{
  "id": 128457,
  "title": "A list of submission errors and how to resolve",
  "url": "/competitions/bengaliai-cv19/discussion/128457",
  "author_name": "Appian",
  "post_date": "2020-01-31T13:22:17.213000",
  "votes": 24,
  "comment_count": 19,
  "views": 0,
  "content": "<p>I found this useful to narrow down and solve mysterious submission errors and thought this should be shared here.</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/733730/14840/kaggle.png\" alt=\"\"></p>\n\n<p>Source: <a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started\">https://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started</a></p>",
  "messages": [
    {
      "id": 733730,
      "postDate": "2020-01-31T13:22:17.213Z",
      "content": "<p>I found this useful to narrow down and solve mysterious submission errors and thought this should be shared here.</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/733730/14840/kaggle.png\" alt=\"\"></p>\n\n<p>Source: <a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started\">https://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started</a></p>",
      "rawMarkdown": "I found this useful to narrow down and solve mysterious submission errors and thought this should be shared here.\n\n![](https://storage.googleapis.com/kaggle-forum-message-attachments/733730/14840/kaggle.png)\n\nSource: https://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started\n\n",
      "votes": 24
    },
    {
      "id": 745784,
      "postDate": "2020-02-14T07:43:53.357Z",
      "content": "<p>(<em>Notebook Exceeded Allowed Compute</em>)When the input size of your model (like densenet) is <strong>128 or bigger</strong>, one-time prediction of a file may cause memory problem. Please try to predict <strong>100</strong> pictures or <strong>1000</strong> pictures at a time.\nI came to this conclusion through multiple submissions. This is a sad story.</p>",
      "rawMarkdown": "(*Notebook Exceeded Allowed Compute*)When the input size of your model (like densenet) is **128 or bigger**, one-time prediction of a file may cause memory problem. Please try to predict **100** pictures or **1000** pictures at a time.\nI came to this conclusion through multiple submissions. This is a sad story.",
      "votes": 1
    },
    {
      "id": 733802,
      "postDate": "2020-01-31T14:27:23.903Z",
      "content": "<p>In my case I encountered <code>Notebook Exceeded Allowed Compute</code> yesterday and had to spend several submissions to solve.</p>\n\n<p>I had no idea what <code>Notebook Exceeded Allowed Compute</code> specifically means. I did not make fishy changes from the last successful submission. Finally I found this list and found out there is <code>External data size limit</code> in other competition. It is not mentioned in this competition and not sure this was the problem. Anyway I excluded past unused uploaded weights(dataset) from notebook and made a submission by rerunning which turned out to be successful.</p>\n\n<p>Actually I'm still not sure <code>External data size limit</code> was the problem because all I got was <code>Notebook Exceeded Allowed Compute</code> after all and this error was not very consistent. But excluding some uploaded dataset solved the issue for now and wanted to share in case someone meets the same problem.</p>\n\n<p>UPDATE:\nI got this weired error again regardless of external dataset. The size of external dataset might not be the problem. It just seems so random. Reforking kernels could work.</p>",
      "rawMarkdown": "In my case I encountered `Notebook Exceeded Allowed Compute` yesterday and had to spend several submissions to solve.\n\nI had no idea what `Notebook Exceeded Allowed Compute` specifically means. I did not make fishy changes from the last successful submission. Finally I found this list and found out there is `External data size limit` in other competition. It is not mentioned in this competition and not sure this was the problem. Anyway I excluded past unused uploaded weights(dataset) from notebook and made a submission by rerunning which turned out to be successful.\n\nActually I'm still not sure `External data size limit` was the problem because all I got was `Notebook Exceeded Allowed Compute` after all and this error was not very consistent. But excluding some uploaded dataset solved the issue for now and wanted to share in case someone meets the same problem.\n\nUPDATE:\nI got this weired error again regardless of external dataset. The size of external dataset might not be the problem. It just seems so random. Reforking kernels could work.",
      "votes": 1,
      "replies": [
        {
          "id": 733810,
          "postDate": "2020-01-31T14:42:20.640Z",
          "content": "<p>How large was your external data when it wasn't working?</p>",
          "rawMarkdown": "How large was your external data when it wasn't working?"
