def test_generation(self):
stream_handler = logging.StreamHandler(sys.stdout)
logger.addHandler(stream_handler)
testargs = ["run_generation.py", "--prompt=Hello", "--length=10", "--seed=42"]
model_type, model_name = ("--model_type=openai-gpt", "--model_name_or_path=openai-gpt")
with patch.object(sys, "argv", testargs + [model_type, model_name]):
result = run_generation.main()
def test_generation(self):
stream_handler = logging.StreamHandler(sys.stdout)
logger.addHandler(stream_handler)
testargs = ["run_generation.py", "--prompt=Hello", "--length=10", "--seed=42"]
model_type, model_name = ("--model_type=gpt2", "--model_name_or_path=sshleifer/tiny-gpt2")
with patch.object(sys, "argv", testargs + [model_type, model_name]):
result = run_generation.main()
self.assertGreaterEqual(len(result[0]), 10)
def test_generation(self):
testargs = ["run_generation.py", "--prompt=Hello", "--length=10", "--seed=42"]
if is_cuda_and_apex_available():
testargs.append("--fp16")
model_type, model_name = (
"--model_type=gpt2",
"--model_name_or_path=sshleifer/tiny-gpt2",
)
with patch.object(sys, "argv", testargs + [model_type, model_name]):
result = run_generation.main()
self.assertGreaterEqual(len(result[0]), 10)
def generate_text(params):
"""Generate text using transformers."""
prompt = ''
if not params['genre'] and not params['title'] and not params['prefix']:
prompt += EOS_TOKEN
if params['genre']:
prompt += params['genre'] + EOG_TOKEN
if params['title']:
prompt += params['title'] + EOT_TOKEN
if params['prefix']:
prompt += params['prefix']
text = run_generation.main([
'--model_type=gpt2', '--model_name_or_path=app/output',
f"--prompt={prompt}" if prompt else '--prompt=""',
f'--temperature={float(params["temp"]) if params["temp"] else uniform(0.7, 1)}',
f'--top_p={float(params["top_p"]) if params["top_p"] else 0}',
'--num_samples=1', '--length=256', f'--stop_token={EOS_TOKEN}'
])
return prompt + text
def generate_text(params):
"""Generate text using transformers."""
if params['genre'] and params['prefix']:
prompt = BOS_TOKEN + params['genre'] + EOG_TOKEN + params['prefix']
elif params['genre']:
prompt = BOS_TOKEN + params['genre'] + EOG_TOKEN
elif params['prefix']:
# If user specifies prefix, the model cannot generate the genre anymore
prompt = params['prefix']
else:
prompt = BOS_TOKEN
text = run_generation.main([
'--model_type=gpt2', '--model_name_or_path=app/output',
f"--prompt={prompt}" if prompt else '--prompt=""',
f'--temperature={float(params["temp"]) if params["temp"] else uniform(0.7, 1)}',
f'--top_p={float(params["top_p"]) if params["top_p"] else 0}',
'--num_samples=1', '--length=256', f'--stop_token={EOS_TOKEN}'
])
return prompt + text
def inference():
SEED = random.randint(0, 9999)
parser = argparse.ArgumentParser()
parser.add_argument('--num_return_sequences',
help='num_return_sequences=1',
type=int,
default=3)
parser.add_argument('--batch_size', type=int, default=1)
parser.add_argument('--no_fp16', help='', type=bool, default=False)
args = parser.parse_args()
num_return_sequences = args.num_return_sequences
batch_size = args.batch_size
if (args.no_fp16):
fp16 = False
else:
fp16 = True
prompt = input("Model prompt >>> ")
args = [
"--num_return_sequences=" + str(num_return_sequences),
"--prompt=" + prompt, "--model_type=" + MODEL_TYPE,
"--model_name_or_path=" + MODEL_PATH, "--seed=" + str(SEED),
"--length=350"
]
if (fp16):
args.append("--fp16")
start_time = time.time()
sequences = run_generation.main(args, batch_size)
print("text generation took --- %s seconds ---" %
(time.time() - start_time))
ts = time.time()
filename = DATA_PATH + "/inference_results_" + datetime.datetime.fromtimestamp(
ts).strftime('%Y-%m-%d_%H:%M') + ".csv"
print("saving results to: ", filename)
with open(filename, 'w', newline='') as myfile:
wr = csv.writer(myfile, quoting=csv.QUOTE_ALL)
wr.writerow(sequences)
import argparse
import logging
import sys
import unittest
from unittest.mock import patch
import run_generation
logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger()
testargs = ["run_generation.py", "--prompt=Hello", "--length=240", "--seed=42"]
model_type, model_name = ("--model_type=gpt2", "--model_name_or_path=gpt2")
with patch.object(sys, "argv", testargs + [model_type, model_name]):
result = run_generation.main()
print(result)