thatdramebaazguy/roberta-base-squad is a forked repo from huggingface. License: cc-by-4-0
Go to file
Aditya Singhal a7a26fd500 Update README.md 2022-07-01 19:12:50 +00:00
.gitattributes allow flax 2021-05-20 22:28:09 +00:00
README.md Update README.md 2022-07-01 19:12:50 +00:00
config.json Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00
flax_model.msgpack upload flax model 2021-05-20 22:28:27 +00:00
merges.txt Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00
pytorch_model.bin Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00
special_tokens_map.json Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00
tf_model.h5 Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00
tokenizer_config.json Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00
vocab.json Adding files - Aditya Singhal 2021-04-26 00:48:27 -04:00

README.md

datasets language thumbnail tags license
SQuAD
English
roberta
roberta-base
question-answering
qa
cc-by-4.0

roberta-base + SQuAD QA

Objective: This is Roberta Base trained to do the SQuAD Task. This makes a QA model capable of answering questions.

model_name = "thatdramebaazguy/roberta-base-squad"
pipeline(model=model_name, tokenizer=model_name, revision="v1.0", task="question-answering")

Overview

Language model: roberta-base
Language: English
Downstream-task: QA
Training data: SQuADv1
Eval data: SQuAD
Infrastructure: 2x Tesla v100
Code: See example

Hyperparameters

Num examples = 88567  
Num Epochs = 10
Instantaneous batch size per device = 32  
Total train batch size (w. parallel, distributed & accumulation) = 64 

Performance

Eval on SQuADv1

  • epoch = 10.0
  • eval_samples = 10790
  • exact_match = 83.6045
  • f1 = 91.1709

Eval on MoviesQA

  • eval_samples = 5032
  • exact_match = 51.6494
  • f1 = 68.2615

Github Repo: