def setup_task(cls, args, **kwargs):
"""Setup the task (e.g., load dictionaries).
Args:
args (argparse.Namespace): parsed command-line arguments
"""
args.left_pad_source = options.eval_bool(args.left_pad_source)
args.left_pad_target = options.eval_bool(args.left_pad_target)
if getattr(args, 'raw_text', False):
utils.deprecation_warning(
'--raw-text is deprecated, please use --dataset-impl=raw')
args.dataset_impl = 'raw'
elif getattr(args, 'lazy_load', False):
utils.deprecation_warning(
'--lazy-load is deprecated, please use --dataset-impl=lazy')
args.dataset_impl = 'lazy'
paths = args.data.split(':')
assert len(paths) > 0
# find language pair automatically
if args.source_lang is None or args.target_lang is None:
args.source_lang, args.target_lang = data_utils.infer_language_pair(
paths[0])
if args.source_lang is None or args.target_lang is None:
raise Exception(
'Could not infer language pair, please provide it explicitly')
# load dictionaries
src_dict = cls.load_dictionary(
os.path.join(paths[0], 'dict.{}.txt'.format(args.source_lang)))
tgt_dict = cls.load_dictionary(
os.path.join(paths[0], 'dict.{}.txt'.format(args.target_lang)))
assert src_dict.pad() == tgt_dict.pad()
assert src_dict.eos() == tgt_dict.eos()
assert src_dict.unk() == tgt_dict.unk()
print('| [{}] dictionary: {} types'.format(args.source_lang,
len(src_dict)))
print('| [{}] dictionary: {} types'.format(args.target_lang,
len(tgt_dict)))
return cls(args, src_dict, tgt_dict)
def get_meter(self, name):
"""[deprecated] Get a specific meter by name."""
from fairseq import meters
if "get_meter" not in self._warn_once:
self._warn_once.add("get_meter")
utils.deprecation_warning(
"Trainer.get_meter is deprecated. Please use fairseq.metrics instead."
)
train_meters = metrics.get_meters("train")
if train_meters is None:
train_meters = {}
if name == "train_loss" and "loss" in train_meters:
return train_meters["loss"]
elif name == "train_nll_loss":
# support for legacy train.py, which assumed this meter is
# always initialized
m = train_meters.get("nll_loss", None)
return m or meters.AverageMeter()
elif name == "wall":
# support for legacy train.py, which assumed this meter is
# always initialized
m = metrics.get_meter("default", "wall")
return m or meters.TimeMeter()
elif name == "wps":
m = metrics.get_meter("train", "wps")
return m or meters.TimeMeter()
elif name in {"valid_loss", "valid_nll_loss"}:
# support for legacy train.py, which assumed these meters
# are always initialized
k = name[len("valid_"):]
m = metrics.get_meter("valid", k)
return m or meters.AverageMeter()
elif name == "oom":
return meters.AverageMeter()
elif name in train_meters:
return train_meters[name]
return None
def setup_task(cls, args, **kwargs):
"""Setup the task (e.g., load dictionaries).
Args:
args (argparse.Namespace): parsed command-line arguments
"""
if getattr(args, 'raw_text', False):
utils.deprecation_warning('--raw-text is deprecated, please use --dataset-impl=raw')
args.dataset_impl = 'raw'
elif getattr(args, 'lazy_load', False):
utils.deprecation_warning('--lazy-load is deprecated, please use --dataset-impl=lazy')
args.dataset_impl = 'lazy'
dictionary = None
output_dictionary = None
if args.data:
paths = args.data.split(':')
assert len(paths) > 0
dictionary = Dictionary.load(os.path.join(paths[0], 'dict.txt'))
print('| dictionary: {} types'.format(len(dictionary)))
output_dictionary = dictionary
if args.output_dictionary_size >= 0:
output_dictionary = TruncatedDictionary(dictionary, args.output_dictionary_size)
# upgrade old checkpoints
if hasattr(args, 'exclude_self_target'):
args.self_target = not args.exclude_self_target
targets = []
if getattr(args, 'self_target', False):
targets.append('self')
if getattr(args, 'future_target', False):
targets.append('future')
if getattr(args, 'past_target', False):
targets.append('past')
if len(targets) == 0:
# standard language modeling
targets = ['future']
return cls(args, dictionary, output_dictionary, targets=targets)
def reduce_metrics(self, logging_outputs, criterion):
"""Aggregate logging outputs from data parallel training."""
