{
  "total": 459,
  "topics": [
    {
      "id": 87235,
      "title": "Official External Data Thread",
      "comment_count": 289,
      "views": 0,
      "votes": 34
    },
    {
      "id": 99102,
      "title": "Update: Leaderboard Finalization Timing",
      "comment_count": 17,
      "views": 0,
      "votes": 17
    },
    {
      "id": 96672,
      "title": "IMPORTANT - Stage 2 - Be Careful!",
      "comment_count": 18,
      "views": 0,
      "votes": 24
    },
    {
      "id": 87232,
      "title": "Welcome to the Jigsaw Unintended Bias in Toxicity Classification Challenge!",
      "comment_count": 30,
      "views": 0,
      "votes": 15
    },
    {
      "id": 87719,
      "title": "Some background on this Kernels-Only format",
      "comment_count": 62,
      "views": 0,
      "votes": 65
    },
    {
      "id": 87242,
      "title": "Welcome to Jigsaw's second Kaggle competition!",
      "comment_count": 6,
      "views": 0,
      "votes": 18
    },
    {
      "id": 391405,
      "title": "Notification sound when cell execution completes",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 320593,
      "title": "Uploading Data to Github for easier usage in notebooks",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87868,
      "title": "Getting Started Guide For NLP Competitions",
      "comment_count": 31,
      "views": 0,
      "votes": 187
    },
    {
      "id": 103280,
      "title": "1ST PLACE SOLUTION",
      "comment_count": 15,
      "views": 0,
      "votes": 89
    },
    {
      "id": 91432,
      "title": "Review Your NLP Knowledge",
      "comment_count": 60,
      "views": 0,
      "votes": 329
    },
    {
      "id": 280273,
      "title": "License for this dataset",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 96876,
      "title": "One trick to free Keras GPU memory in Jupyter Notebook",
      "comment_count": 25,
      "views": 0,
      "votes": 50
    },
    {
      "id": 239707,
      "title": "NLP Knowledge",
      "comment_count": 0,
      "views": 0,
      "votes": 5
    },
    {
      "id": 94004,
      "title": "What is a useful data cleaning",
      "comment_count": 12,
      "views": 0,
      "votes": null
    },
    {
      "id": 91994,
      "title": "Attention Layer",
      "comment_count": 12,
      "views": 0,
      "votes": 4
    },
    {
      "id": 88018,
      "title": "Anyone tried BERT?",
      "comment_count": 65,
      "views": 0,
      "votes": 24
    },
    {
      "id": 89871,
      "title": "Sharing some of the beginner's resources I created",
      "comment_count": 12,
      "views": 0,
      "votes": 31
    },
    {
      "id": 88100,
      "title": "cs224n-2019, taught in Pytorch  (can help you with NLP)",
      "comment_count": 7,
      "views": 0,
      "votes": 36
    },
    {
      "id": 153811,
      "title": "Very confused about the weighted loss",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 95398,
      "title": "how to save fastai pytorch model and load it?",
      "comment_count": 29,
      "views": 0,
      "votes": 0
    },
    {
      "id": 122440,
      "title": "Install an external package in kernel while commit",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 131237,
      "title": "Failed submission",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 93347,
      "title": "How to release GPU memory in kernel?",
      "comment_count": 5,
      "views": 0,
      "votes": 2
    },
    {
      "id": 100811,
      "title": "COMBAT WOMBAT 4th palce solution",
      "comment_count": 25,
      "views": 0,
      "votes": 40
    },
    {
      "id": 97471,
      "title": "3rd place solution",
      "comment_count": 37,
      "views": 0,
      "votes": 108
    },
    {
      "id": 92233,
      "title": "How to avoid fastai learn.load() inserting models/.../.pth ?",
      "comment_count": 8,
      "views": 0,
      "votes": 4
    },
    {
      "id": 100661,
      "title": "2nd place solution",
      "comment_count": 42,
      "views": 0,
      "votes": 110
    },
    {
      "id": 90108,
      "title": "Is anyone using the new optimizer AdaBound ?",
      "comment_count": 4,
      "views": 0,
      "votes": 4
    },
    {
      "id": 91081,
      "title": "CUDA Out of Memeory",
      "comment_count": 16,
      "views": 0,
      "votes": 9
    },
    {
      "id": 89966,
      "title": "Free up some memory for training",
      "comment_count": 5,
      "views": 0,
      "votes": 2
    },
    {
      "id": 89981,
      "title": "NLP for beginners",
      "comment_count": 3,
      "views": 0,
      "votes": 12
    },
    {
      "id": 94779,
      "title": "Dynamic Minibatch Trimming to Improve Bert Training Speed (Pytorch)",
      "comment_count": 33,
      "views": 0,
      "votes": 80
    },
    {
      "id": 101630,
      "title": "10th place late writeup[code available now]",
      "comment_count": 7,
      "views": 0,
      "votes": 31
    },
    {
      "id": 122685,
      "title": "My Public Score is zero!",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 91950,
      "title": "\"This file was created from a Kernel, it does not have a description.\"",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 121365,
      "title": "Not able to create new kernel or edit old kernel",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 90629,
      "title": "Practical implementation on NLP research papers!",
      "comment_count": 2,
      "views": 0,
      "votes": 8
    },
    {
      "id": 119919,
      "title": "Additional data published post-competition",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 100821,
      "title": "14th place solution",
      "comment_count": 18,
      "views": 0,
      "votes": 38
    },
    {
      "id": 100961,
      "title": "8th Place Solution (4 models simple avg)",
      "comment_count": 18,
      "views": 0,
      "votes": 57
    },
    {
      "id": 95220,
      "title": "BERT is not the only one",
      "comment_count": 132,
      "views": 0,
      "votes": 159
    },
    {
      "id": 89498,
      "title": "Efficient GPU and Memory Usage",
      "comment_count": 2,
      "views": 0,
      "votes": 16
    },
    {
      "id": 97135,
      "title": "Distilling BERT — BERT performance using Logistic Regression",
      "comment_count": 7,
      "views": 0,
      "votes": 33
    },
    {
      "id": 95166,
      "title": "Always happening Preparing your Kernel",
      "comment_count": 5,
      "views": 0,
      "votes": 3
    },
    {
      "id": 93557,
      "title": "when I load keras model ,I get the kernel error,need help",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 93230,
      "title": "Don't Save Embedding in RNN Solutions ",
      "comment_count": 53,
      "views": 0,
      "votes": 61
    },
    {
      "id": 92977,
      "title": "BERT Base vs/ BERT Large",
      "comment_count": 29,
      "views": 0,
      "votes": 10
    },
    {
      "id": 90946,
      "title": "Non-deterministic result using Keras cuDNN",
      "comment_count": 14,
      "views": 0,
      "votes": 5
    },
    {
      "id": 87879,
      "title": "403 - This competition only allows kernel submissions.",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 101927,
      "title": "FULL CODE 4th place solution",
      "comment_count": 18,
      "views": 0,
      "votes": 54
    },
    {
      "id": 92867,
      "title": "BERT Kernels 😁",
      "comment_count": 11,
      "views": 0,
      "votes": 18
    },
    {
      "id": 100530,
      "title": "9th place solution",
      "comment_count": 19,
      "views": 0,
      "votes": 50
    },
    {
      "id": 100808,
      "title": "106th place solution (21 models blend)",
      "comment_count": 6,
