from horovod.spark.lightning import TorchEstimator
import torch
from pyspark.ml.param.shared import Param, Params
from pytorch_lightning.utilities import _module_available
from synapse.ml.dl.DeepTextModel import DeepTextModel
from synapse.ml.dl.LitDeepTextModel import LitDeepTextModel
from synapse.ml.dl.utils import keywords_catch, get_or_create_backend
from synapse.ml.dl.PredictionParams import TextPredictionParams
_TRANSFORMERS_AVAILABLE = _module_available("transformers")
if _TRANSFORMERS_AVAILABLE:
import transformers
_TRANSFORMERS_EQUAL_4_15_0 = transformers.__version__ == "4.15.0"
if _TRANSFORMERS_EQUAL_4_15_0:
from transformers import AutoTokenizer
else:
raise RuntimeError(
"transformers should be == 4.15.0, found: {}".format(
transformers.__version__
)
)
else:
raise ModuleNotFoundError("module not found: transformers")
[docs]class DeepTextClassifier(TorchEstimator, TextPredictionParams):
checkpoint = Param(
Params._dummy(), "checkpoint", "checkpoint of the deep text classifier"
)
additional_layers_to_train = Param(
Params._dummy(),
"additional_layers_to_train",
"number of last layers to fine tune for the model, should be larger or equal to 0. default to 3.",
)
num_classes = Param(Params._dummy(), "num_classes", "number of target classes")
loss_name = Param(
Params._dummy(),
"loss_name",
"string representation of torch.nn.functional loss function for the underlying pytorch_lightning model, e.g. binary_cross_entropy",
)
optimizer_name = Param(
Params._dummy(),
"optimizer_name",
"string representation of optimizer function for the underlying pytorch_lightning model",
)
tokenizer = Param(Params._dummy(), "tokenizer", "tokenizer")
max_token_len = Param(Params._dummy(), "max_token_len", "max_token_len")
learning_rate = Param(
Params._dummy(), "learning_rate", "learning rate to be used for the optimizer"
)
train_from_scratch = Param(
Params._dummy(),
"train_from_scratch",
"whether to train the model from scratch or not, if set to False then param additional_layers_to_train need to be specified.",
)
@keywords_catch
def __init__(
self,
checkpoint=None,
additional_layers_to_train=3, # this is needed otherwise the performance is usually bad
num_classes=None,
optimizer_name="adam",
loss_name="cross_entropy",
tokenizer=None,
max_token_len=128,
learning_rate=None,
train_from_scratch=True,
# Classifier args
label_col="label",
text_col="text",
prediction_col="prediction",
# TorchEstimator args
num_proc=None,
backend=None,
store=None,
metrics=None,
loss_weights=None,
sample_weight_col=None,
gradient_compression=None,
input_shapes=None,
validation=None,
callbacks=None,
batch_size=None,
val_batch_size=None,
epochs=None,
verbose=1,
random_seed=None,
shuffle_buffer_size=None,
partitions_per_process=None,
run_id=None,
train_minibatch_fn=None,
train_steps_per_epoch=None,
validation_steps_per_epoch=None,
transformation_fn=None,
transformation_edit_fields=None,
transformation_removed_fields=None,
train_reader_num_workers=None,
trainer_args=None,
val_reader_num_workers=None,
reader_pool_type=None,
label_shapes=None,
inmemory_cache_all=False,
num_gpus=None,
logger=None,
log_every_n_steps=50,
data_module=None,
loader_num_epochs=None,
terminate_on_nan=False,
profiler=None,
debug_data_loader=False,
train_async_data_loader_queue_size=None,
val_async_data_loader_queue_size=None,
use_gpu=True,
mp_start_method=None,
):
super(DeepTextClassifier, self).__init__()
self._setDefault(
checkpoint=None,
additional_layers_to_train=3,
num_classes=None,
optimizer_name="adam",
loss_name="cross_entropy",
tokenizer=None,
max_token_len=128,
learning_rate=None,
train_from_scratch=True,
feature_cols=["text"],
label_cols=["label"],
label_col="label",
text_col="text",
prediction_col="prediction",
)
kwargs = self._kwargs
self._set(**kwargs)
self._update_cols()
self._update_transformation_fn()
model = LitDeepTextModel(
checkpoint=self.getCheckpoint(),