        },
        {
          "id": 733813,
          "postDate": "2020-01-31T14:46:24.960Z",
          "content": "<p>It was more than a few GB at total without this competition's data when it was not working. You meet the same error? </p>",
          "rawMarkdown": "It was more than a few GB at total without this competition's data when it was not working. You meet the same error? "
        },
        {
          "id": 733815,
          "postDate": "2020-01-31T14:49:10.980Z",
          "content": "<p>No, just wanted to know for future reference and so I don't run into it myself.</p>",
          "rawMarkdown": "No, just wanted to know for future reference and so I don't run into it myself.",
          "votes": 1
        },
        {
          "id": 733825,
          "postDate": "2020-01-31T14:56:37.987Z",
          "content": "<p>Okay. Hope this helps when you meet one.</p>",
          "rawMarkdown": "Okay. Hope this helps when you meet one."
        },
        {
          "id": 733848,
          "postDate": "2020-01-31T15:21:41.377Z",
          "content": "<p>I'm getting the same Notebook Exceeded Allowed Compute error message. I submitted a ticket to Kaggle.</p>",
          "rawMarkdown": "I'm getting the same Notebook Exceeded Allowed Compute error message. I submitted a ticket to Kaggle.",
          "votes": 1
        },
        {
          "id": 737020,
          "postDate": "2020-02-04T20:20:54.517Z",
          "content": "<p>I also got a <code>Notebook Exceeded Allowed Compute</code> today, without any external data. I just load my cnn model(300MB) and make a predict in my kernel. Is there a solution to this problem?</p>",
          "rawMarkdown": "I also got a `Notebook Exceeded Allowed Compute` today, without any external data. I just load my cnn model(300MB) and make a predict in my kernel. Is there a solution to this problem?"
        },
        {
          "id": 737084,
          "postDate": "2020-02-04T22:36:31.280Z",
          "content": "<p><a href=\"/xiaohuhayou\">@xiaohuhayou</a> \nI found that the size of external dataset might not be the case after facing another error. It might be something to do with some hidden environments we have no idea. You can try reforking/recreating kernels in that case.</p>",
          "rawMarkdown": "@xiaohuhayou \nI found that the size of external dataset might not be the case after facing another error. It might be something to do with some hidden environments we have no idea. You can try reforking/recreating kernels in that case."
        },
        {
          "id": 737372,
          "postDate": "2020-02-05T08:58:59.323Z",
          "content": "<p>Reforking doesnot work for me...\nI uploaded a simple cnn model (about 100MB, the last is 300MB) to predict, and no errors happend. 😂 😂 </p>",
          "rawMarkdown": "Reforking doesnot work for me...\nI uploaded a simple cnn model (about 100MB, the last is 300MB) to predict, and no errors happend. 😂 😂 \n"
        },
        {
          "id": 744595,
          "postDate": "2020-02-13T00:44:07.477Z",
          "content": "<p>&gt; I'm getting the same Notebook Exceeded Allowed Compute error message. I submitted a ticket to Kaggle.</p>\n\n<p>UPDATE:  Kaggle got back to me and said since it wasn't an issue everyone was experiencing, that they were not going to help.</p>\n\n<p>I created a public notebook on something I found, where reshaping one test set in memory was causing the error for me.  The notebook just shows that I can load each test set into memory and then submit fake prediction data successfully.  Maybe this helps others.  Someone suggested that I run prediction on each row 1 by 1.  So I might try that out or I might look at some working code from other public notebooks, but I think I've found a starting point at least (for me).\n<a href=\"https://www.kaggle.com/yeayates21/cant-submit-1s-whats-wrong-please-help\">https://www.kaggle.com/yeayates21/cant-submit-1s-whats-wrong-please-help</a></p>",