      "views": 0,
      "votes": 13
    },
    {
      "id": 105085,
      "title": "Kernel stuck at \"Preparing your kernel...\" ",
      "comment_count": 1,
      "views": 0,
      "votes": 2
    },
    {
      "id": 97389,
      "title": "Kagengers: How to fall from top (2nd public/very low private)",
      "comment_count": 60,
      "views": 0,
      "votes": 105
    },
    {
      "id": 103427,
      "title": "FAQ: When and where will the data sets in this competition be released after the competition finishes?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 102506,
      "title": "Next NLP competition?",
      "comment_count": 7,
      "views": 0,
      "votes": 10
    },
    {
      "id": 100904,
      "title": "The cool story behind the gold",
      "comment_count": 20,
      "views": 0,
      "votes": 70
    },
    {
      "id": 101263,
      "title": "1st place solution coming soon?",
      "comment_count": 10,
      "views": 0,
      "votes": 14
    },
    {
      "id": 94848,
      "title": "BERT on TPU vs GPU",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 97435,
      "title": "Top Solution Compilation (UPDATED)",
      "comment_count": 8,
      "views": 0,
      "votes": 25
    },
    {
      "id": 101218,
      "title": "How *not* to remove bias in machine learning models",
      "comment_count": 3,
      "views": 0,
      "votes": 23
    },
    {
      "id": 101965,
      "title": "Nuances of bias in machine learning (In Dimension C-132)",
      "comment_count": 0,
      "views": 0,
      "votes": 5
    },
    {
      "id": 100611,
      "title": "7th Place Solution",
      "comment_count": 6,
      "views": 0,
      "votes": 46
    },
    {
      "id": 97422,
      "title": "Public LB Top 10, Private LB Top 26 Solution (Single Bert - 0.94560)",
      "comment_count": 30,
      "views": 0,
      "votes": 57
    },
    {
      "id": 101690,
      "title": "Fake 13.5th Solution (Real 77th) NO PytorchBERT only keras",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 100708,
      "title": "16th @ Jigsaw Unintended Bias",
      "comment_count": 5,
      "views": 0,
      "votes": 33
    },
    {
      "id": 101421,
      "title": "Interview w Ryan Chesler (AAA Team), 12th Pos.",
      "comment_count": 2,
      "views": 0,
      "votes": 8
    },
    {
      "id": 100718,
      "title": "5th place",
      "comment_count": 4,
      "views": 0,
      "votes": 29
    },
    {
      "id": 100799,
      "title": "11th place solution",
      "comment_count": 7,
      "views": 0,
      "votes": 17
    },
    {
      "id": 88604,
      "title": "Keras vs Pytorch vs fastai?",
      "comment_count": 9,
      "views": 0,
      "votes": 10
    },
    {
      "id": 100521,
      "title": "Rerun failed reason(have known)",
      "comment_count": 9,
      "views": 0,
      "votes": 11
    },
    {
      "id": 100713,
      "title": "32nd Place Solution Summary",
      "comment_count": 5,
      "views": 0,
      "votes": 16
    },
    {
      "id": 100936,
      "title": "Any technology to cut down keras inference time？",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 100716,
      "title": "Finetuning word vector",
      "comment_count": 0,
      "views": 0,
      "votes": 6
    },
    {
      "id": 100480,
      "title": "Finally got the results",
      "comment_count": 5,
      "views": 0,
      "votes": 5
    },
    {
      "id": 100559,
      "title": "Has someone got gold using only Kaggle kernels?",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 100478,
      "title": "Rerun phase has just started. Good luck for everyone ))",
      "comment_count": 11,
      "views": 0,
      "votes": 2
    },
    {
      "id": 97443,
      "title": "29th LB And  LB 0.94376 for single Bert-Base ",
      "comment_count": 1,
      "views": 0,
      "votes": 22
    },
    {
      "id": 100543,
      "title": "Is there any way to see the private lb of other submissions?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 100490,
      "title": "Submissions that survived the rerun",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 100486,
      "title": "Only one Kernel being run out of two selections",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 100479,
      "title": "Private LB Scores!",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 97487,
      "title": "Kernel For Training Bert (The real stuff)",
      "comment_count": 13,
      "views": 0,
      "votes": 63
    },
    {
      "id": 99591,
      "title": "has anyone’s kernel been run yet?",
      "comment_count": 9,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97484,
      "title": "52th Public Solution – Newbie Lessons learned",
      "comment_count": 11,
      "views": 0,
      "votes": 24
    },
    {
      "id": 99323,
      "title": "When on earth can the dataset be able to be downloaded?",
      "comment_count": 1,
      "views": 0,
      "votes": -8
    },
    {
      "id": 97400,
      "title": "Loss function and a short report (LB: 5)",
      "comment_count": 5,
      "views": 0,
      "votes": 30
    },
    {
      "id": 98376,
      "title": "When the LB can be finalized?",
      "comment_count": 23,
      "views": 0,
      "votes": 4
    },
    {
      "id": 99122,
      "title": "From 2 to 11 days to go",
      "comment_count": 7,
      "views": 0,
      "votes": 2
    },
    {
      "id": 97468,
      "title": "How did u choose the final submissions?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 97464,
      "title": "27th Public Solution (kernels only)",
      "comment_count": 29,
      "views": 0,
      "votes": 55
    },
    {
      "id": 97425,
      "title": "UET: 18th Private LB (with code)",
      "comment_count": 27,
      "views": 0,
      "votes": 55
    },
    {
      "id": 98960,
      "title": "late submission",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 98852,
      "title": "So... How to use bert in this task appropriately?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 97784,
      "title": "Bert 2 epochs in a single kernel run",
      "comment_count": 8,
      "views": 0,
      "votes": 15
    },
    {
      "id": 98441,
      "title": "Sklearn Logistic Regression Solution!",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97549,
      "title": "Why the limit is 13 days to go?",
      "comment_count": 11,
      "views": 0,
      "votes": 3
    },
    {
      "id": 96879,
      "title": "Please stop private sharing",
      "comment_count": 86,
      "views": 0,
      "votes": 59
    },
    {
      "id": 97268,
      "title": "Good luck to all",
      "comment_count": 28,
      "views": 0,
      "votes": 39
    },
    {
      "id": 97914,
      "title": "Competitions versus real-life projects",
      "comment_count": 5,
      "views": 0,
      "votes": 11
    },
    {
      "id": 97399,
      "title": "Simple tricks to get a solid 0.9420 for single Bert-Base",
      "comment_count": 7,
      "views": 0,
      "votes": 28
    },
    {
      "id": 96773,
      "title": "Finetune XLNet in PyTorch",
      "comment_count": 45,
      "views": 0,
      "votes": 70
    },
    {
      "id": 98089,
      "title": "Best Single BERT Model LB score",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97584,
      "title": "Please share how did you effectively use Kaggle kernels",
      "comment_count": 6,
      "views": 0,
      "votes": 3
    },
    {
      "id": 96714,
      "title": "Authors, Please don't delete public datasets",
      "comment_count": 65,
      "views": 0,
      "votes": 14
    },
    {
      "id": 97463,
      "title": "Another loss weighting",
      "comment_count": 1,
      "views": 0,
      "votes": 17
    },
    {
      "id": 92668,
      "title": "Introducing FastBert ",
      "comment_count": 14,
      "views": 0,
      "votes": 49
    },
    {
      "id": 97413,
      "title": "Simple BERT trick for improving score (head+tail method)",
      "comment_count": 10,
      "views": 0,
      "votes": 6
    },
    {
      "id": 97405,
      "title": "We can run 2nd-stage test-set now !",
      "comment_count": 9,
      "views": 0,
      "votes": 15
    },
    {
      "id": 97401,
      "title": "What's your best LB score for single model?",
      "comment_count": 19,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97391,
      "title": "Tricks to make the last position of silver zone (public) in 6 days.",
      "comment_count": 6,
      "views": 0,
      "votes": 8
    },
    {
      "id": 97387,
      "title": "Lesson #1 for Kernels Competitions. (and for life)",
      "comment_count": 4,
      "views": 0,
      "votes": 8
    },
    {
      "id": 97308,
      "title": "Kernel manipulation",
      "comment_count": 9,
      "views": 0,
      "votes": 14
    },
    {
      "id": 97219,
      "title": "Where is the 51th team",
      "comment_count": 7,
      "views": 0,
      "votes": 6
    },
    {
      "id": 97218,
      "title": "Good Luck to all!",
      "comment_count": 2,
      "views": 0,
      "votes": 15
    },
    {
      "id": 97272,
      "title": "kernel re-run and dataset versions",
      "comment_count": 5,
      "views": 0,
      "votes": 5
    },
    {
      "id": 97453,
      "title": "One trick teams may miss",
      "comment_count": 0,
      "views": 0,
      "votes": 10
    },
    {
      "id": 97316,
      "title": "Shake-up or not ?",
      "comment_count": 4,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97262,
      "title": "[To admin] Submission Error Message: Request could not be processed, even after multiple attempts.",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 97106,
      "title": "Bert pretrained models package added",
      "comment_count": 3,
      "views": 0,
      "votes": 5
    },
    {
      "id": 97565,
      "title": "Lesson learned from this competition",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97386,
      "title": "Fight for the 1st place! 🥇",
      "comment_count": 2,
      "views": 0,
      "votes": 6
    },
    {
      "id": 97395,
      "title": "Do we have to wait for 2 weeks to see the results? (My first kernel competition)",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96306,
      "title": "What should I do in the BERT Second Epoch",
      "comment_count": 38,
      "views": 0,
      "votes": 22
    },
    {
      "id": 97393,
      "title": "Can't wait for top solutions 🥇",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 97132,
      "title": "How long is your kernel run?",
      "comment_count": 12,
      "views": 0,
      "votes": 2
    },
    {
      "id": 97459,
      "title": "Second phase submission are 3% faster",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96742,
      "title": "Leaderboard analysis. Who will win?",
      "comment_count": 13,
      "views": 0,
      "votes": 6
    },
    {
      "id": 97162,
      "title": "Kernels disconnecting and reconnecting over and over",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96956,
      "title": "MASS – A pre-training method that outperforms BERT and GPT",
      "comment_count": 4,
      "views": 0,
      "votes": 16
    },
    {
      "id": 97150,
      "title": "Is it OK to output more than submission.csv for stage 2?",
      "comment_count": 3,
      "views": 0,
      "votes": 4
    },
    {
      "id": 97064,
      "title": "Which version will be selected at time of submission",
      "comment_count": 8,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95531,
      "title": "On my way to another race",
      "comment_count": 50,
      "views": 0,
      "votes": 131
    },
    {
      "id": 96675,
      "title": "kernel environment",
      "comment_count": 11,
      "views": 0,
      "votes": 11
    },
    {
      "id": 96840,
      "title": "Surprising ModuleNotFoundError",
      "comment_count": 4,
      "views": 0,
      "votes": 4
    },
    {
      "id": 97370,
      "title": "Last 2 Hours to run a notebook",
      "comment_count": 2,
      "views": 0,
      "votes": -1
    },
    {
      "id": 97358,
      "title": "Can't wait to good solutions from top or good answers",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 96980,
      "title": "Some Questions about Stage2",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 97390,
      "title": "Ods.ai: Just submitted registration request",
      "comment_count": 1,
      "views": 0,
      "votes": -3
    },
    {
      "id": 97172,
      "title": "Why only 5 submissions available ?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 96674,
      "title": "XLNet is okay to use",
      "comment_count": 14,
      "views": 0,
      "votes": 6
    },
    {
      "id": 97130,
      "title": "Model stops predicting on 2600s",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 97111,
      "title": "[SOLVED]kernel not working suddenly while it was good weeks ago",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 97123,
      "title": "kernel error - libgcc_s.so.1 must be installed for pthread_cancel to work",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96808,
      "title": "head + tail",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 97155,
      "title": "how to save and load fastai model?",
      "comment_count": 5,
      "views": 0,
      "votes": -3
    },
    {
      "id": 96995,
      "title": "Is it OK to run .py script in the Kernel for Stage 2?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 96928,
      "title": "Are 7200 sec of runtime enough for this competition?",
      "comment_count": 7,
      "views": 0,
      "votes": null
    },
    {
      "id": 96935,
      "title": "Are we able to save pre-trained models and load it to a new kernel?",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96972,
      "title": "Can not upload  trained model",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 96377,
      "title": "XLNET",
      "comment_count": 17,
      "views": 0,
      "votes": 15
    },
    {
      "id": 96669,
      "title": "+1 where you should be careful",
      "comment_count": 6,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96408,
      "title": "Ask to kaggle, Is it possible to use xlnet?",
      "comment_count": 11,
      "views": 0,
      "votes": 9
    },
    {
      "id": 96834,
      "title": "Initializing bert config in new version of huggigface repo.",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 96690,
      "title": "SyntaxError: import * only allowed at module level",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96637,
      "title": "Quick Question To Admins: Can we use XLNet?",
      "comment_count": 5,
      "views": 0,
      "votes": 4
    },
    {
      "id": 96809,
      "title": "Expand contraction",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 87340,
      "title": "Kaggle, we'd like to unleash the power of BERT, GPT, ELMo, ULMFiT and other beasts",
      "comment_count": 47,
      "views": 0,
      "votes": 57
    },
    {
      "id": 95054,
      "title": "Lets prevent a heart break ",
      "comment_count": 18,
      "views": 0,
      "votes": 50
    },
    {
      "id": 96842,
      "title": "Help !!!I have crazy about 7400 seconds?",
      "comment_count": 3,
      "views": 0,
      "votes": -4
    },
    {
      "id": 96791,
      "title": "Confusion about the Official External Data Thread",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 96284,
      "title": "Exploring the Bias Metric",