additional_layers_to_train=self.getAdditionalLayersToTrain(),
num_labels=self.getNumClasses(),
optimizer_name=self.getOptimizerName(),
loss_name=self.getLossName(),
label_col=self.getLabelCol(),
text_col=self.getTextCol(),
learning_rate=self.getLearningRate(),
train_from_scratch=self.getTrainFromScratch(),
)
self._set(model=model)
[docs] def setCheckpoint(self, value):
return self._set(checkpoint=value)
[docs] def getCheckpoint(self):
return self.getOrDefault(self.checkpoint)
[docs] def setAdditionalLayersToTrain(self, value):
return self._set(additional_layers_to_train=value)
[docs] def getAdditionalLayersToTrain(self):
return self.getOrDefault(self.additional_layers_to_train)
[docs] def setNumClasses(self, value):
return self._set(num_classes=value)
[docs] def getNumClasses(self):
return self.getOrDefault(self.num_classes)
[docs] def setLossName(self, value):
return self._set(loss_name=value)
[docs] def getLossName(self):
return self.getOrDefault(self.loss_name)
[docs] def setOptimizerName(self, value):
return self._set(optimizer_name=value)
[docs] def getOptimizerName(self):
return self.getOrDefault(self.optimizer_name)
[docs] def setTokenizer(self, value):
return self._set(tokenizer=value)
[docs] def getTokenizer(self):
return self.getOrDefault(self.tokenizer)
[docs] def setMaxTokenLen(self, value):
return self._set(max_token_len=value)
[docs] def getMaxTokenLen(self):
return self.getOrDefault(self.max_token_len)
[docs] def setLearningRate(self, value):
return self._set(learning_rate=value)
[docs] def getLearningRate(self):
return self.getOrDefault(self.learning_rate)
[docs] def setTrainFromScratch(self, value):
return self._set(train_from_scratch=value)
[docs] def getTrainFromScratch(self):
return self.getOrDefault(self.train_from_scratch)
def _update_cols(self):
self.setFeatureCols([self.getTextCol()])
self.setLabelCols([self.getLabelCol()])
def _fit(self, dataset):
return super()._fit(dataset)
# override this method to provide a correct default backend
def _get_or_create_backend(self):
return get_or_create_backend(
self.getBackend(), self.getNumProc(), self.getVerbose(), self.getUseGpu()
)
def _update_transformation_fn(self):
text_col = self.getTextCol()
label_col = self.getLabelCol()
max_token_len = self.getMaxTokenLen()
# load it inside to avoid `Already borrowed` error (https://github.com/huggingface/tokenizers/issues/537)
if self.getTokenizer() is None:
self.setTokenizer(AutoTokenizer.from_pretrained(self.getCheckpoint()))
tokenizer = self.getTokenizer()
def _encoding_text(row):
text = row[text_col]
label = row[label_col]
encoding = tokenizer(
text,
max_length=max_token_len,
padding="max_length",
truncation=True,
return_attention_mask=True,
return_tensors="pt",
)
input_ids = encoding["input_ids"].flatten().numpy()
attention_mask = encoding["attention_mask"].flatten().numpy()
return {
"input_ids": input_ids,
"attention_mask": attention_mask,
"labels": torch.tensor(label, dtype=int),
}
transformation_edit_fields = [
("input_ids", int, None, True),
("attention_mask", int, None, True),
("labels", int, None, False),
]
self.setTransformationEditFields(transformation_edit_fields)
transformation_removed_fields = [self.getTextCol(), self.getLabelCol()]
self.setTransformationRemovedFields(transformation_removed_fields)
self.setTransformationFn(_encoding_text)
[docs] def get_model_class(self):
return DeepTextModel
def _get_model_kwargs(self, model, history, optimizer, run_id, metadata):
return dict(
history=history,
model=model,
optimizer=optimizer,
input_shapes=self.getInputShapes(),
run_id=run_id,
_metadata=metadata,
loss=self.getLoss(),
loss_constructors=self.getLossConstructors(),
tokenizer=self.getTokenizer(),
checkpoint=self.getCheckpoint(),
max_token_len=self.getMaxTokenLen(),
label_col=self.getLabelCol(),
text_col=self.getTextCol(),
prediction_col=self.getPredictionCol(),
)