          "rawMarkdown": "&gt; I'm getting the same Notebook Exceeded Allowed Compute error message. I submitted a ticket to Kaggle.\n\nUPDATE:  Kaggle got back to me and said since it wasn't an issue everyone was experiencing, that they were not going to help.\n\nI created a public notebook on something I found, where reshaping one test set in memory was causing the error for me.  The notebook just shows that I can load each test set into memory and then submit fake prediction data successfully.  Maybe this helps others.  Someone suggested that I run prediction on each row 1 by 1.  So I might try that out or I might look at some working code from other public notebooks, but I think I've found a starting point at least (for me).\nhttps://www.kaggle.com/yeayates21/cant-submit-1s-whats-wrong-please-help"
        },
        {
          "id": 745115,
          "postDate": "2020-02-13T14:08:18.163Z",
          "content": "<p>Thanks for the update <a href=\"/yeayates21\">@yeayates21</a> .\nThis is the part I use to load parquet files in case you are interested.\nThis works just fine. </p>\n\n<p><code>\ndef load_parquet(path):\n    df = pd.read_parquet(path)\n    log('done loading %d rows from %s' % (len(df), path))\n    images = df.iloc[:,1:].values.reshape(-1, 137, 236).astype(np.uint8)\n    image_ids = df.iloc[:,0]\n    return images, image_ids\n</code></p>\n\n<p>It's hard to give an exact answer to this problem but my suggestion could be</p>\n\n<ul>\n<li>Reforking the kernel.</li>\n<li>Try predicting on train_image_data instead of test_image_data to make sure your code runs fine. You can do this without submitting and can see actual error messages.</li>\n</ul>",
          "rawMarkdown": "Thanks for the update @yeayates21 .\nThis is the part I use to load parquet files in case you are interested.\nThis works just fine. \n\n```\ndef load_parquet(path):\n    df = pd.read_parquet(path)\n    log('done loading %d rows from %s' % (len(df), path))\n    images = df.iloc[:,1:].values.reshape(-1, 137, 236).astype(np.uint8)\n    image_ids = df.iloc[:,0]\n    return images, image_ids\n```\n\nIt's hard to give an exact answer to this problem but my suggestion could be\n\n- Reforking the kernel.\n- Try predicting on train\\_image\\_data instead of test\\_image\\_data to make sure your code runs fine. You can do this without submitting and can see actual error messages.\n",
          "votes": 2
        },
        {
          "id": 764831,
          "postDate": "2020-03-06T00:47:15.770Z",
          "content": "<p>News on this? I am encountering this error now. My data is a bit large, but i can easily reduce it (I have attached to the inference notebook things that I am not currently using).</p>\n\n<p>The messages could be more specific to make our life easier.</p>\n\n<p>Thank you</p>",
          "rawMarkdown": "News on this? I am encountering this error now. My data is a bit large, but i can easily reduce it (I have attached to the inference notebook things that I am not currently using).\n\nThe messages could be more specific to make our life easier.\n\nThank you"
        },
        {
          "id": 771814,
          "postDate": "2020-03-14T16:53:59.910Z",
          "content": "<p>My issue was that I was that I was trying to load 3 modeling into memory 1 at a time instead of just building 1 model :-/</p>\n\n<p>Once I switched to just building 1 model, everything worked.</p>",
          "rawMarkdown": "My issue was that I was that I was trying to load 3 modeling into memory 1 at a time instead of just building 1 model :-/\n\nOnce I switched to just building 1 model, everything worked."