      "comment_count": 4,
      "views": 0,
      "votes": 15
    },
    {
      "id": 95603,
      "title": "Fine-Tune BERT for Text Classification",
      "comment_count": 6,
      "views": 0,
      "votes": 25
    },
    {
      "id": 96535,
      "title": "Best usae of info in Identity Columns ",
      "comment_count": 1,
      "views": 0,
      "votes": -1
    },
    {
      "id": 96782,
      "title": "fastai have Problem",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 95913,
      "title": "Talented Newcomer",
      "comment_count": 14,
      "views": 0,
      "votes": 6
    },
    {
      "id": 95371,
      "title": "Where do you train your model? ",
      "comment_count": 8,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96395,
      "title": "Try your model with a fake test",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96345,
      "title": "Is keras bert behave worst than pytorch version?",
      "comment_count": 4,
      "views": 0,
      "votes": 3
    },
    {
      "id": 96577,
      "title": "High Score on LB using SVM",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 96157,
      "title": "Is it unfair to ask for large Bert model weights for this competition?",
      "comment_count": 29,
      "views": 0,
      "votes": -4
    },
    {
      "id": 91240,
      "title": "Beware of keras tokenizer bugs",
      "comment_count": 2,
      "views": 0,
      "votes": 10
    },
    {
      "id": 96299,
      "title": "[SOLVED] import finetuned pretrained BERT model into kaggle kernel?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 96565,
      "title": "Cannot upload model as dateset - getting files exceeded the total size error.",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 96468,
      "title": "Bert inference   (outputs don't match) ",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96287,
      "title": "Reference for mitigating unintented bias",
      "comment_count": 1,
      "views": 0,
      "votes": 9
    },
    {
      "id": 96349,
      "title": "Funny 2999 teams",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96050,
      "title": "An excellent user friendly explanation of LSTMs",
      "comment_count": 3,
      "views": 0,
      "votes": 7
    },
    {
      "id": 94994,
      "title": "What is your best *single BERT* model?",
      "comment_count": 36,
      "views": 0,
      "votes": 17
    },
    {
      "id": 96286,
      "title": "How to prevent shakeup between public LB and private LB?",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96380,
      "title": "XLNet: Outperforming Bert 20 on Tasks !",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96336,
      "title": "Kind of related",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96143,
      "title": "Develop a Weighted Average Ensemble",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 96096,
      "title": "Team up to the Top 10",
      "comment_count": 20,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96233,
      "title": "anyone interested in teaming up?",
      "comment_count": 1,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96074,
      "title": "Teaming up for gold (20th place, 0.94545)",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 95959,
      "title": "Say NO to Magic Numbers",
      "comment_count": 4,
      "views": 0,
      "votes": 16
    },
    {
      "id": 95911,
      "title": "How to improve the speed of bert prediction?",
      "comment_count": 6,
      "views": 0,
      "votes": 2
    },
    {
      "id": 96209,
      "title": "Is this a kernel feature update?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 96037,
      "title": "I want a team and I have a bert model which can get 0.94277 LB, but I don't know how to blend",
      "comment_count": 9,
      "views": 0,
      "votes": 0
    },
    {
      "id": 96139,
      "title": "Any dalao would like to share your gpt-2 kernel model for this competition?",
      "comment_count": 3,
      "views": 0,
      "votes": -7
    },
    {
      "id": 95368,
      "title": "Bert large whole-word-masking models",
      "comment_count": 1,
      "views": 0,
      "votes": 12
    },
    {
      "id": 95658,
      "title": "Congratulations Dieter for becoming Kernels GM",
      "comment_count": 4,
      "views": 0,
      "votes": 14
    },
    {
      "id": 95459,
      "title": "Any one tried BERT in TensorFlow",
      "comment_count": 6,
      "views": 0,
      "votes": 3
    },
    {
      "id": 95970,
      "title": "how to load gpt2 model in  kernel",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 95982,
      "title": "How to ensemble? Does anyone implement stacking or other ensemble techniques?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 95001,
      "title": "[off topic] Replication of GPT2-1.5B",
      "comment_count": 3,
      "views": 0,
      "votes": 16
    },
    {
      "id": 95397,
      "title": "What is your LOCAL VALIDATION score",
      "comment_count": 6,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95892,
      "title": "Mitigate unintented bias",
      "comment_count": 1,
      "views": 0,
      "votes": 0
    },
    {
      "id": 95885,
      "title": "How much training size did you use by bert?",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 95620,
      "title": "Anyone doing well w Keras/TF & might want me on your team?",
      "comment_count": 4,
      "views": 0,
      "votes": 2
    },
    {
      "id": 95733,
      "title": "Can we use two different Deep Learning framework in a single kernels. Like - TensorFlow and Pytorch",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95825,
      "title": "Any one tried TensorFlow Hub",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 93339,
      "title": "What's your best model?",
      "comment_count": 105,
      "views": 0,
      "votes": 42
    },
    {
      "id": 95651,
      "title": "Why choose Glove instead of Word2vec?",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 95466,
      "title": "Looking for Team",
      "comment_count": 8,
      "views": 0,
      "votes": null
    },
    {
      "id": 95639,
      "title": "Looking for Teammate",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95310,
      "title": "Team Merging Request",
      "comment_count": 7,
      "views": 0,
      "votes": 2
    },
    {
      "id": 95339,
      "title": "Mixing BERT and GloVe embeddings",
      "comment_count": 1,
      "views": 0,
      "votes": 5
    },
    {
      "id": 95101,
      "title": "Normal features layer in LSTM",
      "comment_count": 7,
      "views": 0,
      "votes": 4
    },
    {
      "id": 95159,
      "title": "feature engineering",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 95345,
      "title": "Want to know about pre-trained language models (BERT....)?",
      "comment_count": 0,
      "views": 0,
      "votes": 10
    },
    {
      "id": 94681,
      "title": "While waiting for BERT to finish",
      "comment_count": 10,
      "views": 0,
      "votes": 29
    },
    {
      "id": 95570,
      "title": "How good is the bert base model, where is the error and the causes?",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95150,
      "title": "Issues with translating data (too many requests)",
      "comment_count": 4,
      "views": 0,
      "votes": 8
    },
    {
      "id": 95530,
      "title": "Help about the LB score not change when I continue train my bert.",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 94833,
      "title": "Why label should be set to 0,1",
      "comment_count": 6,
      "views": 0,
      "votes": 3
    },
    {
      "id": 95445,
