def loadRobertaCheckpoint(pathBERTCheckpoint, pathData, from_pretrained=False):
"""
Load Roberta model from checkpoint.
If load a pretrained model from fairseq, set from_pretrained=True.
"""
if from_pretrained: # Require connection to download bpe, possible errors for trained checkpoint that contains cfg
roberta = RobertaModel.from_pretrained(dirname(pathBERTCheckpoint),
basename(pathBERTCheckpoint),
pathData)
else:
# Set up the args Namespace
model_args = argparse.Namespace(task='masked_lm',
seed=-1,
output_dictionary_size=-1,
data=pathData,
path=pathBERTCheckpoint)
# Setup task
task = tasks.setup_task(model_args)
# Load model
models, _model_args = checkpoint_utils.load_model_ensemble(
[model_args.path], task=task)
model = models[0]
# Wrap-up to RobertaHubInterface (to be consistent with RobertaModel.from_pretrained)
roberta = RobertaHubInterface(_model_args, task, model)
return roberta
def _load_model(self, path: str, bpe: str, bpe_filename:str) -> RobertaHubInterface:
if path == "xlmr.large" or path == "xlmr.base":
return hub.load("pytorch/fairseq", path, force_reload=True)
else:
checkpoint_file = "model.pt" if os.path.exists(os.path.join(path, "model.pt")) else "checkpoint_best.pt"
loaded = hub_utils.from_pretrained(
model_name_or_path=path,
checkpoint_file=checkpoint_file,
data_name_or_path=path,
bpe=bpe,
sentencepiece_vocab=os.path.join(path, bpe_filename),
sentencepiece_model=os.path.join(path, bpe_filename),
load_checkpoint_heads=True,
archive_map=RobertaModel.hub_models(),
cpu=False
)
return RobertaHubInterface(loaded['args'], loaded['task'], loaded['models'][0])
def evaluate_task(self):
checkpoints_output_dir = os.path.join("checkpoints", self.model_name,
self.task.spec().output_path())
checkpoint_file = "checkpoint_last.pt" if self.task.spec(
).no_dev_set else "checkpoint_best.pt"
loaded = hub_utils.from_pretrained(
model_name_or_path=checkpoints_output_dir,
checkpoint_file=checkpoint_file,
data_name_or_path=self.task_output_dir,
bpe="sentencepiece",
sentencepiece_vocab=os.path.join(self.model_dir,
"sentencepiece.bpe.model"),
load_checkpoint_heads=True,
archive_map=RobertaModel.hub_models())
roberta = RobertaHubInterface(loaded['args'], loaded['task'],
loaded['models'][0])
evaluator = TaskEvaluator(self.task, self.task_id, roberta,
self.input_dir, checkpoints_output_dir)
return evaluator.evaluate()
from fairseq import hub_utils
from fairseq.models.roberta import RobertaModel, RobertaHubInterface
import os
from tqdm import tqdm
model_path = "polish_roberta_large_no_finetune"
loaded = hub_utils.from_pretrained(model_name_or_path=model_path,
data_name_or_path=model_path,
bpe="sentencepiece",
sentencepiece_vocab=os.path.join(
model_path, "sentencepiece.bpe.model"),
load_checkpoint_heads=True,
archive_map=RobertaModel.hub_models(),
cpu=False)
roberta = RobertaHubInterface(loaded['args'], loaded['task'],
loaded['models'][0])
roberta.eval()
roberta.cuda()
preds = roberta.fill_mask('Ala <mask>, kota', topk=3)
#import pdb; pdb.set_trace()
def predict(f_in_path, f_out_path):
f_in = open(f_in_path, 'r', newline='\n')
f_out = open(f_out_path, 'w', newline='\n')
for line in tqdm(f_in, total=19986):
_, _, before, after = line.split('\t')
before = ' '.join(
vocab_model_file="wikipedia_upper_voc_32000_sen10000000.model"
vocab_path = os.path.join(root_path, "vocab", vocab_model_file)
#%%
loaded = hub_utils.from_pretrained(
model_name_or_path=model_path,
checkpoint_file=checkpoint_file,
data_name_or_path='./',
bpe="sentencepiece",
sentencepiece_vocab=vocab_path,
load_checkpoint_heads=True,
archive_map=RobertaModel.hub_models(),
cpu=True
)
roberta = RobertaHubInterface(loaded['args'], loaded['task'], loaded['models'][0])
roberta.eval()
#%%
def print_mask(s, predicted):
print(s)
for p in predicted:
print(f'\t{p[2]} - {p[0]} - confidence {p[1]}')
sentences = [
'Bolesław Bierut objął rządy w <mask> roku.', #1948
'Największym <mask> we współczesnym świecie jest głód.',
'Wikipedia powstała jako projekt uzupełniający dla <mask>, darmowej encyklopedii internetowej', #Nupedii
'W Olsztynie pracował Mikołaj Kopernik, ten który <mask> ziemię a wstrzymał słońce.',