reddit_t5_summarization.py

>>> from transformers import T5Tokenizer, T5ForConditionalGeneration
2020-11-10 18:10:11.206223: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
>>> model = T5ForConditionalGeneration.from_pretrained('t5-base')
>>> tokenizer = T5Tokenizer.from_pretrained('t5-base')
>>> text = ".".join(comment_section)
>>> Preprocessed_text = "summarize: " + text
>>> tokens_input = tokenizer.encode(Preprocessed_text,return_tensors="pt", max_length=512, truncation=True)
>>> summary_ids = model.generate(tokens_input,
...                              min_length=60,
...                              max_length=180,
...                              length_penalty=4.0)
>>>
>>> summary = tokenizer.decode(summary_ids[0])
>>> print(summary)
british queen elizabeth sends birthday cards to people who turn 100 on their birthday. 
i think it's a weird phenomenon because it's not automatic. 
i've never heard of a queen of england so i can't help but think it's a joke. 
i'd love to see a british queen congratulate her on her 100th birthday.