      "title": "Torchtext, Embeddings loading from cache",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95528,
      "title": "I got many 0.94157. I have many models get LB 0.94157, Interesting.",
      "comment_count": 2,
      "views": 0,
      "votes": -3
    },
    {
      "id": 95213,
      "title": "Focal loss- way to go?",
      "comment_count": 5,
      "views": 0,
      "votes": -13
    },
    {
      "id": 95340,
      "title": "Any Experience with GloVe Twitter Embeddings",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95109,
      "title": "My kernel break because of memory limit ",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 95265,
      "title": "Why do Keras kernels seem to show lower performances than Pytorch",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 95060,
      "title": "Accidentally deleted kernel!",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95134,
      "title": "Text augmentation ",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 95096,
      "title": "what are the things that I can try?",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 95114,
      "title": "Looking for a teammate",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 95112,
      "title": "Teammate wanted",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 94955,
      "title": "Sorted Minibatch Prediction to Improve Bert Inference Speed (Pytorch)",
      "comment_count": 0,
      "views": 0,
      "votes": 9
    },
    {
      "id": 94602,
      "title": "LM finetuning",
      "comment_count": 8,
      "views": 0,
      "votes": null
    },
    {
      "id": 95045,
      "title": "problems with kernels?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 95032,
      "title": "Does anyone find LB test unstable?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 94985,
      "title": "Wonder if anyone actually check the comment and label to see if they make sense",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 94800,
      "title": "is there a way to average the word embeddings?",
      "comment_count": 14,
      "views": 0,
      "votes": 3
    },
    {
      "id": 94738,
      "title": "What is the best indicator of overfitting?",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 94792,
      "title": "Trends of top 20 teams up to this point",
      "comment_count": 3,
      "views": 0,
      "votes": 11
    },
    {
      "id": 94939,
      "title": "Batch Size and learning rate for bert ",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 91691,
      "title": "Non-BERT Single Model Score",
      "comment_count": 35,
      "views": 0,
      "votes": 12
    },
    {
      "id": 94677,
      "title": "Fast Computation of convert_lines in bert ?",
      "comment_count": 5,
      "views": 0,
      "votes": 2
    },
    {
      "id": 94904,
      "title": "Alternative ways to use bert",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 94678,
      "title": "Managing Ensemble <120 Min?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 94552,
      "title": "How to detect if two phrases have the same information?",
      "comment_count": 1,
      "views": 0,
      "votes": 7
    },
    {
      "id": 94784,
      "title": "Jumping Back In",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 92737,
      "title": "How does bert handles unrecognized words?",
      "comment_count": 34,
      "views": 0,
      "votes": 22
    },
    {
      "id": 93969,
      "title": "Can I combine Bert with LSTM in model level?",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 94367,
      "title": "Train BERT for 1 epoch.",
      "comment_count": 9,
      "views": 0,
      "votes": -2
    },
    {
      "id": 94393,
      "title": "Kernel competition rules ",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 94041,
      "title": "Is it FAIR?",
      "comment_count": 32,
      "views": 0,
      "votes": 1
    },
    {
      "id": 94525,
      "title": "How to use old version of a dataset in kernel?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 92505,
      "title": "Kernel for training BERT using Pytorch",
      "comment_count": 57,
      "views": 0,
      "votes": 140
    },
    {
      "id": 94576,
      "title": "Should there not be time shown while commit?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 94228,
      "title": "FastAI vs Keras+TensorFlow",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 93550,
      "title": "Fast & unified embeddings loading",
      "comment_count": 2,
      "views": 0,
      "votes": 19
    },
    {
      "id": 94301,
      "title": "Is the RNN kernels using the correct metric?",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 94260,
      "title": "Can we use BERT in stage2?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 94384,
      "title": "Load pretrain model in stage2 ?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90447,
      "title": "Your kernel cannot use kernel outputs as a data source for this competition.",
      "comment_count": 6,
      "views": 0,
      "votes": 3
    },
    {
      "id": 94253,
      "title": "Do we predict binary or fraction values for target?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 94100,
      "title": "How can I upload my model trained on kaggle to kaggle's dataset?",
      "comment_count": 10,
      "views": 0,
      "votes": null
    },
    {
      "id": 93457,
      "title": "Regular Expression. Cleaning text",
      "comment_count": 1,
      "views": 0,
      "votes": 4
    },
    {
      "id": 90487,
      "title": "When we get new test data...",
      "comment_count": 7,
      "views": 0,
      "votes": 4
    },
    {
      "id": 94009,
      "title": "Question on the target/labels",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 93905,
      "title": "How to interpret the Magic Blending results?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 92872,
      "title": "Which cloud is best to pre-train BERT?",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 92186,
      "title": "LF teammate",
      "comment_count": 5,
      "views": 0,
      "votes": 0
    },
    {
      "id": 90291,
      "title": "Kaggle reading group videos on NLP research papers",
      "comment_count": 14,
      "views": 0,
      "votes": 62
    },
    {
      "id": 92487,
      "title": "Why fastai public kernel is better than pytorch public kernels?",
      "comment_count": 2,
      "views": 0,
      "votes": 5
    },
    {
      "id": 93872,
      "title": " Pytorch - Save/Load Model and predict ",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 93614,
      "title": "! pip install giving SyntaxErrorr in kernel script",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 93436,
      "title": "Ensemble question ",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 93346,
      "title": "Ideas for what to look into beyond BERT",
      "comment_count": 1,
      "views": 0,
      "votes": 17
    },
    {
      "id": 93369,
      "title": "Is everyone using the weighted cross-entropy loss?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 93719,
      "title": "Sharing Data between kernels",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 91956,
      "title": "How is everyone dealing with memory constraints in the kernel?",
      "comment_count": 7,
      "views": 0,
      "votes": null
    },
    {
      "id": 91475,
      "title": "Looking for a teammate! ",
      "comment_count": 6,
      "views": 0,
      "votes": 2
    },
    {
      "id": 93200,
      "title": "Batch size ",
      "comment_count": 7,
      "views": 0,
      "votes": 7
    },
    {
      "id": 93397,
      "title": "BERT",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 93348,
