I trained a keras model on the training data, and from the output (a model.h5 and a history.csv files) I created a private dataset. Then, in a separate (inference) kernel, I predict the test set using the model I loaded from my private dataset. The model is clearly under 1GB.
\n\nHowever, when i try to submit submission.csv from my inference, it keeps telling me: \"Your notebook cannot use non-standard datasets in this competition\".
\n\nAny idea on how I can submit anything?
\n\nEDIT: solved by manually uploading model.h5 instead of generating from kernel output.
", "messages": [ { "id": "695184", "postDate": "12/14/2019 18:06:17", "content": "I trained a keras model on the training data, and from the output (a model.h5 and a history.csv files) I created a private dataset. Then, in a separate (inference) kernel, I predict the test set using the model I loaded from my private dataset. The model is clearly under 1GB.
\n\nHowever, when i try to submit submission.csv from my inference, it keeps telling me: \"Your notebook cannot use non-standard datasets in this competition\".
\n\nAny idea on how I can submit anything?
\n\nEDIT: solved by manually uploading model.h5 instead of generating from kernel output.
", "rawMarkdown": "I trained a keras model on the training data, and from the output (a model.h5 and a history.csv files) I created a private dataset. Then, in a separate (inference) kernel, I predict the test set using the model I loaded from my private dataset. The model is clearly under 1GB.\n\nHowever, when i try to submit submission.csv from my inference, it keeps telling me: \"Your notebook cannot use non-standard datasets in this competition\". \n\nAny idea on how I can submit anything?\n\nEDIT: solved by manually uploading model.h5 instead of generating from kernel output.", "votes": null }, { "id": "695190", "postDate": "12/14/2019 18:14:05", "content": "did you attach output of the training kernel? if yes, thats not allowed.
", "rawMarkdown": "did you attach output of the training kernel? if yes, thats not allowed.", "votes": null }, { "id": "695191", "postDate": "12/14/2019 18:14:23", "content": "Per the \"Code Requirements\" section:
\n\nNo using other Kaggle notebooks or utility scripts as inputs to your submission notebook. External data also cannot be externally-referenced (i.e. via BigQuery, Github, external URL). Instead, load external models or datasets directly as an external data source.\n
You'll need to upload the h5 and csv files as a dataset, instead of using the output from a previous kernel.
Try using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.
", "rawMarkdown": "Per the \"Code Requirements\" section:\n\n```No using other Kaggle notebooks or utility scripts as inputs to your submission notebook. External data also cannot be externally-referenced (i.e. via BigQuery, Github, external URL). Instead, load external models or datasets directly as an external data source.\n```\n\nYou'll need to upload the `h5` and `csv` files as a dataset, instead of using the output from a previous kernel.\n\nTry using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.", "votes": null }, { "id": "695192", "postDate": "12/14/2019 18:16:16", "content": "No I created a new dataset from the kernel, and loaded that dataset into my inference kernel. I didn't directly load the kernel output.
", "rawMarkdown": "No I created a new dataset from the kernel, and loaded that dataset into my inference kernel. I didn't directly load the kernel output.", "votes": null }, { "id": "695193", "postDate": "12/14/2019 18:16:45", "content": "\n\n\nTry using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.
\n
Exactly what i did lol
", "rawMarkdown": "> Try using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.\n\nExactly what i did lol", "votes": null }, { "id": "695194", "postDate": "12/14/2019 18:20:16", "content": "Oh ¯_(ツ)_/¯ maybe kaggle knows that the new dataset was created from a kernel. Did you try downloading and re-uploading into a dataset not linked to the first kernel?
", "rawMarkdown": "Oh ¯\\_(ツ)_/¯ maybe kaggle knows that the new dataset was created from a kernel. Did you try downloading and re-uploading into a dataset not linked to the first kernel?", "votes": null }, { "id": "695195", "postDate": "12/14/2019 18:22:31", "content": "That's a good idea, i'm trying this right now haha
", "rawMarkdown": "That's a good idea, i'm trying this right now haha", "votes": null }, { "id": "695211", "postDate": "12/14/2019 19:09:13", "content": "This worked :)
", "rawMarkdown": "This worked :)", "votes": null }, { "id": "723027", "postDate": "01/19/2020 12:20:32", "content": "I had the same error when trying to create a new dataset with a file from the URL option. After downloading from this URL to my machine -> creating a new dataset -> actually uploading the file to the dataset - it worked.
", "rawMarkdown": "I had the same error when trying to create a new dataset with a file from the URL option. After downloading from this URL to my machine -> creating a new dataset -> actually uploading the file to the dataset - it worked.", "votes": null }, { "id": "949851", "postDate": "07/29/2020 02:41:39", "content": "I want to call keras's inception_v3 model (via ' from keras.applications import inception_v3 ') inside my notebook. Is this allowed? I can't do this while being offline though so I cant get the submit option up. this is as when i call keras. inception_v3.InceptionV3 it tries to load a h5 file online from the url ' ttps://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5 ' ??
\n\nIs there a way for me to use this locally via my call to keras.applications as
\n\n``TimeDistributed(inception_v3.InceptionV3(include_top=False,pooling='max',weights='imagenet'))(img_input)
\n\nin my code
\n\nPlease heeeelp Im stuck :(
", "rawMarkdown": "I want to call keras's inception_v3 model (via ' from keras.applications import inception_v3 ') inside my notebook. Is this allowed? I can't do this while being offline though so I cant get the submit option up. this is as when i call keras. inception_v3.InceptionV3 it tries to load a h5 file online from the url ' ttps://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5 ' ?? \n\nIs there a way for me to use this locally via my call to keras.applications as \n\n``TimeDistributed(inception_v3.InceptionV3(include_top=False,pooling='max',weights='imagenet'))(img_input)\n\nin my code\n\nPlease heeeelp Im stuck :(", "votes": null } ], "comments": [ { "id": 695190, "author_name": "abhishek", "author_url": "", "post_date": "12/14/2019 18:14:05", "content": "did you attach output of the training kernel? if yes, thats not allowed.
