T5¶
DISCLAIMER: This model is still a work in progress, if you see something strange, file a Github Issue
Overview¶
The T5 model was presented in Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer by Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu in Here the abstract:
Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new “Colossal Clean Crawled Corpus”, we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code.
The Authors’ code can be found here .
Training¶
T5 is an encoder-decoder model and converts all NLP problems into a text-to-text format. It is trained using teacher forcing.
This means that for training we always need an input sequence and a target sequence.
The input sequence is fed to the model using input_ids. The target sequence is shifted to the right, i.e. perprended by a start-sequence token and fed to the decoder using the decoder_input_ids. In teacher-forcing style, the target sequence is then appended by the EOS token and corresponds to the lm_labels. The PAD token is hereby used as the start-sequence token.
T5 can be trained / fine-tuned both in a supervised and unsupervised fashion.
Unsupervised denoising training In this setup spans of the input sequence are masked by so-called sentinel tokens (a.k.a unique mask tokens) and the output sequence is formed as a concatenation of the same sentinel tokens and the real masked tokens. Each sentinel tokens represents a unique mask token for this sentence and should start with
<extra_id_1>,<extrac_id_2>, … up to<extra_id_100>. As a default 100 sentinel tokens are available inT5Tokenizer. E.g. the sentence “The cute dog walks in the park” with the masks put on “cute dog” and “the” should be processed as follows:
input_ids = tokenizer.encode('The <extra_id_1> walks in <extra_id_2> park', return_tensors='pt')
lm_labels = tokenizer.encode('<extra_id_1> cute dog <extra_id_2> the <extra_id_3> </s>', return_tensors='pt')
# the forward function automatically creates the correct decoder_input_ids
model(input_ids=input_ids, lm_labels=lm_labels)
Supervised training In this setup the input sequence and output sequence are standard sequence to sequence input output mapping. In translation, e.g. the input sequence “The house is wonderful.” and output sequence “Das Haus ist wunderbar.” should be processed as follows:
input_ids = tokenizer.encode('translate English to German: The house is wonderful. </s>', return_tensors='pt')
lm_labels = tokenizer.encode('Das Haus ist wunderbar. </s>', return_tensors='pt')
# the forward function automatically creates the correct decoder_input_ids
model(input_ids=input_ids, lm_labels=lm_labels)
Tips¶
T5 is an encoder-decoder model pre-trained on a multi-task mixture of unsupervised and supervised tasks and for which each task is converted into a text-to-text format. T5 works well on a variety of tasks out-of-the-box by prepending a different prefix to the input corresponding to each task, e.g.: for translation: translate English to German: …, summarize: …. For more information about which prefix to use, it is easiest to look into Appendix D of the paper .
For sequence to sequence generation, it is recommended to use
T5ForConditionalGeneration.generate(). The method takes care of feeding the encoded input via cross-attention layers to the decoder and auto-regressively generates the decoder output.T5 uses relative scalar embeddings. Encoder input padding can be done on the left and on the right.