      "title": "any advice on save and load model using fastai?",
      "comment_count": 6,
      "views": 0,
      "votes": 4
    },
    {
      "id": 93171,
      "title": "BERT validation and lb scores",
      "comment_count": 2,
      "views": 0,
      "votes": 4
    },
    {
      "id": 93332,
      "title": "BERT in R",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 93260,
      "title": "AssertionError: Torch not compiled with CUDA enabled",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91777,
      "title": "What's the point of (CV) Cross Validation?",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 93329,
      "title": "AttentionWithContext Layer for Tensorflow 1.13 with Eager Execution.",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91982,
      "title": "What's your best BERT single model score?",
      "comment_count": 71,
      "views": 0,
      "votes": 24
    },
    {
      "id": 87586,
      "title": "How to Speed Things Up??",
      "comment_count": 28,
      "views": 0,
      "votes": 30
    },
    {
      "id": 90527,
      "title": "Making evaluation easy",
      "comment_count": 14,
      "views": 0,
      "votes": 114
    },
    {
      "id": 93194,
      "title": "how can I save my preprocessed text and load efficiently?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 89876,
      "title": "Anyone tried Universal Sentence Encoder or ELMo?",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 92192,
      "title": "Some tips on preprocessing for GloVe",
      "comment_count": 7,
      "views": 0,
      "votes": 71
    },
    {
      "id": 92304,
      "title": "Teaming up?",
      "comment_count": 3,
      "views": 0,
      "votes": 3
    },
    {
      "id": 90888,
      "title": "'rating' data",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 93012,
      "title": "why can not use all train data to train by using pre-trained bert model",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 92864,
      "title": "Don't stop",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 92811,
      "title": "two or three kinds of word embedding",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 91972,
      "title": "why removing punctuations when there is actually embeddings for them",
      "comment_count": 7,
      "views": 0,
      "votes": 4
    },
    {
      "id": 92477,
      "title": "Please update fastai and clarify \"custom packages\".",
      "comment_count": 4,
      "views": 0,
      "votes": 17
    },
    {
      "id": 92880,
      "title": "Language Translation",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 92850,
      "title": "InvalidArgumentError",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 92804,
      "title": "Pytorch DataLoader Problem ",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 92555,
      "title": "A picture tell you how to do model ensemble . (blending/stacking)",
      "comment_count": 3,
      "views": 0,
      "votes": 3
    },
    {
      "id": 92661,
      "title": "Question about BERT GPU memory used",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 92649,
      "title": "Memory profiling in Python",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 90936,
      "title": "my first NLP challenge in Kaggle",
      "comment_count": 39,
      "views": 0,
      "votes": 148
    },
    {
      "id": 90202,
      "title": "how high can the metric be for a single model?",
      "comment_count": 52,
      "views": 0,
      "votes": 24
    },
    {
      "id": 92582,
      "title": "Problem about the rules?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 92545,
      "title": "could there be any tips to save time?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 90143,
      "title": "Ratio of Toxic to Non-Toxic for Key Subgroups",
      "comment_count": 2,
      "views": 0,
      "votes": 5
    },
    {
      "id": 92375,
      "title": "CV constantly higher than LB",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 92384,
      "title": "why learning rate is so sensitive?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 92378,
      "title": "Can I save my model to break the time limitation?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91619,
      "title": "Is your cv score consistent with lb?",
      "comment_count": 15,
      "views": 0,
      "votes": null
    },
    {
      "id": 91928,
      "title": "Adversarial Learning for Debiasing",
      "comment_count": 4,
      "views": 0,
      "votes": 5
    },
    {
      "id": 92140,
      "title": "what's the best max_seq_length for bert?",
      "comment_count": 6,
      "views": 0,
      "votes": 2
    },
    {
      "id": 91848,
      "title": "Knowledge Distillation",
      "comment_count": 8,
      "views": 0,
      "votes": 3
    },
    {
      "id": 92149,
      "title": "Fast convert_lines",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 92146,
      "title": "Data Augmentation Ideas",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 91513,
      "title": "Looking to join / create a team!",
      "comment_count": 8,
      "views": 0,
      "votes": null
    },
    {
      "id": 92021,
      "title": "Why, if I do something, my results get worse?",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91918,
      "title": "Loss Function",
      "comment_count": 12,
      "views": 0,
      "votes": 2
    },
    {
      "id": 91528,
      "title": "How to save and load the fine-tuning BERT model?",
      "comment_count": 9,
      "views": 0,
      "votes": 4
    },
    {
      "id": 91523,
      "title": "From where can I learn NLP from the very basic concept.",
      "comment_count": 6,
      "views": 0,
      "votes": 1
    },
    {
      "id": 92068,
      "title": "id, Parent_id, Publication_id, Article_id",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 91168,
      "title": "kaggle manager for automating workflow for kernel only competitions",
      "comment_count": 1,
      "views": 0,
      "votes": 4
    },
    {
      "id": 91822,
      "title": "Should I trust my model ?",
      "comment_count": 6,
      "views": 0,
      "votes": -1
    },
    {
      "id": 91896,
      "title": "Is there a limit to use of number of external models?",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 91702,
      "title": "New text cleaning functions",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 91830,
      "title": "Train Models Locally or By Kernel",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 91672,
      "title": "How to Ensemble?",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91385,
      "title": "Logistic Regression Classifier With 89% accuracy",
      "comment_count": 4,
      "views": 0,
      "votes": 5
    },
    {
      "id": 88533,
      "title": "Bert stops after several hours of training",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 91682,
      "title": "How do you set trainable or non-trainable when using Bert?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90426,
      "title": "Flair RNN document embeddings--Useful?",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 91533,
      "title": "Problem with editing my kernel",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91576,
      "title": "A Thought Provoking Kernel for your Competition Consideration",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 91696,
      "title": "How to set up CV for Bias Detection ",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 90988,
      "title": "How to submit prediction",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 91581,
      "title": "Sequence bucketing with Attention",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 91578,
