{
  "id": 664985,
  "title": "Some more data",
  "url": "/competitions/deep-past-initiative-machine-translation/discussion/664985",
  "author_name": "",
  "post_date": "2025-12-29T12:19:48.189032900Z",
  "votes": 5,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi everyone, \nI made some <a href=\"https://www.kaggle.com/datasets/aneeshmukkamala/cleandata\" target=\"_blank\">new data</a>. </p>\n<p>I put out a few samples (only 80) to gather your thoughts, comments, reviews. \nLet me know if these aid in your training or if they seem useful.\nI will generate even more. </p>",
  "messages": [
    {
      "id": "3383109",
      "postDate": "12/29/2025 12:19:48",
      "content": "<p>Hi everyone, \nI made some <a href=\"https://www.kaggle.com/datasets/aneeshmukkamala/cleandata\" target=\"_blank\">new data</a>. </p>\n<p>I put out a few samples (only 80) to gather your thoughts, comments, reviews. \nLet me know if these aid in your training or if they seem useful.\nI will generate even more. </p>",
      "rawMarkdown": "Hi everyone, \nI made some [new data](https://www.kaggle.com/datasets/aneeshmukkamala/cleandata). \n\nI put out a few samples (only 80) to gather your thoughts, comments, reviews. \nLet me know if these aid in your training or if they seem useful.\nI will generate even more.",
      "votes": null
    },
    {
      "id": "3383112",
      "postDate": "12/29/2025 12:37:19",
      "content": "<p>It's unaccessible <a href=\"https://www.kaggle.com/aneeshmukkamala\" target=\"_blank\">@aneeshmukkamala</a> . :)</p>",
      "rawMarkdown": "It's unaccessible @aneeshmukkamala . :)",
      "votes": null
    },
    {
      "id": "3383116",
      "postDate": "12/29/2025 12:45:41",
      "content": "<p>my bad, it was private</p>\n<p>i made it public</p>",
      "rawMarkdown": "my bad, it was private\n\ni made it public",
      "votes": null
    },
    {
      "id": "3383798",
      "postDate": "12/30/2025 21:38:53",
      "content": "<p>You will generate? What does that mean? Where are you getting this data?</p>",
      "rawMarkdown": "You will generate? What does that mean? Where are you getting this data?",
      "votes": null
    },
    {
      "id": "3383859",
      "postDate": "12/31/2025 00:32:59",
      "content": "<p>I pre trained a multi modal base model on Akkadian textbooks, rulebooks, literature. Getting it to work neatly for translation was really difficult due to lack of pair wise data for SFT and also is computationaly not usable on T4x2s of Kaggle. So I'm using it as a RAG kinda of thing and then pass that info to another LLM for data generation. There's a lot more going on than this but this is the high level gist. </p>",
      "rawMarkdown": "I pre trained a multi modal base model on Akkadian textbooks, rulebooks, literature. Getting it to work neatly for translation was really difficult due to lack of pair wise data for SFT and also is computationaly not usable on T4x2s of Kaggle. So I'm using it as a RAG kinda of thing and then pass that info to another LLM for data generation. There's a lot more going on than this but this is the high level gist.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3383112,
      "author_name": "angantyr",
      "author_url": "",
      "post_date": "12/29/2025 12:37:19",
      "content": "<p>It's unaccessible <a href=\"https://www.kaggle.com/aneeshmukkamala\" target=\"_blank\">@aneeshmukkamala</a> . :)</p>",
      "votes": null,
      "replies": [
        {
          "id": 3383116,
          "author_name": "",
          "author_url": "",
          "post_date": "12/29/2025 12:45:41",
          "content": "<p>my bad, it was private</p>\n<p>i made it public</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3383798,
      "author_name": "alvarochaveste",
      "author_url": "",
      "post_date": "12/30/2025 21:38:53",
      "content": "<p>You will generate? What does that mean? Where are you getting this data?</p>",
      "votes": null,
      "replies": [
        {
          "id": 3383859,
          "author_name": "",
          "author_url": "",
          "post_date": "12/31/2025 00:32:59",
          "content": "<p>I pre trained a multi modal base model on Akkadian textbooks, rulebooks, literature. Getting it to work neatly for translation was really difficult due to lack of pair wise data for SFT and also is computationaly not usable on T4x2s of Kaggle. So I'm using it as a RAG kinda of thing and then pass that info to another LLM for data generation. There's a lot more going on than this but this is the high level gist. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3383109": "Hi everyone, \nI made some [new data](https://www.kaggle.com/datasets/aneeshmukkamala/cleandata). \n\nI put out a few samples (only 80) to gather your thoughts, comments, reviews. \nLet me know if these aid in your training or if they seem useful.\nI will generate even more.",
    "3383112": "It's unaccessible @aneeshmukkamala . :)",
    "3383116": "my bad, it was private\n\ni made it public",
    "3383798": "You will generate? What does that mean? Where are you getting this data?",
    "3383859": "I pre trained a multi modal base model on Akkadian textbooks, rulebooks, literature. Getting it to work neatly for translation was really difficult due to lack of pair wise data for SFT and also is computationaly not usable on T4x2s of Kaggle. So I'm using it as a RAG kinda of thing and then pass that info to another LLM for data generation. There's a lot more going on than this but this is the high level gist."
  },
  "source": "meta"
}