allenai/unifiedqa

Issues

• How much GPU memory is needed to run the 11B model? and how much about 3B model?

  #48 opened a year ago by bigbrother001
  0

• Finetuning the v2 version of the base model

  #47 opened 2 years ago by saggu
  0

• low loss in fine tuning but generated answers are not correct

  #46 opened 3 years ago by cnut1648
  7

• Is there a way to get SQuAD V2 style "No Answer"?

  #45 opened 3 years ago by tingofurro
  1

• Motivation for two checkpoints in UnifiedQA v2

  #44 opened 3 years ago by marksverdhei
  1

• Could I have your processed data?

  #12 opened 4 years ago by Sanqiang
  3

• Reproducing Evaluation Numbers

  #43 opened 3 years ago by carlos-gemmell
  8

• What is the purpose for the regex substitution in preprocessing?

  #42 opened 3 years ago by marksverdhei
  1

• AutoTokenizer broken for v2 models

  #41 opened 3 years ago by shwang
  2

• What's new in V2?

  #39 opened 3 years ago by chen-yifu
  2

• ValueError: exactly one of keep_prob and rate should be set

  #34 opened 3 years ago by ayush714
  1

• How to do Span Extraction with T5/BART-style model ?

  #37 opened 3 years ago by ZenMoore
  1

• What learning rate should be used to fine-tune T5-large and T5-3B?

  #40 opened 3 years ago by chen-yifu
  1

• Will the V2 models be added to the HuggingFace Hub?

  #38 opened 3 years ago by tingofurro
  1

• Getting the same error even after using the slightly older versions.

  #33 opened 3 years ago by ayush714
  2

• spelling mistake in unifiedqa/bart/run.py:158

  #36 opened 3 years ago by christineaa
  1

• What are "flatten", "in_metadata" and "out_metadata" in dataloader?

  #35 opened 3 years ago by WindChimeRan
  1

• Using Unified QA T5-large Saying ( Getting a Error ) Make Layer Stack Param Not found,

  #32 opened 3 years ago by ayush714
  3

• How to fit T5-11B(and T5-3b) into a v3-8 TPU

  #31 opened 3 years ago by Sanqiang
  2

• How to fit T5-11B(and T5-3b) into a v3-8 TPU

  #30 opened 3 years ago by Sanqiang
  0

• how to convert text format for multiple answers

  #26 opened 3 years ago by zirui
  2

• Unable to download and use any of the models

  #25 opened 3 years ago by alisonreboud
  3

• how many candidates can test model(unifiedQA-uncased/best-model.pt) involve?

  #22 opened 4 years ago by wgkwgk
  1

• Evaluation scores

  #21 opened 4 years ago by NacirB
  1

• Cuda Out Of Memory

  #20 opened 4 years ago by NacirB
  6

• error from generate_from_string method for BART

  #17 opened 4 years ago by timhartill
  7

• Confidence Values and N-Best for Prediction

  #18 opened 4 years ago by alexander-fichtl
  1

• Training params for smaller models

  #15 opened 4 years ago by ag1988
  5

• Finetuning Parameters

  #16 opened 4 years ago by corypaik
  1

• score associated with a result?

  #14 opened 4 years ago by SourAlderson
  1

• Further Fine-tuning on T5

  #13 opened 4 years ago by chen-yifu
  1

• Multi-Paragraph

  #10 opened 4 years ago by jiajudu
  3

• Huggingface 3B and 11B models not configured properly

  #8 opened 4 years ago by Shamdan17
  10

• Not exactly reproducing the example with huggingface

  #11 opened 4 years ago by wenlongzhao094
  7

• BART weights are unavailable

  #9 opened 4 years ago by tshrjn
  1

• Unable to load models in huggingface, tf throws DataLossError

  #6 opened 4 years ago by tshrjn
  1

• What is the performance of BART checkpoint ?

  #4 opened 4 years ago by ratthachat
  2

• Do the checkpoints rely on deep learning framework?

  #5 opened 4 years ago by hi-weiyuan
  1

• Can I use your pretrained model in transformers?

  #1 opened 4 years ago by ky941122
  3

• Checkpoint 1000700 of t5-large is much worse than others

  #2 opened 4 years ago by joe32140
  1