema(BaseModel):
optimizer: Optional["Optimizer"]
class training(BaseModel):
patience: int = 10
eval_frequency: int = 100
dropout: float = 0.2
init_tok2vec: Optional[FilePath] = None
vectors: Optional[str] = None
max_epochs: int = 100
orth_variant_level: float = 0.0
gold_preproc: bool = False
max_length: int = 0
use_gpu: int = 0
scores: List[str] = ["ents_p", "ents_r", "ents_f"]
score_weights: Dict[str, Union[int, float]] = {"ents_f": 1.0}
limit: int = 0
batch_size: Union[Sequence[int], int]
class nlp(BaseModel):
lang: str
vectors: Optional[str]
pipeline: Optional[Dict[str, PipelineComponent]]
class Config:
extra = "allow"
# Of course, these would normally decorate the functions where they're defined.
# But for now...
@registry.architectures.register("hash_embed_cnn.v1")
def hash_embed_cnn(
pretrained_vectors, width, depth, embed_size, maxout_pieces, window_size
):
return component_models.Tok2Vec(
width=width,
embed_size=embed_size,
pretrained_vectors=pretrained_vectors,
conv_depth=depth,
cnn_maxout_pieces=maxout_pieces,
bilstm_depth=0,
window_size=window_size,
)
@registry.architectures.register("hash_embed_bilstm.v1")
def hash_embed_bilstm_v1(pretrained_vectors, width, depth, embed_size):
return component_models.Tok2Vec(
width=width,
embed_size=embed_size,
pretrained_vectors=pretrained_vectors,
bilstm_depth=depth,
conv_depth=0,
cnn_maxout_pieces=0,
)
@registry.architectures.register("tagger_model.v1")
def build_tagger_model_v1(tok2vec):
return component_models.build_tagger_model(nr_class=None, tok2vec=tok2vec)
@registry.architectures.register("transition_based_parser.v1")
def create_tb_parser_model(
tok2vec: Model,
nr_feature_tokens: StrictInt = 3,
hidden_width: StrictInt = 64,
maxout_pieces: StrictInt = 3,
):
from thinc.layers import Linear, chain, list2array
from spacy.ml._layers import PrecomputableAffine
from spacy.syntax._parser_model import ParserModel
from thinc.api import use_ops, zero_init
token_vector_width = tok2vec.get_dim("nO")
tok2vec = chain(tok2vec, list2array())
tok2vec.set_dim("nO", token_vector_width)
lower = PrecomputableAffine(
hidden_width, nF=nr_feature_tokens, nI=tok2vec.get_dim("nO"), nP=maxout_pieces
)
lower.set_dim("nP", maxout_pieces)
with use_ops("numpy"):
# Initialize weights at zero, as it's a classification layer.
upper = Linear(init_W=zero_init)
return ParserModel(tok2vec, lower, upper)
@plac.annotations(
# fmt: off
train_path=("Location of JSON-formatted training data", "positional", None, Path),
dev_path=("Location of JSON-formatted development data", "positional", None, Path),
config_path=("Path to config file", "positional", None, Path),
output_path=("Output directory to store model in", "option", "o", Path),
meta_path=("Optional path to meta.json to use as base.", "option", "m", Path),
raw_text=("Path to jsonl file with unlabelled text documents.", "option", "rt", Path),
# fmt: on
)
def train_from_config_cli(
train_path,
dev_path,
config_path,
output_path=None,
meta_path=None,
raw_text=None,
debug=False,
verbose=False,
):
"""
Train or update a spaCy model. Requires data to be formatted in spaCy's
JSON format. To convert data from other formats, use the `spacy convert`
command.
"""
if not config_path or not config_path.exists():
msg.fail("Config file not found", config_path, exits=1)
if not train_path or not train_path.exists():
msg.fail("Training data not found", train_path, exits=1)
if not dev_path or not dev_path.exists():
msg.fail("Development data not found", dev_path, exits=1)
if meta_path is not None and not meta_path.exists():
msg.fail("Can't find model meta.json", meta_path, exits=1)
if output_path is not None and not output_path.exists():
output_path.mkdir()
try:
train_from_config(
config_path,
{"train": train_path, "dev": dev_path},
output_path=output_path,
meta_path=meta_path,
raw_text=raw_text,
)
except KeyboardInterrupt:
msg.warn("Cancelled.")
def train_from_config(
config_path,
data_paths,
raw_text=None,
meta_path=None,
output_path=None,
):
msg.info("Loading config from: {}".format(config_path))
config = util.load_from_config(config_path, create_objects=True)
use_gpu = config["training"]["use_gpu"]
if use_gpu >= 0:
msg.info("Using GPU")
else:
msg.info("Using CPU")
msg.info("Creating nlp from config")
nlp = create_nlp_from_config(**config["nlp"])
optimizer = config["optimizer"]
limit = config["training"]["limit"]
msg.info("Loading training corpus")
corpus = GoldCorpus(data_paths["train"], data_paths["dev"], limit=limit)
msg.info("Initializing the nlp pipeline")
nlp.begin_training(
lambda: corpus.train_examples, device=use_gpu
)
train_batches = create_train_batches(nlp, corpus, config["training"])
evaluate = create_evaluation_callback(nlp, optimizer, corpus, config["training"])
# Create iterator, which yields out info after each optimization step.
msg.info("Start training")
training_step_iterator = train_while_improving(
nlp,
optimizer,
train_batches,
evaluate,
config["training"]["dropout"],
config["training"]["patience"],
config["training"]["eval_frequency"],
)
msg.info("Training. Initial learn rate: {}".format(optimizer.learn_rate))
print_row = setup_printer(config)
try:
progress = tqdm.tqdm(total=config["training"]["eval_frequency"], leave=False)
for batch, info, is_best_checkpoint in training_step_iterator:
progress.update(1)
if is_best_checkpoint is not None:
progress.close()
print_row(info)
if is_best_checkpoint and output_path is not None:
nlp.to_disk(output_path)
progress = tqdm.tqdm(
total=config["training"]["eval_frequency"], leave=False
)
finally:
if output_path is not None:
with nlp.use_params(optimizer.averages):
final_model_path = output_path / "model-final"
nlp.to_disk(final_model_path)
msg.good("Saved model to output directory", final_model_path)
# with msg.loading("Creating best model..."):
# best_model_path = _collate_best_model(meta, output_path, nlp.pipe_names)
# msg.good("Created best model", best_model_path)
def create_nlp_from_config(lang, vectors, pipeline):
lang_class = spacy.util.get_lang_class(lang)
nlp = lang_class()
if vectors is not None:
spacy.cli.train._load_vectors(nlp, vectors)
for name, component_cfg in pipeline.items():
factory = component_cfg.pop("factory")
component = nlp.create_pipe(factory, config=component_cfg)
nlp.add_pipe(component, name=name)
return nlp
def create_train_batches(nlp, corpus, cfg):
while True:
train_examples = corpus.train_dataset(
nlp,
noise_level=0.0,
orth_variant_level=cfg["orth_variant_level"],
gold_preproc=cfg["gold_preproc"],
max_length=cfg["max_length"],
ignore_misaligned=True,
)
for batch in util.minibatch_by_words(train_examples, size=cfg["batch_size"]):
yield batch
def create_evaluation_callback(nlp, optimizer, corpus, cfg):
def evaluate():
with nlp.use_params(optimizer.averages):
dev_examples = list(
corpus.dev_dataset(
nlp, gold_preproc=cfg["gold_preproc"], ignore_misaligned=True
)
)
scorer = nlp.evaluate(dev_examples)
scores = scorer.scores
# Calculate a weighted sum based on score_weights for the main score
weights = cfg["score_weights"]
weighted_score = sum(scores[s] * weights.get(s, 0.0) for s in weights)
return weighted_score, scorer.scores
return evaluate
def train_while_improving(
nlp, optimizer, train_data, evaluate, dropout, patience, eval_frequency
):
"""Train until an evaluation stops improving. Works as a generator,
with each iteration yielding a tuple `(batch, info, is_best_checkpoint)`,
where info is a dict, and is_best_checkpoint is in [True, False, None] --
None indicating that the iteration was not evaluated as a checkpoint.
The evaluation is conducted by calling the evaluate callback, which should
Positional arguments:
nlp: The spaCy pipeline to evaluate.
train_data (Iterable[Batch]): A generator of batches, with the training
data. Each batch should be a Sized[Tuple[Input, Annot]]. The training
data iterable needs to take care of iterating over the epochs and
shuffling.
evaluate (Callable[[], Tuple[float, Any]]): A callback to perform evaluation.
The callback should take no arguments and return a tuple
`(main_score, other_scores)`. The main_score should be a float where
higher is better. other_scores can be any object.
Every iteration, the function yields out a tuple with:
* batch: A zipped sequence of Tuple[Doc, GoldParse] pairs.
* info: A dict with various information about the last update (see below).
* is_best_checkpoint: A value in None, False, True, indicating whether this
was the best evaluation so far. You should use this to save the model
checkpoints during training. If None, evaluation was not conducted on
that iteration. False means evaluation was conducted, but a previous
evaluation was better.
The info dict provides the following information:
epoch (int): How many passes over the data have been completed.
step (int): How many steps have been completed.
score (float): The main score form the last evaluation.
other_scores: : The other scores from the last evaluation.
loss: The accumulated losses throughout training.
checkpoints: A list of previous results, where each result is a
(score, step, epoch) tuple.
"""
if isinstance(dropout, float):
dropouts = thinc.schedules.constant(dropout)
else:
dropouts = dropout
results = []
losses = {}
for step, batch in enumerate(train_data):
dropout = next(dropouts)
for subbatch in subdivide_batch(batch):
nlp.update(subbatch, drop=dropout, losses=losses, sgd=False)
for name, proc in nlp.pipeline:
if hasattr(proc, "model"):
proc.model.finish_update(optimizer)
optimizer.step_schedules()
if not (step % eval_frequency):
score, other_scores = evaluate()
results.append((score, step))
is_best_checkpoint = score == max(results)[0]
else:
score, other_scores = (None, None)
is_best_checkpoint = None
info = {
"step": step,
"score": score,
"other_scores": other_scores,
"losses": losses,
"checkpoints": results,
}
yield batch, info, is_best_checkpoint
if is_best_checkpoint is not None:
losses = {}
# Stop if no improvement in `patience` updates
best_score, best_step = max(results)
if (step - best_step) >= patience:
break
def subdivide_batch(batch):
return [batch]
def setup_printer(config):
score_cols = config["training"]["scores"]
score_widths = [max(len(col), 6) for col in score_cols]
loss_cols = ["Loss {}".format(pipe) for pipe in config["nlp"]["pipeline"]]
loss_widths = [max(len(col), 8) for col in loss_cols]
table_header = ["#"] + loss_cols + score_cols + ["Score"]
table_header = [col.upper() for col in table_header]
table_widths = [6] + loss_widths + score_widths + [6]
table_aligns = ["r" for _ in table_widths]
msg.row(table_header, widths=table_widths)
msg.row(["-" * width for width in table_widths])
def print_row(info):
losses = [
"{0:.2f}".format(info["losses"].get(col, 0.0))
for col in config["nlp"]["pipeline"]
]
scores = [
"{0:.2f}".format(info["other_scores"].get(col, 0.0))
for col in config["training"]["scores"]
]
data = [info["step"]] + losses + scores + ["{0:.2f}".format(info["score"])]
msg.row(data, widths=table_widths, aligns=table_aligns)
return print_row
@registry.architectures.register("tok2vec_tensors.v1")
def tok2vec_tensors_v1(width):
tok2vec = Tok2VecListener("tok2vec", width=width)
return tok2vec