", "votes": null, "replies": [ { "id": 695192, "author_name": "xhlulu", "author_url": "", "post_date": "12/14/2019 18:16:16", "content": "No I created a new dataset from the kernel, and loaded that dataset into my inference kernel. I didn't directly load the kernel output.
", "votes": null, "replies": [] } ] }, { "id": 695191, "author_name": "robikscube", "author_url": "", "post_date": "12/14/2019 18:14:23", "content": "Per the \"Code Requirements\" section:
\n\nNo using other Kaggle notebooks or utility scripts as inputs to your submission notebook. External data also cannot be externally-referenced (i.e. via BigQuery, Github, external URL). Instead, load external models or datasets directly as an external data source.\n
You'll need to upload the h5 and csv files as a dataset, instead of using the output from a previous kernel.
Try using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.
", "votes": null, "replies": [ { "id": 695193, "author_name": "xhlulu", "author_url": "", "post_date": "12/14/2019 18:16:45", "content": "\n\n\nTry using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.
\n
Exactly what i did lol
", "votes": null, "replies": [] }, { "id": 695194, "author_name": "robikscube", "author_url": "", "post_date": "12/14/2019 18:20:16", "content": "Oh ¯_(ツ)_/¯ maybe kaggle knows that the new dataset was created from a kernel. Did you try downloading and re-uploading into a dataset not linked to the first kernel?
", "votes": null, "replies": [] }, { "id": 695195, "author_name": "xhlulu", "author_url": "", "post_date": "12/14/2019 18:22:31", "content": "That's a good idea, i'm trying this right now haha
", "votes": null, "replies": [] }, { "id": 695211, "author_name": "xhlulu", "author_url": "", "post_date": "12/14/2019 19:09:13", "content": "This worked :)
", "votes": null, "replies": [] } ] }, { "id": 723027, "author_name": "odedgolden", "author_url": "", "post_date": "01/19/2020 12:20:32", "content": "I had the same error when trying to create a new dataset with a file from the URL option. After downloading from this URL to my machine -> creating a new dataset -> actually uploading the file to the dataset - it worked.
", "votes": null, "replies": [] }, { "id": 949851, "author_name": "pollgorm1", "author_url": "", "post_date": "07/29/2020 02:41:39", "content": "I want to call keras's inception_v3 model (via ' from keras.applications import inception_v3 ') inside my notebook. Is this allowed? I can't do this while being offline though so I cant get the submit option up. this is as when i call keras. inception_v3.InceptionV3 it tries to load a h5 file online from the url ' ttps://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5 ' ??
\n\nIs there a way for me to use this locally via my call to keras.applications as
\n\n``TimeDistributed(inception_v3.InceptionV3(include_top=False,pooling='max',weights='imagenet'))(img_input)
\n\nin my code
\n\nPlease heeeelp Im stuck :(
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "695184": "I trained a keras model on the training data, and from the output (a model.h5 and a history.csv files) I created a private dataset. Then, in a separate (inference) kernel, I predict the test set using the model I loaded from my private dataset. The model is clearly under 1GB.\n\nHowever, when i try to submit submission.csv from my inference, it keeps telling me: \"Your notebook cannot use non-standard datasets in this competition\". \n\nAny idea on how I can submit anything?\n\nEDIT: solved by manually uploading model.h5 instead of generating from kernel output.", "695190": "did you attach output of the training kernel? if yes, thats not allowed.", "695191": "Per the \"Code Requirements\" section:\n\n```No using other Kaggle notebooks or utility scripts as inputs to your submission notebook. External data also cannot be externally-referenced (i.e. via BigQuery, Github, external URL). Instead, load external models or datasets directly as an external data source.\n```\n\nYou'll need to upload the `h5` and `csv` files as a dataset, instead of using the output from a previous kernel.\n\nTry using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.", "695192": "No I created a new dataset from the kernel, and loaded that dataset into my inference kernel. I didn't directly load the kernel output.", "695193": "> Try using the the \"New Dataset\" button next to the output files in your first kernel. And then reference that dataset in your second.\n\nExactly what i did lol", "695194": "Oh ¯\\_(ツ)_/¯ maybe kaggle knows that the new dataset was created from a kernel. Did you try downloading and re-uploading into a dataset not linked to the first kernel?", "695195": "That's a good idea, i'm trying this right now haha", "695211": "This worked :)", "723027": "I had the same error when trying to create a new dataset with a file from the URL option. After downloading from this URL to my machine -> creating a new dataset -> actually uploading the file to the dataset - it worked.", "949851": "I want to call keras's inception_v3 model (via ' from keras.applications import inception_v3 ') inside my notebook. Is this allowed? I can't do this while being offline though so I cant get the submit option up. this is as when i call keras. inception_v3.InceptionV3 it tries to load a h5 file online from the url ' ttps://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5 ' ?? \n\nIs there a way for me to use this locally via my call to keras.applications as \n\n``TimeDistributed(inception_v3.InceptionV3(include_top=False,pooling='max',weights='imagenet'))(img_input)\n\nin my code\n\nPlease heeeelp Im stuck :(" }, "source": "meta" }