# backward compatibility for tasks that override aggregate_logging_outputs
base_func = FairseqTask.aggregate_logging_outputs
self_func = getattr(self, 'aggregate_logging_outputs').__func__
if self_func is not base_func:
utils.deprecation_warning(
'Tasks should implement the reduce_metrics API. '
'Falling back to deprecated aggregate_logging_outputs API.')
agg_logging_outputs = self.aggregate_logging_outputs(
logging_outputs, criterion)
for k, v in agg_logging_outputs.items():
metrics.log_scalar(k, v)
return
if not any('ntokens' in log for log in logging_outputs):
warnings.warn(
'ntokens not found in Criterion logging outputs, cannot log wpb or wps'
)
else:
ntokens = utils.item(
sum(log.get('ntokens', 0) for log in logging_outputs))
metrics.log_scalar('wpb', ntokens, priority=180, round=1)
metrics.log_speed('wps',
ntokens,
ignore_first=10,
priority=90,
round=1)
if not any('nsentences' in log for log in logging_outputs):
warnings.warn(
'nsentences not found in Criterion logging outputs, cannot log bsz'
)
else:
nsentences = utils.item(
sum(log.get('nsentences', 0) for log in logging_outputs))
metrics.log_scalar('bsz', nsentences, priority=190, round=1)
criterion.__class__.reduce_metrics(logging_outputs)
def setup_task(cls, args, **kwargs):
"""Setup the task (e.g., load dictionaries).
Args:
args (argparse.Namespace): parsed command-line arguments
"""
args.left_pad_source = options.eval_bool(args.left_pad_source)
args.left_pad_target = options.eval_bool(args.left_pad_target)
if getattr(args, 'raw_text', False):
utils.deprecation_warning('--raw-text is deprecated, please use --dataset-impl=raw')
args.dataset_impl = 'raw'
elif getattr(args, 'lazy_load', False):
utils.deprecation_warning('--lazy-load is deprecated, please use --dataset-impl=lazy')
args.dataset_impl = 'lazy'
paths = args.data.split(':')
assert len(paths) > 0
# find language pair automatically
if args.source_lang is None or args.target_lang is None:
args.source_lang, args.target_lang = data_utils.infer_language_pair(paths[0])
if args.source_lang is None or args.target_lang is None:
raise Exception('Could not infer language pair, please provide it explicitly')
# load dictionaries
assert args.target_lang == 'actions', 'target extension must be "actions"'
args.target_lang_nopos = 'actions_nopos' # only build dictionary without pointer values
args.target_lang_pos = 'actions_pos'
args.target_lang_vocab_nodes = 'actions.vocab.nodes'
args.target_lang_vocab_others = 'actions.vocab.others'
src_dict = cls.load_dictionary(os.path.join(paths[0], 'dict.{}.txt'.format(args.source_lang)))
# tgt_dict = cls.load_dictionary(os.path.join(paths[0], 'dict.{}.txt'.format(args.target_lang_nopos)))
# NOTE rebuild the dictionary every time
tgt_dict = cls.build_dictionary_bart_extend(
node_freq_min=args.node_freq_min,
node_file_path=os.path.join(paths[0], args.target_lang_vocab_nodes),
others_file_path=os.path.join(paths[0], args.target_lang_vocab_others)
)
# TODO target dictionary 'actions_nopos' is hard coded now; change it later
assert src_dict.pad() == tgt_dict.pad()
assert src_dict.eos() == tgt_dict.eos()
assert src_dict.unk() == tgt_dict.unk()
print('| [{}] dictionary: {} types'.format(args.source_lang, len(src_dict)))
print('| [{}] dictionary: {} types'.format(args.target_lang_nopos, len(tgt_dict)))
# ========== load the pretrained BART model ==========
if getattr(args, 'arch', None):
# training time: pretrained BART needs to be used for initialization
if 'bart_base' in args.arch or 'bartsv_base' in args.arch:
print('-' * 10 + ' loading pretrained bart.base model ' + '-' * 10)
bart = torch.hub.load('pytorch/fairseq', 'bart.base')
elif 'bart_large' in args.arch or 'bartsv_large' in args.arch:
print('-' * 10 + 'loading pretrained bart.large model ' + '-' * 10)
bart = torch.hub.load('pytorch/fairseq', 'bart.large')
else:
raise ValueError
else:
# inference time: pretrained BART is only used for dictionary related things; size does not matter
# NOTE size does matter; update this later in model initialization if model is with "bart.large"
print('-' * 10 + ' (for bpe vocab and embed size at inference time) loading pretrained bart.base model '
+ '-' * 10)
bart = torch.hub.load('pytorch/fairseq', 'bart.base')
bart.eval() # the pretrained BART model is only for assistance
# ====================================================
return cls(args, src_dict, tgt_dict, bart)
def setup_task(cls, args, **kwargs):
"""Setup the task (e.g., load dictionaries).
