metadata
license: apache-2.0
datasets:
- McGill-NLP/FaithDial
language:
- en
metrics:
- bleu
- bertscore
- accuracy
pipeline_tag: conversational
T3 stands for Terribly Tiny Transformers that are an efficient way of creating tiny distilled (student) models for hallucination-free LLM models in parameter-constrained environment (edge devices). The base model is a T3 adaptation of T5 model. The paradigm of T3 models can be extended to all types of models ( encoder only, decoder only & seq2seq)