I am trying to dive into tensorflow 2.0, since TPU looks like a hit of this season.
\n\nSo far I am running a kernels, like thise one beautiful kernel by Chris. But am facing a following error:
\n\n\nUnimplementedError: 7 root error(s) found.\n (0) Unimplemented: {{function_node __inference_train_function_233480}} File system scheme '[local]' not implemented (file: '/content/tfrecords/train06-2175.tfrec')\n [[{{node MultiDeviceIteratorGetNextFromShard}}]]\n [[RemoteCall]]\n [[IteratorGetNextAsOptional_6]]\n
I see there is a closed issue on github with a resolution that using local filesystem is not allowed on colab and they team is working to fix this.
\n\nThe question is - is there any success in resolving this?
", "messages": [ { "id": 956584, "postDate": "2020-08-03T16:00:05.570Z", "content": "I am trying to dive into tensorflow 2.0, since TPU looks like a hit of this season.
\n\nSo far I am running a kernels, like thise one beautiful kernel by Chris. But am facing a following error:
\n\n\nUnimplementedError: 7 root error(s) found.\n (0) Unimplemented: {{function_node __inference_train_function_233480}} File system scheme '[local]' not implemented (file: '/content/tfrecords/train06-2175.tfrec')\n [[{{node MultiDeviceIteratorGetNextFromShard}}]]\n [[RemoteCall]]\n [[IteratorGetNextAsOptional_6]]\n
I see there is a closed issue on github with a resolution that using local filesystem is not allowed on colab and they team is working to fix this.
\n\nThe question is - is there any success in resolving this?
", "rawMarkdown": "I am trying to dive into tensorflow 2.0, since TPU looks like a hit of this season.\n\nSo far I am running a kernels, like [thise one](https://www.kaggle.com/cdeotte/triple-stratified-kfold-with-tfrecords) beautiful kernel by Chris. But am facing a following error:\n\n```\nUnimplementedError: 7 root error(s) found.\n (0) Unimplemented: {{function_node __inference_train_function_233480}} File system scheme '[local]' not implemented (file: '/content/tfrecords/train06-2175.tfrec')\n\t [[{{node MultiDeviceIteratorGetNextFromShard}}]]\n\t [[RemoteCall]]\n\t [[IteratorGetNextAsOptional_6]]\n```\n\nI see there is a [closed issue](https://github.com/tensorflow/tensorflow/issues/32651) on github with a resolution that using local filesystem is not allowed on colab and they team is working to fix this.\n\nThe question is - is there any success in resolving this?", "votes": 6 }, { "id": 956681, "postDate": "2020-08-03T17:27:49.173Z", "content": "Roman, you need to load your tfrecords from a Google Cloud Storage bucket rather than from your local file system or Google Drive. Otherwise it won't work with a TPU. Check out In[13] in my public notebook for an example. And here is another public kernel that I wrote for identifying the most recent GCS addresses of some popular tfrecord data sets. Hope it helps.
", "rawMarkdown": "Roman, you need to load your tfrecords from a Google Cloud Storage bucket rather than from your local file system or Google Drive. Otherwise it won't work with a TPU. Check out In[13] in [my public notebook](https://www.kaggle.com/graf10a/efficientnet-bn-tabular-features-tf-cv5-512x512) for an example. And here is another [public kernel](https://www.kaggle.com/graf10a/siim-show-gcs-bucket-addresses) that I wrote for identifying the most recent GCS addresses of some popular tfrecord data sets. Hope it helps.", "votes": 3 }, { "id": 957808, "postDate": "2020-08-04T15:13:44.623Z", "content": "Hi @nroman! Since your original question was addressed to the Kaggle team, I wanted to chime in and say that the advice you've received from the community here is correct regarding data storage locations. Let us know if there's anything else we can do to help out!
", "rawMarkdown": "Hi @nroman! Since your original question was addressed to the Kaggle team, I wanted to chime in and say that the advice you've received from the community here is correct regarding data storage locations. Let us know if there's anything else we can do to help out!", "votes": 1, "replies": [ { "id": 957826, "postDate": "2020-08-04T15:23:04.117Z", "rawMarkdown": "", "isDeleted": true }, { "id": 960758, "postDate": "2020-08-06T16:51:40.770Z", "content": "yes, and thank you for citing the Kaggle documentation!
", "rawMarkdown": "yes, and thank you for citing the Kaggle documentation!" } ] }, { "id": 956708, "postDate": "2020-08-03T17:49:07.843Z", "content": "@nroman I've also faced the same things. However, @graf10a is right.