Args:
args (argparse.Namespace): parsed command-line arguments
"""
print("These are the arguments", args)
print("\n")
if getattr(args, 'raw_text', False):
utils.deprecation_warning(
'--raw-text is deprecated, please use --dataset-impl=raw')
args.dataset_impl = 'raw'
elif getattr(args, 'lazy_load', False):
utils.deprecation_warning(
'--lazy-load is deprecated, please use --dataset-impl=lazy')
args.dataset_impl = 'lazy'
dictionary = None
output_dictionary = None
def read_config(path):
with open(path) as config:
import yaml
contents = config.read()
data = yaml.load(contents)
return data
# if args.data:
# paths = args.data.split(':')
# assert len(paths) > 0
# dictionary = Dictionary.load(os.path.join(paths[0], 'dict.txt'))
# print('| dictionary: {} types'.format(len(dictionary)))
# output_dictionary = dictionary
# if args.output_dictionary_size >= 0:
# output_dictionary = TruncatedDictionary(dictionary, args.output_dictionary_size)
######################### CHANGE THIS #################################################
args.data = read_config(
"/content/drive/My Drive/IIIT-H RA/ICON/fairseq-working/config.yaml"
)
########################################################################################
if (args.data):
path = args.data['dictionary']['src']
dictionary = Dictionary.load(path)
print('| dictionary: {} types'.format(len(dictionary)))
output_dictionary = dictionary
if (args.output_dictionary_size >= 0):
output_dictionary = TruncatedDictionary(
dictionary, args.output_dictionary_size)
# upgrade old checkpoints
if hasattr(args, 'exclude_self_target'):
args.self_target = not args.exclude_self_target
targets = []
if getattr(args, 'self_target', False):
targets.append('self')
if getattr(args, 'future_target', False):
targets.append('future')
if getattr(args, 'past_target', False):
targets.append('past')
if len(targets) == 0:
# standard language modeling
targets = ['future']
return cls(args,
args.data,
dictionary,
output_dictionary,
targets=targets)
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
utils.deprecation_warning('hello', stacklevel=4)
def setup_task(cls, args, **kwargs):
"""Setup the task (e.g., load dictionaries).
Args:
args (argparse.Namespace): parsed command-line arguments
"""
args.left_pad_source = options.eval_bool(args.left_pad_source)
args.left_pad_target = options.eval_bool(args.left_pad_target)
if getattr(args, 'raw_text', False):
utils.deprecation_warning(
'--raw-text is deprecated, please use --dataset-impl=raw')
args.dataset_impl = 'raw'
elif getattr(args, 'lazy_load', False):
utils.deprecation_warning(
'--lazy-load is deprecated, please use --dataset-impl=lazy')
args.dataset_impl = 'lazy'
paths = args.data.split(':')
assert len(paths) > 0
# find language pair automatically
if args.source_lang is None or args.target_lang is None:
raise Exception(
'Could not infer language pair, please provide it explicitly'
) # See comment below?
# TRY ADDING: --source-lang true_w,true_p --target-lang reco_w
# OR ENABLING BELOW HARDCODED DEFAULTS.
# args.source_lang = "true_w,true_p"
# args.target_lang = "reco_w"
# args.source_lang, args.target_lang = data_utils.infer_language_pair(paths[0])
assert type(args.source_lang) == str
if ',' not in args.source_lang:
raise Exception(
"source-lang is " + args.source_lang +
" source-lang needs to contain two comma separated strings")
# load dictionaries
src_lang1, src_lang2 = args.source_lang.split(',')
src_dict1 = cls.load_dictionary(
os.path.join(paths[0], 'dict.{}.txt'.format(src_lang1)))
src_dict2 = cls.load_dictionary(
os.path.join(paths[0], 'dict.{}.txt'.format(src_lang2)))
if ',' in args.target_lang:
assert args.criterion == 'cross_entropy_dual'
target_lang, target_lang_extra = args.target_lang.split(',')
tgt_dict = cls.load_dictionary(
os.path.join(paths[0], 'dict.{}.txt'.format(target_lang)))
tgt_dict_extra = cls.load_dictionary(
os.path.join(paths[0],
'dict.{}.txt'.format(target_lang_extra)))
dual_decoder = True
else:
tgt_dict = cls.load_dictionary(
os.path.join(paths[0], 'dict.{}.txt'.format(args.target_lang)))
dual_decoder = False
tgt_dict_extra = None
assert src_dict1.pad() == tgt_dict.pad()
assert src_dict1.eos() == tgt_dict.eos()
assert src_dict1.unk() == tgt_dict.unk()
assert src_dict2.pad() == tgt_dict.pad()
assert src_dict2.eos() == tgt_dict.eos()
assert src_dict2.unk() == tgt_dict.unk()
print('| [{}] dictionary: {} types'.format(src_lang1, len(src_dict1)))
print('| [{}] dictionary: {} types'.format(src_lang2, len(src_dict2)))
if dual_decoder:
print('| [{}] dictionary: {} types'.format(target_lang,
len(tgt_dict)))
print('| [{}] dictionary: {} types'.format(target_lang_extra,
len(tgt_dict_extra)))
else:
print('| [{}] dictionary: {} types'.format(args.target_lang,
len(tgt_dict)))
return cls(args, src_dict1, src_dict2, tgt_dict, tgt_dict_extra)
