{
  "total": 237,
  "topics": [
    {
      "id": 127409,
      "title": "Collections of gold solutions",
      "comment_count": 9,
      "views": 0,
      "votes": 58
    },
    {
      "id": 114727,
      "title": "Official External Data Thread",
      "comment_count": 25,
      "views": 0,
      "votes": 3
    },
    {
      "id": 131675,
      "title": "TensorFlow 2.0 Prize Winners!",
      "comment_count": 4,
      "views": 0,
      "votes": 6
    },
    {
      "id": 127262,
      "title": "TensorFlow 2.0 Prizes: Call for eligible submissions!",
      "comment_count": 17,
      "views": 0,
      "votes": 4
    },
    {
      "id": 118609,
      "title": "Request Free Cloud TPU Quota!",
      "comment_count": 38,
      "views": 0,
      "votes": 31
    },
    {
      "id": 114718,
      "title": "Welcome!",
      "comment_count": 20,
      "views": 0,
      "votes": 11
    },
    {
      "id": 114729,
      "title": "Looking for a Team Thread",
      "comment_count": 68,
      "views": 0,
      "votes": 7
    },
    {
      "id": 121506,
      "title": "Leaderboard update complete",
      "comment_count": 1,
      "views": 0,
      "votes": 8
    },
    {
      "id": 120931,
      "title": "We're going to be updating the leaderboard...",
      "comment_count": 16,
      "views": 0,
      "votes": 7
    },
    {
      "id": 114730,
      "title": "New to Machine Learning or Kaggle?",
      "comment_count": 8,
      "views": 0,
      "votes": 5
    },
    {
      "id": 114720,
      "title": "New to NLP? Check this out",
      "comment_count": 3,
      "views": 0,
      "votes": 51
    },
    {
      "id": 127339,
      "title": "3rd place solution",
      "comment_count": 30,
      "views": 0,
      "votes": 77
    },
    {
      "id": 127309,
      "title": "Training BERT for different languages",
      "comment_count": 7,
      "views": 0,
      "votes": 9
    },
    {
      "id": 114724,
      "title": "19 Great Articles About Natural Language Processing (NLP)",
      "comment_count": 74,
      "views": 0,
      "votes": 106
    },
    {
      "id": 120327,
      "title": "How do you guys \"Clean\" the data",
      "comment_count": 6,
      "views": 0,
      "votes": 1
    },
    {
      "id": 115553,
      "title": "cannot commit: \"Your Notebook cannot use internet access in this competition\" ",
      "comment_count": 16,
      "views": 0,
      "votes": 15
    },
    {
      "id": 123434,
      "title": "HuggingFace question / BertForQuestionAnswering",
      "comment_count": 8,
      "views": 0,
      "votes": 5
    },
    {
      "id": 202655,
      "title": "Huggingface - models",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 127371,
      "title": "4th place Solution",
      "comment_count": 11,
      "views": 0,
      "votes": 29
    },
    {
      "id": 117716,
      "title": "Hugging Face Transformers: PyTorch Setup",
      "comment_count": 4,
      "views": 0,
      "votes": 14
    },
    {
      "id": 127212,
      "title": "Congrats to the winners!!!",
      "comment_count": 11,
      "views": 0,
      "votes": 11
    },
    {
      "id": 127551,
      "title": "1st place solution with code",
      "comment_count": 55,
      "views": 0,
      "votes": 151
    },
    {
      "id": 127523,
      "title": "AMA Interview with Dieter | 3rd Pos Solution | Chai Time Data Science Show",
      "comment_count": 9,
      "views": 0,
      "votes": 12
    },
    {
      "id": 127223,
      "title": "BERT,RoBERTa,ALBERT and...What pretrained model have you used?",
      "comment_count": 5,
      "views": 0,
      "votes": 2
    },
    {
      "id": 114861,
      "title": "Question Answering Datasets/Leaderboards",
      "comment_count": 1,
      "views": 0,
      "votes": 25
    },
    {
      "id": 127259,
      "title": "7th place solution",
      "comment_count": 16,
      "views": 0,
      "votes": 52
    },
    {
      "id": 127521,
      "title": "6th place solution",
      "comment_count": 8,
      "views": 0,
      "votes": 26
    },
    {
      "id": 118868,
      "title": "Is anyone able to do fine tuning Bert model using TF2.0 or Keras?",
      "comment_count": 21,
      "views": 0,
      "votes": 8
    },
    {
      "id": 128832,
      "title": "🥇🥈🥉 TF2.0 QA Medal Solutions 🥉🥈🥇",
      "comment_count": 2,
      "views": 0,
      "votes": 13
    },
    {
      "id": 131064,
      "title": "Is Negative Sampling the huge trick in this competation?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 128278,
      "title": "9th place solution",
      "comment_count": 9,
      "views": 0,
      "votes": 38
    },
    {
      "id": 130059,
      "title": "How to evaluate the predictions (metrics)",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 127333,
      "title": "2nd place solution",
      "comment_count": 42,
      "views": 0,
      "votes": 92
    },
    {
      "id": 127545,
      "title": "8th place solution",
      "comment_count": 4,
      "views": 0,
      "votes": 30
    },
    {
      "id": 127350,
      "title": "Brief summary of 13th place solution (hide the pain Harold)",
      "comment_count": 6,
      "views": 0,
      "votes": 28
    },
    {
      "id": 126702,
      "title": "BERT & Friends Reference",
      "comment_count": 21,
      "views": 0,
      "votes": 37
    },
    {
      "id": 128747,
      "title": "Kernels and posts of winners of this competition in one kernel",
      "comment_count": 3,
      "views": 0,
      "votes": 6
    },
    {
      "id": 127215,
      "title": "Wow",
      "comment_count": 17,
      "views": 0,
      "votes": 28
    },
    {
      "id": 128188,
      "title": "30th Place Solution",
      "comment_count": 8,
      "views": 0,
      "votes": 8
    },
    {
      "id": 127241,
      "title": "17th Place solution [bert-disjoint] kernel + all utility scripts",
      "comment_count": 4,
      "views": 0,
      "votes": 15
    },
    {
      "id": 128140,
      "title": "21th place solution, puzzlingly shaking from public LB 3th",
      "comment_count": 6,
      "views": 0,
      "votes": 5
    },
    {
      "id": 127374,
      "title": "23rd place solution: ensemble, rank passage and predict span",
      "comment_count": 3,
      "views": 0,
      "votes": 11
    },
    {
      "id": 127266,
      "title": "47th Place Solution Write-Up (Ensembling)",
      "comment_count": 16,
      "views": 0,
      "votes": 17
    },
    {
      "id": 127219,
      "title": "Bo is awesome. ",
      "comment_count": 5,
      "views": 0,
      "votes": 16
    },
    {
      "id": 128186,
      "title": "How to pretrain model by SQuAD2",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 127512,
      "title": "The actual competition metric (unofficial)",
      "comment_count": 19,
      "views": 0,
      "votes": 12
    },
    {
      "id": 127232,
      "title": "31st solution with custom loss ",
      "comment_count": 10,
      "views": 0,
      "votes": 22
    },
    {
      "id": 127258,
      "title": "45th solution, my journey and learnings, feeling grateful :) ",
      "comment_count": 2,
      "views": 0,
      "votes": 14
    },
    {
      "id": 127385,
      "title": "AMA Interview with Comp Winner: Guanshuo Xu | Chai Time Data Science ",
      "comment_count": 2,
      "views": 0,
      "votes": 13
    },
    {
      "id": 127257,
      "title": "rank68_the_ simplest idea to get the first medal",
      "comment_count": 3,
      "views": 0,
      "votes": 6
    },
    {
      "id": 127214,
      "title": "what is the best TF2.0 single model?",
      "comment_count": 14,
      "views": 0,
      "votes": 6
    },
    {
      "id": 127473,
      "title": "Third decimal point visibility after submission deadline",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 127220,
      "title": "Rank 93 Solution: Or how you can still win a medal without any model fine-tuning",
      "comment_count": 4,
      "views": 0,
      "votes": 7
    },
    {
      "id": 127299,
      "title": "27th solution with luck and some questions from me",
      "comment_count": 3,
      "views": 0,
      "votes": 9
    },
    {
      "id": 127250,
      "title": "MVP award for prvi ",
      "comment_count": 0,
      "views": 0,
      "votes": 10
    },
    {
      "id": 127346,
      "title": "[Reference]Frustratingly Easy Natural Question Answering",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 127267,
      "title": "\"This competition has completed. This leaderboard reflects preliminary final standings. The result will become final after the competition organizers verify the results.\"",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 127351,
      "title": "NQ&GQ",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 127209,
      "title": "So it seems private data really match the public data... ",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 127463,
      "title": "AutoML on TensorFlow 2.0 Question Answering",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 126609,
      "title": "bert-joint model answers thresholds",
      "comment_count": 21,
      "views": 0,
      "votes": 4
    },
    {
      "id": 127185,
      "title": "Zero score",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 127029,
      "title": "TF2.0 QA Nostalgia ",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 127188,
      "title": "Good luck and last 3 hours",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 127173,
      "title": "Simultaneous Submissions?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 127171,
      "title": "Why this will happen...",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 126974,
      "title": "[TF2.0 Question] Another way to feed data instead of TFRecords",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 126371,
      "title": "Possible to work with Albert ?",
      "comment_count": 23,
      "views": 0,
      "votes": 8
    },
    {
      "id": 126887,
      "title": "How do you submit ?",
      "comment_count": 3,
      "views": 0,
      "votes": 0
    },
    {
      "id": 126801,
      "title": "Any models other than Bert ?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 126543,
      "title": "[TF2.0] Way to accumulate gradients in custom loop TF2.0",
      "comment_count": 0,
      "views": 0,
      "votes": 10
    },
    {
      "id": 126742,
      "title": "can't submit submission.csv file",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 126488,
      "title": "Competition Closing on 1/22",
      "comment_count": 1,
      "views": 0,
      "votes": 6
    },
    {
      "id": 126724,
      "title": "Validation up but lb down?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 125594,
      "title": "question (or my problem) about CV and LB",
      "comment_count": 1,
      "views": 0,
      "votes": 4
    },
    {
      "id": 126372,
      "title": "MS Marco website down ???",
      "comment_count": 3,
      "views": 0,
      "votes": -1
    },
    {
      "id": 126455,
      "title": "Question about long answer metrics",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 125938,
      "title": "Ensembling",
      "comment_count": 5,
      "views": 0,
      "votes": 3
    },
    {
      "id": 126113,
      "title": "Upgrading from Colab TPU to paid TPU",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 125817,
      "title": "Notebook errors",
      "comment_count": 11,
      "views": 0,
      "votes": null
    },
    {
      "id": 124158,
      "title": "Best Single Model Score",
      "comment_count": 25,
      "views": 0,
      "votes": 11
    },
    {
      "id": 125641,
      "title": "Paper 'A Survey on Machine Reading Comprehension Systems'",
      "comment_count": 2,
      "views": 0,
      "votes": 8
    },
    {
      "id": 125675,
      "title": "Clarification on multispan short answers",
      "comment_count": 8,
      "views": 0,
      "votes": 2
    },
    {
      "id": 125923,
      "title": "Notebooks on GCP - change path",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 125746,
      "title": "Kernels and posts of Winners of Prize Competitions for NLP&TF",
      "comment_count": 0,
      "views": 0,
      "votes": 8
    },
    {
      "id": 125775,
      "title": "Are we supposed to create nq_test_tfrecord offline ?",
      "comment_count": 9,
      "views": 0,
      "votes": null
    },
    {
      "id": 125750,
      "title": "Eval Metric - my guess in a kernel",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 123266,
      "title": "Question about getting short_answer end_token",
      "comment_count": 23,
      "views": 0,
      "votes": 3
    },
    {
      "id": 125627,
      "title": "Interim quick submission to public LB.",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 124914,
      "title": "My python script for training on GCP + TPU",
      "comment_count": 7,
      "views": 0,
      "votes": 5
    },
    {
      "id": 124662,
      "title": "Fine-tuning pipeline for large model",
      "comment_count": 24,
      "views": 0,
      "votes": 2
    },
    {
      "id": 123977,
      "title": "Clarification on Kernel time limit",
      "comment_count": 35,
      "views": 0,
      "votes": 2
    },
    {
      "id": 125079,
      "title": "What is token_map used for?",
      "comment_count": 4,
      "views": 0,
      "votes": 2
    },
    {
      "id": 125400,
      "title": "DrQA Reader",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 125233,
      "title": "TensorFlow 2.1 eligibility for TensorFlow 2.0 Prizes",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 124710,
      "title": "Natural Questions Data Preparation",
      "comment_count": 4,
      "views": 0,
      "votes": 4
    },
    {
      "id": 116625,
      "title": "Data Augmentation for NLP",
      "comment_count": 8,
      "views": 0,
      "votes": 31
    },
    {
      "id": 120240,
      "title": "Congrats to Dieter for becoming Kaggle Discussion Grandmaster",
      "comment_count": 19,
      "views": 0,
      "votes": 28
    },
    {
      "id": 121572,
      "title": "Using BERT on Cloud TPUs with TensorFlow 2.1 RC",
      "comment_count": 23,
      "views": 0,
      "votes": 21
    },
    {
      "id": 118129,
      "title": "Save time committing & submitting",
      "comment_count": 13,
      "views": 0,
      "votes": 25
    },
    {
      "id": 120461,
      "title": "Visualizing data with Google's nq_browser",
      "comment_count": 25,
      "views": 0,
      "votes": 49
    },
    {
      "id": 123345,
      "title": "My TF2 Inference + Validation kernel is done.",
      "comment_count": 18,
      "views": 0,
      "votes": 23
    },
    {
      "id": 119784,
      "title": "Cannot submit to competion: redirected and 404",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 122853,
      "title": "Anyone tried models other than BERT",
      "comment_count": 4,
      "views": 0,
      "votes": 2
    },
    {
      "id": 122293,
      "title": "[PyTorch] Question about pre-trained Bert specific to this competition",
      "comment_count": 21,
      "views": 0,
      "votes": 4
    },
    {
      "id": 119959,
      "title": "TF2 Training kernel - Use Hugging Face's Tensorflow 2 transformer models",
      "comment_count": 9,
      "views": 0,
      "votes": 7
    },
    {
      "id": 121972,
      "title": "Fine tuning",
      "comment_count": 9,
      "views": 0,
      "votes": 3
    },
    {
      "id": 122799,
      "title": "BERT Base vs/ BERT Large",
      "comment_count": 16,
      "views": 0,
      "votes": 3
    },
    {
      "id": 123466,
      "title": "Scoring error again",
      "comment_count": 4,
      "views": 0,
      "votes": 3
    },
    {
      "id": 123781,
      "title": "Dataset conflict",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 123533,
      "title": "Submission scoring error",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 123458,
      "title": "Run out of GPU quota for a week",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 116409,
      "title": "Scoring error on submission - even with sample_submission.csv",
      "comment_count": 19,
      "views": 0,
      "votes": 13
    },
    {
      "id": 123634,
      "title": "Is the GPU notebook time limit three hours from the beginning?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 121286,
      "title": "Please provide implementation of the competition metric",
      "comment_count": 10,
      "views": 0,
      "votes": 45
    },
    {
      "id": 115874,
      "title": "More Natural Language Processing Articles",
      "comment_count": 11,
      "views": 0,
      "votes": 22
    },
    {
      "id": 120030,
      "title": "A bit confused, using the new metric, offline ~0.5 but online 0.7+",
      "comment_count": 27,
      "views": 0,
      "votes": 8
    },
    {
      "id": 123408,
      "title": "What happen if the Wiki page Layout changes/Updated?",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 123392,
      "title": "How does the joint model fit this competition?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 123298,
      "title": "anyone using tf1？",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 122680,
      "title": "Customizing BERT",
      "comment_count": 2,
      "views": 0,
      "votes": 10
    },
    {
      "id": 121998,
      "title": "nq-test.tfrecords",
      "comment_count": 4,
      "views": 0,
      "votes": 3
    },
    {
      "id": 116896,
      "title": "GPT-2 and BERT Pretrained Weights (pytorch)",
      "comment_count": 10,
      "views": 0,
      "votes": 18
    },
    {
      "id": 122914,
      "title": "How to get public score quickly?",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 122976,
      "title": "What does score < 1.5 mean?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 122859,
      "title": "[Remark] - My TF2 inference kernel",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 122855,
      "title": "[Warning] -- Wrong model in my TF2 traning kernel",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 122776,
      "title": "Advice to beginners starting out?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 118454,
      "title": "is Colab good for this competition ?",
      "comment_count": 18,
      "views": 0,
      "votes": 4
    },
    {
      "id": 122459,
      "title": "TPU access received. Now what?",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 119957,
      "title": "Reproducing joint BERT baseline",
      "comment_count": 33,
      "views": 0,
      "votes": 21
    },
    {
      "id": 122041,
      "title": "[SOLVED] How many test sets do we have?",
      "comment_count": 6,
      "views": 0,
      "votes": 10
    },
    {
      "id": 119530,
      "title": "Problem with a basic submission",
      "comment_count": 8,
      "views": 0,
      "votes": null
    },
    {
      "id": 121860,
      "title": "Why running private dataset every time?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 121852,
      "title": "Is it worth participating in this competition?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 121394,
      "title": "Key to success in this competition",
      "comment_count": 8,
      "views": 0,
      "votes": 4
    },
    {
      "id": 121610,
      "title": "GCP: can't use credits on GPU",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 115676,
      "title": "Go from the basics - Attention mechanism, transformers, BERT",
      "comment_count": 2,
      "views": 0,
      "votes": 6
    },
    {
      "id": 120658,
      "title": "substantial code competition",
      "comment_count": 4,
      "views": 0,
      "votes": 10
    },
    {
      "id": 121666,
      "title": "still confused by the metric",
      "comment_count": 3,
      "views": 0,
      "votes": 4
    },
    {
      "id": 121581,
      "title": "how can I get the files of \"input/bert-joint-baseline/model_cpkt-1*\"",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 120535,
      "title": "How to use high-performance colab",
      "comment_count": 3,
      "views": 0,
      "votes": 9
    },
    {
      "id": 121080,
      "title": "Wrong metric waste too much time.Is current metric correct now? ",
      "comment_count": 13,
      "views": 0,
      "votes": 10
    },
    {
      "id": 117007,
      "title": "Ready-to-use:Transformed jsonl to dataframe",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 120074,
      "title": "Beginner-level study material for learning BERT",
      "comment_count": 2,
      "views": 0,
      "votes": 4
    },
    {
      "id": 119000,
      "title": "PyTorch BERT training kernel",
      "comment_count": 11,
      "views": 0,
      "votes": 23
    },
    {
      "id": 115202,
      "title": "Google colab starter kit!",
      "comment_count": 4,
      "views": 0,
      "votes": 12
    },
    {
      "id": 121032,
      "title": "From 0.7 to 0.48 !!!",
      "comment_count": 8,
      "views": 0,
      "votes": 5
    },
    {
      "id": 121255,
      "title": "Anyone using PyTorch?",
      "comment_count": 11,
      "views": 0,
      "votes": 2
    },
    {
      "id": 120061,
      "title": "My implementation of competition metric",
      "comment_count": 3,
      "views": 0,
      "votes": 36
    },
    {
      "id": 121275,
      "title": "about submissions",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 120973,
      "title": "What are the best resources to learn about BERT?",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 121304,
      "title": "Any good way to restore tf1 checkpoint?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 119604,
      "title": "Submitting the public kernel produces an error",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 120803,
      "title": "Question: How to make a prediction from N strides",
      "comment_count": 4,
      "views": 0,
      "votes": 4
    },
    {
      "id": 121015,
      "title": "Leaderboard is changing",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 119239,
      "title": "Tensorflow 1 and bert-tensorflow installation",
      "comment_count": 8,
      "views": 0,
      "votes": 4
    },
    {
      "id": 120419,
      "title": "GPU quota bug, and be careful everyone.",
      "comment_count": 4,
      "views": 0,
      "votes": 6
    },
    {
      "id": 118758,
      "title": "Please provide an official metric implementation or clarification",
      "comment_count": 20,
      "views": 0,
      "votes": 21
    },
    {
      "id": 119992,
      "title": "Loss used in BertForQuestionAnswering",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 117370,
      "title": "Understand how does training input look like",
      "comment_count": 4,
      "views": 0,
      "votes": 47
    },
    {
      "id": 116306,
      "title": "about the way BERT-joint (baseline) processes long answers",
      "comment_count": 5,
      "views": 0,
      "votes": 5
    },
    {
      "id": 120573,
      "title": "Expected output for no-answer cases?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 114787,
      "title": "State-of-the-art results on the SQuAD",
      "comment_count": 4,
      "views": 0,
      "votes": 9
    },
    {
      "id": 115018,
      "title": "Sharing my utility script: Convert line-delimited JSON to pandas",
      "comment_count": 4,
      "views": 0,
      "votes": 34
    },
    {
      "id": 120209,
      "title": "BERT code analysis",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 117473,
      "title": "LB probing ",
      "comment_count": 9,
      "views": 0,
      "votes": 14
    },
    {
      "id": 119606,
      "title": "What is the metric",
      "comment_count": 15,
      "views": 0,
      "votes": 5
    },
    {
      "id": 119462,
      "title": "CV vs LB",
      "comment_count": 10,
      "views": 0,
      "votes": 6
    },
    {
      "id": 120131,
      "title": "long answer candidates in test data",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 116095,
      "title": "Types of questions in the dataset",
      "comment_count": 1,
      "views": 0,
      "votes": 2
    },
    {
      "id": 119143,
      "title": "Time spent so far",
      "comment_count": 5,
      "views": 0,
      "votes": 22
    },
    {
      "id": 119280,
      "title": "A story about 2 submission files",
      "comment_count": 14,
      "views": 0,
      "votes": 7
    },
    {
      "id": 120025,
      "title": "What is the significance of multiple long_answer_candidates ?",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 116096,
      "title": "Annotation process",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 120102,
      "title": "Does Layer Norm still need batch size (>=4)?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 114719,
      "title": "Past NLP Competitions",
      "comment_count": 4,
      "views": 0,
      "votes": 34
    },
    {
      "id": 114717,
      "title": "When next NLP match begins, one question",
      "comment_count": 9,
      "views": 0,
      "votes": 6
    },
    {
      "id": 115695,
      "title": "GPU limit reached",
      "comment_count": 1,
      "views": 0,
      "votes": -1
    },
    {
      "id": 119768,
      "title": "I can't understand meaning of {'start_token': -1,' candidate_index': -1,' end_token': -1}",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 119560,
      "title": "BERT Tutorial Kernel",
      "comment_count": 2,
      "views": 0,
      "votes": 4
    },
    {
      "id": 114835,
      "title": "GPU vs TPU for BERT",
      "comment_count": 12,
      "views": 0,
      "votes": 4
    },
    {
      "id": 119191,
      "title": "PLEASE PROVIDE AN OFFICAL METRIC IMPLEMENTATION!",
      "comment_count": 0,
      "views": 0,
      "votes": 13
    },
    {
      "id": 119134,
      "title": "Evaluation Score Do not Change",
      "comment_count": 4,
      "views": 0,
      "votes": 5
    },
    {
      "id": 118825,
      "title": "Some Long answer indicies larger than total number of tokens",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 116566,
      "title": "Anyone getting deterministic results with TensorFlow 2.0 or new Keras",
      "comment_count": 5,
      "views": 0,
      "votes": 4
    },
    {
      "id": 119363,
      "title": "Requesting clarifications on submissions",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 118759,
      "title": "Are you predicting multiple short answers?",
      "comment_count": 2,
      "views": 0,
      "votes": 7
    },
    {
      "id": 119040,
      "title": "Interesting article on NLP Transfer Learning",
      "comment_count": 1,
      "views": 0,
      "votes": 2
    },
    {
      "id": 119035,
      "title": "Looking to join a team ",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 118265,
      "title": "Use the  tensorboard-2.0 in kaggle .it show 404:not Found ,why？",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 118073,
      "title": "TPU and GCP usage in kernels",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 115715,
      "title": "top_level tag in long_answer_candidates json",
      "comment_count": 2,
      "views": 0,
      "votes": 5
    },
    {
      "id": 116801,
      "title": "what is the significance of candidate_index in annotations field  and top_level in long_answer_candidates field ",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 117113,
      "title": "Submission Scoring Error - No additional details provided for this error",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 118453,
      "title": "Check this out",
      "comment_count": 1,
      "views": 0,
      "votes": 14
    },
    {
      "id": 116268,
      "title": "Memory efficient JSON file reader",
      "comment_count": 2,
      "views": 0,
      "votes": 13
    },
    {
      "id": 118715,
      "title": "Kernel doesn't connect",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 116890,
      "title": "Question about short answer evaluation",
      "comment_count": 12,
      "views": 0,
      "votes": 10
    },
    {
      "id": 118294,
      "title": "After reading this nice paper, I have 2 questions",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 117182,
      "title": "Confused about the evaluation score",
      "comment_count": 11,
      "views": 0,
      "votes": 17
    },
    {
      "id": 116325,
      "title": "How accurate are the annotations of answers in the training data?",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 117344,
      "title": "my submissions take longer time than commiting my notebook",
      "comment_count": 10,
      "views": 0,
      "votes": 2
    },
    {
      "id": 117647,
      "title": "commiting",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 117356,
      "title": "Manual question answering?",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 117619,
      "title": "History of Question Answering Systems",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 116974,
      "title": "One of the best collection of papers about BERT",
      "comment_count": 4,
      "views": 0,
      "votes": 11
    },
    {
      "id": 115349,
      "title": "Kaggle Kernels disconnecting AGAIN and AGAIN",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 115110,
      "title": "Baseline script translation to TF 2.0",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 116153,
      "title": "Using/installing custom packages",
      "comment_count": 3,
      "views": 0,
      "votes": 8
    },
    {
      "id": 116092,
      "title": "missing data in annotations?",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 114771,
      "title": "Can't made a GCP Credits Requests",
      "comment_count": 7,
      "views": 0,
      "votes": 2
    },
    {
      "id": 117018,
      "title": "Ready-to-train dataset - Fully converted",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 116461,
      "title": "Can I upload my offline prediction file as dataset, then read it in kernel and make submission?",
      "comment_count": 7,
      "views": 0,
      "votes": 2
    },
    {
      "id": 116682,
      "title": "Confusion about max public LB score",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 114838,
      "title": "Difference between micro-F1 and F1",
      "comment_count": 1,
      "views": 0,
      "votes": 15
    },
    {
      "id": 115836,
      "title": "Do not fill the NaNs with \"\"!",
      "comment_count": 4,
      "views": 0,
      "votes": 10
    },
    {
      "id": 115179,
      "title": "Submission scoring error",
      "comment_count": 7,
      "views": 0,
      "votes": 2
    },
    {
      "id": 115815,
      "title": "Want to know difference between one-time train and checkpoints train...",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 115752,
      "title": "How to get long answer text?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 115359,
      "title": "Useful Lectures/Materials/Videos",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 115046,
      "title": "Question about Code Only Competition",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 115407,
      "title": "Jeff Dean Explains BERT at TF World Keynote",
      "comment_count": 0,
      "views": 0,
      "votes": 7
    },
    {
      "id": 115492,
      "title": "Do we need to use tensorflow 2.0 only?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 114974,
      "title": "Why is TensorFlow 2 much slower than TensorFlow 1?",
      "comment_count": 3,
      "views": 0,
      "votes": 10
    },
    {
      "id": 115048,
      "title": "Confuse about the kernel running time",
      "comment_count": 3,
      "views": 0,
      "votes": 3
    },
    {
      "id": 114769,
      "title": "BERT for Question and Answering",
      "comment_count": 1,
      "views": 0,
      "votes": 9
    },
    {
      "id": 115071,
      "title": "Tensorflow2.1 will come soon",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 115082,
      "title": "What's the difference between code-only and kernel-only?",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 114767,
      "title": "Starter papers for Open Domain Question Answering",
      "comment_count": 1,
      "views": 0,
      "votes": 8
    },
    {
      "id": 114793,
      "title": "BERT basic",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 114773,
      "title": "why the benchmark is surpassed within 4 hrs after launch?",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 114979,
      "title": "Transfer Learning?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 114829,
      "title": "Simple BERT explaination",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 114841,
      "title": "Interested in teaming up ",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 114758,
      "title": "Confused about the deadline",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 114726,
      "title": "Deleted sorry I was misuderstanding something at first",
      "comment_count": 2,
      "views": 0,
      "votes": null
    }
  ],
  "errors": []
}