Spaces:
Runtime error
Runtime error
File size: 1,614 Bytes
fa6856c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 |
# Generates positive movie reviews by tuning a pretrained model on IMDB dataset
# with a sentiment reward function
import json
import os
import sys
from typing import List
import torch
from datasets import load_dataset
from transformers import pipeline
import trlx
from trlx.data.default_configs import TRLConfig, default_ppo_config
def get_positive_score(scores):
"Extract value associated with a positive sentiment from pipeline's output"
return dict(map(lambda x: tuple(x.values()), scores))["POSITIVE"]
def main(hparams={}):
# Merge sweep config with default config if given
config = TRLConfig.update(default_ppo_config().to_dict(), hparams)
if torch.cuda.is_available():
device = int(os.environ.get("LOCAL_RANK", 0))
else:
device = -1
sentiment_fn = pipeline(
"sentiment-analysis",
"lvwerra/distilbert-imdb",
top_k=2,
truncation=True,
batch_size=256,
device=device,
)
def reward_fn(samples: List[str], **kwargs) -> List[float]:
sentiments = list(map(get_positive_score, sentiment_fn(samples)))
return sentiments
# Take few words off of movies reviews as prompts
imdb = load_dataset("imdb", split="train+test")
prompts = [" ".join(review.split()[:4]) for review in imdb["text"]]
trlx.train(
reward_fn=reward_fn,
prompts=prompts,
eval_prompts=["I don't know much about Hungarian underground"] * 256,
config=config,
)
if __name__ == "__main__":
hparams = {} if len(sys.argv) == 1 else json.loads(sys.argv[1])
main(hparams)
|