", "rawMarkdown": "@nroman I've also faced the same things. However, @graf10a is right.", "votes": 1 }, { "id": 956834, "postDate": "2020-08-03T20:29:15.277Z", "content": "You need either to make the data public (in Kaggle Datasets) or to upload it in a bucket of your Google Cloud Account (the latter is not free)
", "rawMarkdown": "You need either to make the data public (in Kaggle Datasets) or to upload it in a bucket of your Google Cloud Account (the latter is not free)" }, { "id": 957025, "postDate": "2020-08-04T02:24:51.953Z", "rawMarkdown": "", "isDeleted": true }, { "id": 957015, "postDate": "2020-08-04T02:10:40.627Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 956681, "author_name": "Alexey Pronin", "author_url": "", "post_date": "2020-08-03T17:27:49.173000", "content": "Roman, you need to load your tfrecords from a Google Cloud Storage bucket rather than from your local file system or Google Drive. Otherwise it won't work with a TPU. Check out In[13] in my public notebook for an example. And here is another public kernel that I wrote for identifying the most recent GCS addresses of some popular tfrecord data sets. Hope it helps.
", "votes": 3, "replies": [] }, { "id": 957808, "author_name": "Jesse Mostipak", "author_url": "", "post_date": "2020-08-04T15:13:44.623000", "content": "Hi @nroman! Since your original question was addressed to the Kaggle team, I wanted to chime in and say that the advice you've received from the community here is correct regarding data storage locations. Let us know if there's anything else we can do to help out!
", "votes": 1, "replies": [ { "id": 957826, "author_name": "", "author_url": "", "post_date": "2020-08-04T15:23:04.117000", "content": "", "votes": 0, "replies": [] }, { "id": 960758, "author_name": "Jesse Mostipak", "author_url": "", "post_date": "2020-08-06T16:51:40.770000", "content": "yes, and thank you for citing the Kaggle documentation!
", "votes": 0, "replies": [] } ] }, { "id": 956708, "author_name": "Innat", "author_url": "", "post_date": "2020-08-03T17:49:07.843000", "content": "@nroman I've also faced the same things. However, @graf10a is right.
", "votes": 1, "replies": [] }, { "id": 956834, "author_name": "Serigne ", "author_url": "", "post_date": "2020-08-03T20:29:15.277000", "content": "You need either to make the data public (in Kaggle Datasets) or to upload it in a bucket of your Google Cloud Account (the latter is not free)
", "votes": 0, "replies": [] }, { "id": 957025, "author_name": "", "author_url": "", "post_date": "2020-08-04T02:24:51.953000", "content": "", "votes": 0, "replies": [] }, { "id": 957015, "author_name": "", "author_url": "", "post_date": "2020-08-04T02:10:40.627000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "956584": "I am trying to dive into tensorflow 2.0, since TPU looks like a hit of this season.\n\nSo far I am running a kernels, like [thise one](https://www.kaggle.com/cdeotte/triple-stratified-kfold-with-tfrecords) beautiful kernel by Chris. But am facing a following error:\n\n```\nUnimplementedError: 7 root error(s) found.\n (0) Unimplemented: {{function_node __inference_train_function_233480}} File system scheme '[local]' not implemented (file: '/content/tfrecords/train06-2175.tfrec')\n\t [[{{node MultiDeviceIteratorGetNextFromShard}}]]\n\t [[RemoteCall]]\n\t [[IteratorGetNextAsOptional_6]]\n```\n\nI see there is a [closed issue](https://github.com/tensorflow/tensorflow/issues/32651) on github with a resolution that using local filesystem is not allowed on colab and they team is working to fix this.\n\nThe question is - is there any success in resolving this?", "956681": "Roman, you need to load your tfrecords from a Google Cloud Storage bucket rather than from your local file system or Google Drive. Otherwise it won't work with a TPU. Check out In[13] in [my public notebook](https://www.kaggle.com/graf10a/efficientnet-bn-tabular-features-tf-cv5-512x512) for an example. And here is another [public kernel](https://www.kaggle.com/graf10a/siim-show-gcs-bucket-addresses) that I wrote for identifying the most recent GCS addresses of some popular tfrecord data sets. Hope it helps.", "957808": "Hi @nroman! Since your original question was addressed to the Kaggle team, I wanted to chime in and say that the advice you've received from the community here is correct regarding data storage locations. Let us know if there's anything else we can do to help out!", "956708": "@nroman I've also faced the same things. However, @graf10a is right.", "956834": "You need either to make the data public (in Kaggle Datasets) or to upload it in a bucket of your Google Cloud Account (the latter is not free)", "957025": "", "957015": "" } }