        }
      ]
    },
    {
      "id": 733951,
      "postDate": "2020-01-31T18:08:44.523Z",
      "content": "<p>In my case I got <code>Submission Scoring Error</code> when I was generating 3 channels data in my data loader. It definitely not a <code>submission.csv</code> format problem, because it can run successfully with 12 test data. \nI followed this tips <a href=\"https://www.kaggle.com/c/bengaliai-cv19/discussion/126054\">https://www.kaggle.com/c/bengaliai-cv19/discussion/126054</a>  , to change 3 channels to 1 channel data, then problem fixed. It should be Memory problem.  It was very confusing to have a <code>Submission Scoring Error</code> for a Memory problem.</p>",
      "rawMarkdown": "In my case I got `Submission Scoring Error` when I was generating 3 channels data in my data loader. It definitely not a `submission.csv` format problem, because it can run successfully with 12 test data. \nI followed this tips https://www.kaggle.com/c/bengaliai-cv19/discussion/126054  , to change 3 channels to 1 channel data, then problem fixed. It should be Memory problem.  It was very confusing to have a `Submission Scoring Error` for a Memory problem.",
      "votes": 2,
      "replies": [
        {
          "id": 734185,
          "postDate": "2020-02-01T04:14:54.713Z",
          "content": "<p>Hi chicm,\nThank you for sharing and very confusing error message indeed.</p>",
          "rawMarkdown": "Hi chicm,\nThank you for sharing and very confusing error message indeed."
        }
      ]
    },
    {
      "id": 748141,
      "postDate": "2020-02-17T07:23:03.537Z",
      "content": "<p>I faced the <code>Submission Scoring Error</code> because I sent the prediction of just one instead of all four Test.parquet file. I also had the wrong sort order in the row_id column, though I'm not sure if you need the right order in the submission file.</p>",
      "rawMarkdown": "I faced the `Submission Scoring Error` because I sent the prediction of just one instead of all four Test.parquet file. I also had the wrong sort order in the row_id column, though I'm not sure if you need the right order in the submission file."
    },
    {
      "id": 772125,
      "postDate": "2020-03-15T04:02:37.740Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 733847,
      "postDate": "2020-01-31T15:20:54.957Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 745784,
      "author_name": "CXDZB",
      "author_url": "",
      "post_date": "2020-02-14T07:43:53.357000",
      "content": "<p>(<em>Notebook Exceeded Allowed Compute</em>)When the input size of your model (like densenet) is <strong>128 or bigger</strong>, one-time prediction of a file may cause memory problem. Please try to predict <strong>100</strong> pictures or <strong>1000</strong> pictures at a time.\nI came to this conclusion through multiple submissions. This is a sad story.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 733802,
      "author_name": "Appian",
      "author_url": "",
      "post_date": "2020-01-31T14:27:23.903000",
      "content": "<p>In my case I encountered <code>Notebook Exceeded Allowed Compute</code> yesterday and had to spend several submissions to solve.</p>\n\n<p>I had no idea what <code>Notebook Exceeded Allowed Compute</code> specifically means. I did not make fishy changes from the last successful submission. Finally I found this list and found out there is <code>External data size limit</code> in other competition. It is not mentioned in this competition and not sure this was the problem. Anyway I excluded past unused uploaded weights(dataset) from notebook and made a submission by rerunning which turned out to be successful.</p>\n\n<p>Actually I'm still not sure <code>External data size limit</code> was the problem because all I got was <code>Notebook Exceeded Allowed Compute</code> after all and this error was not very consistent. But excluding some uploaded dataset solved the issue for now and wanted to share in case someone meets the same problem.</p>\n\n<p>UPDATE:\nI got this weired error again regardless of external dataset. The size of external dataset might not be the problem. It just seems so random. Reforking kernels could work.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 733810,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2020-01-31T14:42:20.640000",
          "content": "<p>How large was your external data when it wasn't working?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 733813,
          "author_name": "Appian",
          "author_url": "",
          "post_date": "2020-01-31T14:46:24.960000",
