Merge branch 'master' into change-dash-m-method

Author: ramon-astudillo

src/transition_amr_parser/parse.py

@@ -61,7 +61,8 @@ def argument_parsing():
  parser.add_argument(
  '--in-doc',
  type=str,
- help='File with one __tokenized__ sentence per line and one newline separating each doc (and the end)'
+ help='File with one __tokenized__ sentence per line and one newline '
+ 'separating each doc (and the end)'
  )
  parser.add_argument(
  '--in-amr',
@@ -219,7 +220,8 @@ def argument_parsing():
  assert (
  bool(args.in_tokenized_sentences) or bool(args.in_amr)
  ) or bool(args.service) or bool(args.in_doc), \
- "Must either specify --in-tokenized-sentences or --in-doc or set --service"
+ "Must either specify --in-tokenized-sentences or --in-doc or set " \
+ "--service"
 
  if not (bool(args.model_name) ^ bool(args.in_checkpoint)):
  raise Exception("Use either --model-name or --in-checkpoint")
@@ -231,7 +233,10 @@ def argument_parsing():
  )
  if bool(args.in_actions) and bool(args.in_doc):
  print(yellow_font(
- 'WARNING: Given force actions will be used superseding the force actions generated by the inhouse force action generator. Make sure it is in the correct format '))
+ 'WARNING: Given force actions will be used superseding the force '
+ 'actions generated by the inhouse force action generator. Make '
+ 'sure it is in the correct format '
+ ))
 
  # num samples replaces beam search
  if args.num_samples:
@@ -605,14 +610,14 @@ def from_pretrained(cls, model_name, dict_dir=None,
  roberta_cache_path=None, fp16=False,
  inspector=None, beam=1, nbest=1, num_samples=None,
  sampling_topp=-1, temperature=1.0):
- """ Load model checkpoints from available model names; 
- Will check if the model is downloaded to cache, if not, download from cloud storage; 
+ """ Load model checkpoints from available model names;
+ Will check if the model is downloaded to cache, if not, download from cloud storage;
  Below is a list of available modelnames to trun
  {
  'AMR3.0':'https://s3.us-east.cloud-object-storage.appdomain.cloud/cloud-object-storage-xc-cos-standard-htg/amr3.0-structured-bart-large-neur-al-sampling5.zip'
  }
  Args:
- modelname (str): a model name within our pretrained model library. 
+ modelname (str): a model name within our pretrained model library.
  dict_dir (_type_, optional): _description_. Defaults to None.
  roberta_cache_path (_type_, optional): _description_. Defaults to None.
  fp16 (bool, optional): _description_. Defaults to False.
@@ -1263,7 +1268,6 @@ def run_service(args, parser):
 
 def prepare_data(args, parser):
 
- force_actions = None
  if args.in_amr:
 
  # align mode: read input AMR to be aligned
@@ -1398,8 +1402,8 @@ def main():
  if args.in_actions:
  with open(args.in_actions) as fact:
  force_actions = [eval(line.strip()) + [[]] for line in fact]
- assert len(tok_sentences) == len(
- force_actions), "Number of force actions doesn't match the number of sentences"
+ assert len(tok_sentences) == len(force_actions), \
+ "Number of force actions doesn't match the number of sentences"
 
  # sampling needs copy of force actions N times
  if args.num_samples is not None:
@@ -1416,15 +1420,17 @@ def main():
  start = time.time()
 
  if args.in_doc:
- annotations, machines = parser.parse_docs(tok_sentences,
- gold_amrs=gold_amrs,
- window_size=args.window_size,
- window_overlap=args.window_overlap,
- batch_size=args.batch_size,
- roberta_batch_size=args.roberta_batch_size,
- beam=args.beam,
- jamr=args.jamr,
- no_isi=args.no_isi)
+ annotations, machines = parser.parse_docs(
+ tok_sentences,
+ gold_amrs=gold_amrs,
+ window_size=args.window_size,
+ window_overlap=args.window_overlap,
+ batch_size=args.batch_size,
+ roberta_batch_size=args.roberta_batch_size,
+ beam=args.beam,
+ jamr=args.jamr,
+ no_isi=args.no_isi
+ )
 
  else: