def main(args):
# If output_dir not provided, a folder will be generated in pwd
if not args.output_dir:
args.output_dir = os.path.join(
"./results",
f"{args.task}_{time.strftime('%Y%m%d_%H%M%S')}",
)
os.makedirs(args.output_dir)
model = SummarizationTrainer(args)
trainer = generic_train(model, args)
# Optionally, predict on dev set and write to output_dir
if args.do_predict:
# See https://github.com/huggingface/transformers/issues/3159
# pl use this format to create a checkpoint:
# https://github.com/PyTorchLightning/pytorch-lightning/blob/master\
# /pytorch_lightning/callbacks/model_checkpoint.py#L169
checkpoints = list(
sorted(
glob.glob(os.path.join(args.output_dir,
"checkpointepoch=*.ckpt"),
recursive=True)))
model = model.load_from_checkpoint(checkpoints[-1])
trainer.test(model)
def main(args):
# If output_dir not provided, a folder will be generated in pwd
if not args.output_dir:
args.output_dir = os.path.join("./results", f"{args.task}_{args.model_type}_{time.strftime('%Y%m%d_%H%M%S')}",)
os.makedirs(args.output_dir)
model = SummarizationTrainer(args)
trainer = generic_train(model, args)
# Optionally, predict on dev set and write to output_dir
if args.do_predict:
checkpoints = list(sorted(glob.glob(os.path.join(args.output_dir, "checkpointepoch=*.ckpt"), recursive=True)))
SummarizationTrainer.load_from_checkpoint(checkpoints[-1])
trainer.test(model)
parser.add_argument(
"--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
)
parser.add_argument(
"--tags", nargs='+', type=str, help="experiment tags for neptune.ai", default=['FT', 'last-layer']
)
return parser
if __name__ == "__main__":
parser = argparse.ArgumentParser()
add_generic_args(parser, os.getcwd())
parser = GLUETransformer.add_model_specific_args(parser, os.getcwd())
args = parser.parse_args()
# If output_dir not provided, a folder will be generated in pwd
if args.output_dir is None:
args.output_dir = os.path.join("./results", f"{args.task}_{time.strftime('%Y%m%d_%H%M%S')}",)
os.makedirs(args.output_dir)
model = GLUETransformer(args)
trainer = generic_train(model, args)
# Optionally, predict on dev set and write to output_dir
if args.do_predict:
checkpoints = list(sorted(glob.glob(os.path.join(args.output_dir, "checkpointepoch=*.ckpt"), recursive=True)))
model = model.load_from_checkpoint(checkpoints[-1])
trainer.test(model)
def main(args):
# If output_dir not provided, a folder will be generated in pwd
if not args.output_dir:
args.output_dir = os.path.join(
"./results",
f"{args.task}_{time.strftime('%Y%m%d_%H%M%S')}",
)
os.makedirs(args.output_dir)
model = SummarizationTrainer(args)
trainer = generic_train(model, args)
# Optionally, predict on dev set and write to output_dir
if args.do_predict:
# See https://github.com/huggingface/transformers/issues/3159
# pl use this format to create a checkpoint:
# https://github.com/PyTorchLightning/pytorch-lightning/blob/master\
# /pytorch_lightning/callbacks/model_checkpoint.py#L169
checkpoints = list(
sorted(
glob.glob(os.path.join(args.output_dir,
"checkpointepoch=*.ckpt"),
recursive=True)))
print(str(checkpoints))
model = model.load_from_checkpoint(checkpoints[-1])
# trainer.test(model)
tokenizer = T5Tokenizer.from_pretrained(args.model_name_or_path)
test_examples = [
x.rstrip() for x in open('./csqa.train.qac.src').readlines()
]
test_fout = open('train_csqa.txt', 'w')
val_examples = [
x.rstrip() for x in open('./csqa.dev.qac.src').readlines()
]
val_fout = open('val_csqa.txt', 'w')
max_length = 24
min_length = 1
def chunks(lst, n):
for i in range(0, len(lst), n):
yield lst[i:i + n]
device = "cuda" if torch.cuda.is_available() else "cpu"
print(device)
model.to(device)
for batch in tqdm(list(chunks(test_examples, 8))):
dct = tokenizer.batch_encode_plus(batch,
max_length=64,
return_tensors="pt",
pad_to_max_length=True)
summaries = model.model.generate(
input_ids=dct["input_ids"].to(device),
attention_mask=dct["attention_mask"].to(device),
num_beams=5,
length_penalty=0.6,
max_length=max_length +
2, # +2 from original because we start at step=1 and stop before max_length
min_length=min_length +
1, # +1 from original because we start at step=1
no_repeat_ngram_size=3,
early_stopping=True,
decoder_start_token_id=model.config.eos_token_id,
)
dec = [
tokenizer.decode(g,
skip_special_tokens=True,
clean_up_tokenization_spaces=False)
for g in summaries
]
for hypothesis in dec:
test_fout.write(hypothesis + "\n")
test_fout.flush()
for batch in tqdm(list(chunks(val_examples, 8))):
dct = tokenizer.batch_encode_plus(batch,
max_length=64,
return_tensors="pt",
pad_to_max_length=True)
summaries = model.model.generate(
input_ids=dct["input_ids"].to(device),
attention_mask=dct["attention_mask"].to(device),
num_beams=5,
length_penalty=0.6,
max_length=max_length +
2, # +2 from original because we start at step=1 and stop before max_length
min_length=min_length +
1, # +1 from original because we start at step=1
no_repeat_ngram_size=3,
early_stopping=True,
decoder_start_token_id=model.config.eos_token_id,
)
dec = [
tokenizer.decode(g,
skip_special_tokens=True,
clean_up_tokenization_spaces=False)
for g in summaries
]
for hypothesis in dec:
val_fout.write(hypothesis + "\n")
val_fout.flush()