          "content": "<p>It was more than a few GB at total without this competition's data when it was not working. You meet the same error? </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 733815,
          "author_name": "GreatGameDota",
          "author_url": "",
          "post_date": "2020-01-31T14:49:10.980000",
          "content": "<p>No, just wanted to know for future reference and so I don't run into it myself.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 733825,
          "author_name": "Appian",
          "author_url": "",
          "post_date": "2020-01-31T14:56:37.987000",
          "content": "<p>Okay. Hope this helps when you meet one.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 733848,
          "author_name": "Matt Yates",
          "author_url": "",
          "post_date": "2020-01-31T15:21:41.377000",
          "content": "<p>I'm getting the same Notebook Exceeded Allowed Compute error message. I submitted a ticket to Kaggle.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 737020,
          "author_name": "Orion",
          "author_url": "",
          "post_date": "2020-02-04T20:20:54.517000",
          "content": "<p>I also got a <code>Notebook Exceeded Allowed Compute</code> today, without any external data. I just load my cnn model(300MB) and make a predict in my kernel. Is there a solution to this problem?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 737084,
          "author_name": "Appian",
          "author_url": "",
          "post_date": "2020-02-04T22:36:31.280000",
          "content": "<p><a href=\"/xiaohuhayou\">@xiaohuhayou</a> \nI found that the size of external dataset might not be the case after facing another error. It might be something to do with some hidden environments we have no idea. You can try reforking/recreating kernels in that case.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 737372,
          "author_name": "Orion",
          "author_url": "",
          "post_date": "2020-02-05T08:58:59.323000",
          "content": "<p>Reforking doesnot work for me...\nI uploaded a simple cnn model (about 100MB, the last is 300MB) to predict, and no errors happend. 😂 😂 </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 744595,
          "author_name": "Matt Yates",
          "author_url": "",
          "post_date": "2020-02-13T00:44:07.477000",
          "content": "<p>&gt; I'm getting the same Notebook Exceeded Allowed Compute error message. I submitted a ticket to Kaggle.</p>\n\n<p>UPDATE:  Kaggle got back to me and said since it wasn't an issue everyone was experiencing, that they were not going to help.</p>\n\n<p>I created a public notebook on something I found, where reshaping one test set in memory was causing the error for me.  The notebook just shows that I can load each test set into memory and then submit fake prediction data successfully.  Maybe this helps others.  Someone suggested that I run prediction on each row 1 by 1.  So I might try that out or I might look at some working code from other public notebooks, but I think I've found a starting point at least (for me).\n<a href=\"https://www.kaggle.com/yeayates21/cant-submit-1s-whats-wrong-please-help\">https://www.kaggle.com/yeayates21/cant-submit-1s-whats-wrong-please-help</a></p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 745115,
          "author_name": "Appian",
          "author_url": "",
          "post_date": "2020-02-13T14:08:18.163000",
          "content": "<p>Thanks for the update <a href=\"/yeayates21\">@yeayates21</a> .\nThis is the part I use to load parquet files in case you are interested.\nThis works just fine. </p>\n\n<p><code>\ndef load_parquet(path):\n    df = pd.read_parquet(path)\n    log('done loading %d rows from %s' % (len(df), path))\n    images = df.iloc[:,1:].values.reshape(-1, 137, 236).astype(np.uint8)\n    image_ids = df.iloc[:,0]\n    return images, image_ids\n</code></p>\n\n<p>It's hard to give an exact answer to this problem but my suggestion could be</p>\n\n<ul>\n<li>Reforking the kernel.</li>\n<li>Try predicting on train_image_data instead of test_image_data to make sure your code runs fine. You can do this without submitting and can see actual error messages.</li>\n</ul>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 764831,
          "author_name": "Vasco Mano",
          "author_url": "",
          "post_date": "2020-03-06T00:47:15.770000",
          "content": "<p>News on this? I am encountering this error now. My data is a bit large, but i can easily reduce it (I have attached to the inference notebook things that I am not currently using).</p>\n\n<p>The messages could be more specific to make our life easier.</p>\n\n<p>Thank you</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 771814,
