{
  "id": 438484,
  "title": "Finetune training pipline",
  "url": "/competitions/bengaliai-speech/discussion/438484",
  "author_name": "",
  "post_date": "2023-09-11T12:21:56.403950400Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi, everyone. I have published a training and inference pipeline to share with you.<br>\nThe training notebook: <a href=\"https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-training\" target=\"_blank\">Bengali SR wav2vec_v1_bengali [Training]</a><br>\nThe inference notebook: <a href=\"https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-inference\" target=\"_blank\">Bengali SR wav2vec_v1_bengali [Inference]</a><br>\nMy first version scored 0.445 on the leaderboard. You can use the training notebook to finetune a better model and improve your score.<br>\nThanks.</p>",
  "messages": [
    {
      "id": "2433225",
      "postDate": "09/11/2023 12:21:56",
      "content": "<p>Hi, everyone. I have published a training and inference pipeline to share with you.<br>\nThe training notebook: <a href=\"https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-training\" target=\"_blank\">Bengali SR wav2vec_v1_bengali [Training]</a><br>\nThe inference notebook: <a href=\"https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-inference\" target=\"_blank\">Bengali SR wav2vec_v1_bengali [Inference]</a><br>\nMy first version scored 0.445 on the leaderboard. You can use the training notebook to finetune a better model and improve your score.<br>\nThanks.</p>",
      "rawMarkdown": "Hi, everyone. I have published a training and inference pipeline to share with you.\nThe training notebook: [Bengali SR wav2vec_v1_bengali [Training]](https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-training)\nThe inference notebook: [Bengali SR wav2vec_v1_bengali [Inference]](https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-inference)\nMy first version scored 0.445 on the leaderboard. You can use the training notebook to finetune a better model and improve your score.\nThanks.",
      "votes": null
    },
    {
      "id": "2434460",
      "postDate": "09/12/2023 10:06:44",
      "content": "<p>how long do you train on your first version, i find that it takes long time to train😭</p>",
      "rawMarkdown": "how long do you train on your first version, i find that it takes long time to train😭",
      "votes": null
    },
    {
      "id": "2434503",
      "postDate": "09/12/2023 10:40:33",
      "content": "<p>Around 5 hours with A100 40G.</p>",
      "rawMarkdown": "Around 5 hours with A100 40G.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2434460,
      "author_name": "algerwang",
      "author_url": "",
      "post_date": "09/12/2023 10:06:44",
      "content": "<p>how long do you train on your first version, i find that it takes long time to train😭</p>",
      "votes": null,
      "replies": [
        {
          "id": 2434503,
          "author_name": "takanashihumbert",
          "author_url": "",
          "post_date": "09/12/2023 10:40:33",
          "content": "<p>Around 5 hours with A100 40G.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2433225": "Hi, everyone. I have published a training and inference pipeline to share with you.\nThe training notebook: [Bengali SR wav2vec_v1_bengali [Training]](https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-training)\nThe inference notebook: [Bengali SR wav2vec_v1_bengali [Inference]](https://www.kaggle.com/code/takanashihumbert/bengali-sr-wav2vec-v1-bengali-inference)\nMy first version scored 0.445 on the leaderboard. You can use the training notebook to finetune a better model and improve your score.\nThanks.",
    "2434460": "how long do you train on your first version, i find that it takes long time to train😭",
    "2434503": "Around 5 hours with A100 40G."
  },
  "source": "meta"
}