      "title": "Labeling Ambiguity in Training Data?",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 90753,
      "title": "Google's Perspective API demo for Jigsaw",
      "comment_count": 6,
      "views": 0,
      "votes": 3
    },
    {
      "id": 91272,
      "title": "How to optimize for evaluation metric instead of BCE?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 91422,
      "title": "Which of the subgroups will be used for AUC evaluation",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 88173,
      "title": "[solved] How do you cope with the duplication in training data?",
      "comment_count": 7,
      "views": 0,
      "votes": 4
    },
    {
      "id": 91552,
      "title": "Is pytorch faster than keras?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 89672,
      "title": "Whats your CV AUC like?",
      "comment_count": 36,
      "views": 0,
      "votes": 14
    },
    {
      "id": 91529,
      "title": "Request for advice on how to proceed with NLP Competitions",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 91093,
      "title": "Many words are not found in embeddings because of a strange font",
      "comment_count": 6,
      "views": 0,
      "votes": 13
    },
    {
      "id": 88220,
      "title": "How to augment text in this Comp",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 91094,
      "title": "Torchtext. Who used?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91445,
      "title": "Some cool resources about Machine Learning Fairness",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87756,
      "title": "Some Insights about the last Quora Competition",
      "comment_count": 10,
      "views": 0,
      "votes": 41
    },
    {
      "id": 90014,
      "title": "Need Team members",
      "comment_count": 15,
      "views": 0,
      "votes": null
    },
    {
      "id": 90986,
      "title": "How ROC-AUC evaluation metric is calculated?",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 91097,
      "title": "Is it better to add paddings at the start or at the end of the text?",
      "comment_count": 5,
      "views": 0,
      "votes": 2
    },
    {
      "id": 90958,
      "title": "How to build custom loss function?",
      "comment_count": 2,
      "views": 0,
      "votes": 5
    },
    {
      "id": 91235,
      "title": "Are manually handcrafted dictionaries allowed?",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 91026,
      "title": "some common errors encountered when you do NLP",
      "comment_count": 4,
      "views": 0,
      "votes": 5
    },
    {
      "id": 91200,
      "title": "Exit Code 137 During K Fold CV",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 91113,
      "title": "Is it a bad idea to update word embeddings during training ?",
      "comment_count": 7,
      "views": 0,
      "votes": null
    },
    {
      "id": 91198,
      "title": "Fast estimation of AUC for validation",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 90912,
      "title": "Is this competition friendly to beginners?",
      "comment_count": 6,
      "views": 0,
      "votes": 2
    },
    {
      "id": 91108,
      "title": "Does embeddings include Capitalized words?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 91024,
      "title": "model over the tf.keras.models.Model/.Model problem",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 88558,
      "title": "LB0.93351 best 6 kernel!",
      "comment_count": 11,
      "views": 0,
      "votes": 13
    },
    {
      "id": 90706,
      "title": "Questions about the rules.",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90794,
      "title": "Questions about external data",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 90647,
      "title": "What embeddings are you Using?",
      "comment_count": 4,
      "views": 0,
      "votes": 3
    },
    {
      "id": 89695,
      "title": "Well i create a kernel; how can i upload an embedding?",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 89897,
      "title": "How to prepare a kernel for running on private test set",
      "comment_count": 5,
      "views": 0,
      "votes": 3
    },
    {
      "id": 90836,
      "title": "Vocabulary problem",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90623,
      "title": "Fixing random seeds",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90218,
      "title": "Workaround for \"Unknown backend error\" or GCS error in commit log",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90721,
      "title": "A Recipe for Training Neural Networks",
      "comment_count": 1,
      "views": 0,
      "votes": 6
    },
    {
      "id": 90765,
      "title": "Final Evaluation Metric Callback for FastAI ",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 89773,
      "title": "Probably better to enable non-kernel submission in Stage 1",
      "comment_count": 5,
      "views": 0,
      "votes": 21
    },
    {
      "id": 90793,
      "title": "Hinge Loss vs Cross Entropy",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 90711,
      "title": "How do we select lstm or gru as encoder/decoder?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90551,
      "title": "Exited with code 137",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90746,
      "title": "Load a tensorflow model on the kaggle kernel issue",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87325,
      "title": "Some of my commited kernels never complete",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90246,
      "title": "Is blending legal?",
      "comment_count": 8,
      "views": 0,
      "votes": 2
    },
    {
      "id": 89530,
      "title": "How to convert multiple json to a single dataframe ?",
      "comment_count": 4,
      "views": 0,
      "votes": 2
    },
    {
      "id": 90589,
      "title": "Old toxic dataset",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90475,
      "title": "Applying weight during training to reduce bias",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 90435,
      "title": "Why published kernels are trained for 4 epochs?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90526,
      "title": "< 0.25 rule not kicking in",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90456,
      "title": "Part of Speech Tagging kernel",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 90357,
      "title": "Running, running,... for a long time",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90422,
      "title": "Using additional columns for training...",
      "comment_count": 3,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87286,
      "title": "First time 1st place",
      "comment_count": 15,
      "views": 0,
      "votes": 8
    },
    {
      "id": 90453,
      "title": "Different Kernel Run Times",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90449,
      "title": "how do you speed up training process",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90425,
      "title": "Using Multiple GPUs",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 90356,
      "title": "NN or LightBGM ?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 90256,
      "title": "How to tune or decide the network?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 90274,
      "title": "Looking for a Team",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 90089,
      "title": "Loss And Metrics",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 90105,
      "title": "kernel problem",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 90180,
      "title": "GPU mode can not load input",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 90198,