          "author_name": "Matt Yates",
          "author_url": "",
          "post_date": "2020-03-14T16:53:59.910000",
          "content": "<p>My issue was that I was that I was trying to load 3 modeling into memory 1 at a time instead of just building 1 model :-/</p>\n\n<p>Once I switched to just building 1 model, everything worked.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 733951,
      "author_name": "chicm",
      "author_url": "",
      "post_date": "2020-01-31T18:08:44.523000",
      "content": "<p>In my case I got <code>Submission Scoring Error</code> when I was generating 3 channels data in my data loader. It definitely not a <code>submission.csv</code> format problem, because it can run successfully with 12 test data. \nI followed this tips <a href=\"https://www.kaggle.com/c/bengaliai-cv19/discussion/126054\">https://www.kaggle.com/c/bengaliai-cv19/discussion/126054</a>  , to change 3 channels to 1 channel data, then problem fixed. It should be Memory problem.  It was very confusing to have a <code>Submission Scoring Error</code> for a Memory problem.</p>",
      "votes": 2,
      "replies": [
        {
          "id": 734185,
          "author_name": "Appian",
          "author_url": "",
          "post_date": "2020-02-01T04:14:54.713000",
          "content": "<p>Hi chicm,\nThank you for sharing and very confusing error message indeed.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 748141,
      "author_name": "Jo Tom",
      "author_url": "",
      "post_date": "2020-02-17T07:23:03.537000",
      "content": "<p>I faced the <code>Submission Scoring Error</code> because I sent the prediction of just one instead of all four Test.parquet file. I also had the wrong sort order in the row_id column, though I'm not sure if you need the right order in the submission file.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 772125,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-03-15T04:02:37.740000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 733847,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-01-31T15:20:54.957000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "733730": "I found this useful to narrow down and solve mysterious submission errors and thought this should be shared here.\n\n![](https://storage.googleapis.com/kaggle-forum-message-attachments/733730/14840/kaggle.png)\n\nSource: https://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started\n\n",
    "745784": "(*Notebook Exceeded Allowed Compute*)When the input size of your model (like densenet) is **128 or bigger**, one-time prediction of a file may cause memory problem. Please try to predict **100** pictures or **1000** pictures at a time.\nI came to this conclusion through multiple submissions. This is a sad story.",
    "733802": "In my case I encountered `Notebook Exceeded Allowed Compute` yesterday and had to spend several submissions to solve.\n\nI had no idea what `Notebook Exceeded Allowed Compute` specifically means. I did not make fishy changes from the last successful submission. Finally I found this list and found out there is `External data size limit` in other competition. It is not mentioned in this competition and not sure this was the problem. Anyway I excluded past unused uploaded weights(dataset) from notebook and made a submission by rerunning which turned out to be successful.\n\nActually I'm still not sure `External data size limit` was the problem because all I got was `Notebook Exceeded Allowed Compute` after all and this error was not very consistent. But excluding some uploaded dataset solved the issue for now and wanted to share in case someone meets the same problem.\n\nUPDATE:\nI got this weired error again regardless of external dataset. The size of external dataset might not be the problem. It just seems so random. Reforking kernels could work.",
    "733951": "In my case I got `Submission Scoring Error` when I was generating 3 channels data in my data loader. It definitely not a `submission.csv` format problem, because it can run successfully with 12 test data. \nI followed this tips https://www.kaggle.com/c/bengaliai-cv19/discussion/126054  , to change 3 channels to 1 channel data, then problem fixed. It should be Memory problem.  It was very confusing to have a `Submission Scoring Error` for a Memory problem.",
    "748141": "I faced the `Submission Scoring Error` because I sent the prediction of just one instead of all four Test.parquet file. I also had the wrong sort order in the row_id column, though I'm not sure if you need the right order in the submission file.",
    "772125": "",
    "733847": ""
  }
}