      "title": "Keep my folder architecture in Kaggle Kernel",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89765,
      "title": "Kaggle performance",
      "comment_count": 4,
      "views": 0,
      "votes": 13
    },
    {
      "id": 90127,
      "title": "Best package in R for string manipulation, sentiment analysis and NLP",
      "comment_count": 2,
      "views": 0,
      "votes": 0
    },
    {
      "id": 89693,
      "title": "Embedding Penalty",
      "comment_count": 6,
      "views": 0,
      "votes": 7
    },
    {
      "id": 89964,
      "title": "GPU or CPU ?",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89863,
      "title": "Jigsaw EDA and Modelling kernel",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89836,
      "title": "Is offline training on competition data allowed?",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89869,
      "title": "little confuse about the auc",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 89982,
      "title": "Question about submission format.",
      "comment_count": 0,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89296,
      "title": "Anyone interested in working together on this challenge?",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 89898,
      "title": "Approach to saving and versions in kernels",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 89891,
      "title": "Encoding errors on LSTM kernels",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 89837,
      "title": "anybody can contribute a CV model?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 89696,
      "title": "Is this really toxic?",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89759,
      "title": "Could anyone explain a little bit about the \"identity columns\"",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 88691,
      "title": "Looking for team up!",
      "comment_count": 16,
      "views": 0,
      "votes": 1
    },
    {
      "id": 89587,
      "title": "Can we save the after-training result?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 89522,
      "title": "how is target value calculated?",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 89616,
      "title": "Clarification On Metrics",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87537,
      "title": "word embedding or should we make custom word embedding for this competition",
      "comment_count": 4,
      "views": 0,
      "votes": 2
    },
    {
      "id": 89221,
      "title": "A question about using multiple kernel .",
      "comment_count": 4,
      "views": 0,
      "votes": 3
    },
    {
      "id": 89482,
      "title": "Validation?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 89437,
      "title": "Some insights from Quora Insincere Questions Classification",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 87648,
      "title": "Useful references from past competitions",
      "comment_count": 4,
      "views": 0,
      "votes": 46
    },
    {
      "id": 89256,
      "title": "Eligible statistical software (Python, R, SAS, Stata..)",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 88420,
      "title": "Missing characters/words or error with pandas reading csv ?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 88404,
      "title": "About text cleaning and capsule networks",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 88319,
      "title": "Different data in previous competition",
      "comment_count": 7,
      "views": 0,
      "votes": 7
    },
    {
      "id": 87245,
      "title": "Overview of past toxic competition and framework code",
      "comment_count": 12,
      "views": 0,
      "votes": 100
    },
    {
      "id": 88765,
      "title": "How to use Pytorch for text tokenization",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 88415,
      "title": "How many data will be in stage-2 for private lb？",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87758,
      "title": "Sorry guys, I have a BTC mining farm :D",
      "comment_count": 8,
      "views": 0,
      "votes": -14
    },
    {
      "id": 87950,
      "title": "How are you going to utilize the extra columns in train data",
      "comment_count": 4,
      "views": 0,
      "votes": 10
    },
    {
      "id": 88191,
      "title": "Monitoring how much resources your kernel uses",
      "comment_count": 3,
      "views": 0,
      "votes": 9
    },
    {
      "id": 88300,
      "title": "Anyone tried FastAI AWD-LSTM?",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 88218,
      "title": "Codec can't decode",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 88336,
      "title": "Cannot submit predictions. Internet isssues",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 88021,
      "title": "What does it mean for No custom packages enabled in kernels?",
      "comment_count": 2,
      "views": 0,
      "votes": 4
    },
    {
      "id": 88313,
      "title": "How are people using the other columns",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 88229,
      "title": "AllenNLP for Kaggle and a starter kit",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 88200,
      "title": "Small subgroups in test may cause large score fluctutations ",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 87959,
      "title": "Clever competition design - \"inference kernel\" type requires new strategies",
      "comment_count": 0,
      "views": 0,
      "votes": 9
    },
    {
      "id": 88150,
      "title": "cleaned data",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 87269,
      "title": "BERT - key component for this competition?",
      "comment_count": 10,
      "views": 0,
      "votes": 12
    },
    {
      "id": 87991,
      "title": "where is the gold medal",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 87778,
      "title": "Some helpful links...",
      "comment_count": 1,
      "views": 0,
      "votes": 4
    },
    {
      "id": 87578,
      "title": "Deep single model on one fold or simpler models on k-fold?",
      "comment_count": 4,
      "views": 0,
      "votes": 4
    },
    {
      "id": 87732,
      "title": "Old version of Modin means slow Pandas (can we have it updated)",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 87683,
      "title": "Does extract features from raw data and save as a private dataset ok？",
      "comment_count": 5,
      "views": 0,
      "votes": null
    },
    {
      "id": 87368,
      "title": "interesting identity behavior",
      "comment_count": 0,
      "views": 0,
      "votes": 17
    },
    {
      "id": 87613,
      "title": "how to understand the other labels except “target”",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 87545,
      "title": "Padding: ex-ante or ex-post?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 87453,
      "title": "i`m newbie. i have a question!",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 87410,
      "title": "Kernel running for very long time",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 87357,
      "title": "Can I use the data from this competition for academic project?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 87420,
      "title": "Other CPU models good for word vectors?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 87387,
      "title": "Loss functions",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 87364,
      "title": "How can i download the embed files",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 87271,
      "title": "Public and Private Leaderboard will use the same data?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    }
  ],
  "errors": []
}