\n
![\"\"](\"https://i.imgur.com/HJ82SvS.png\")
Okay, now that the competition is over, I implore you to help me understand, why this notebook will not submit for evaluation, i.e. why it cannot find the submission.csv file. It it a mix of Stefan Kahls two notebooks, the one for training and the one for submitting to the competition.
\nIf you have access to the versions, I will tell you that version 4 succeded in showing that a submission.csv file was created, only I had called it submission1.csv and submission2.csv, and thus it could not be used for anything. BUT IT COULD TELL THAT IT WAS THERE. However, all other versions fail.
\nI asked Kaggle for the difference, between e.g. version 6 and version 4, and this is what it shows:
\n
The ONLY difference was the name. And the fact that I used a maximum of 5 files in version 4 and 1500 in all others.
\n![\"\"](\"https://i.imgur.com/15sQ99k.png\")
![\"\"](\"https://i.imgur.com/a4Q7r9h.png\")
I know that I'm saving the submissions.csv file two times, to different paths, that resolve to the same place. But as far as I can tell, it makes no difference, whether I do both of them or only one of them.
\nIf you make a commit and run all version of the latest version and try to submit it to the competition, what then? Does the submissions file appear for you? If not, can you spot why it wouldn't!?
\nhttps://www.kaggle.com/tobiasbonnesen/birdclef2021-model-training-thesis-baseline
\nI'm training my real model on my own computer, so I'm not super dependent on this for further scoring. It just annoys me so much, that it doesn't work, when as far as I can tell, it should.
\nAnd no, it's not because of training time. This takes no time to train. You can even change the max audio files settings to allow for super-duper fast training time. 🤓
\nThank you 🙏🙏🙏🙏🙏🙏🙏
", "messages": [ { "id": "1340378", "postDate": "06/07/2021 20:32:48", "content": "Okay, now that the competition is over, I implore you to help me understand, why this notebook will not submit for evaluation, i.e. why it cannot find the submission.csv file. It it a mix of Stefan Kahls two notebooks, the one for training and the one for submitting to the competition.
\nIf you have access to the versions, I will tell you that version 4 succeded in showing that a submission.csv file was created, only I had called it submission1.csv and submission2.csv, and thus it could not be used for anything. BUT IT COULD TELL THAT IT WAS THERE. However, all other versions fail.
\nI asked Kaggle for the difference, between e.g. version 6 and version 4, and this is what it shows:
\n
The ONLY difference was the name. And the fact that I used a maximum of 5 files in version 4 and 1500 in all others.
\n
I know that I'm saving the submissions.csv file two times, to different paths, that resolve to the same place. But as far as I can tell, it makes no difference, whether I do both of them or only one of them.
\nIf you make a commit and run all version of the latest version and try to submit it to the competition, what then? Does the submissions file appear for you? If not, can you spot why it wouldn't!?
\nhttps://www.kaggle.com/tobiasbonnesen/birdclef2021-model-training-thesis-baseline
\nI'm training my real model on my own computer, so I'm not super dependent on this for further scoring. It just annoys me so much, that it doesn't work, when as far as I can tell, it should.
\nAnd no, it's not because of training time. This takes no time to train. You can even change the max audio files settings to allow for super-duper fast training time. 🤓
\nThank you 🙏🙏🙏🙏🙏🙏🙏
", "rawMarkdown": "Okay, now that the competition is over, I implore you to help me understand, why this notebook will not submit for evaluation, i.e. why it cannot find the submission.csv file. It it a mix of Stefan Kahls two notebooks, the one for training and the one for submitting to the competition. \n\nIf you have access to the versions, I will tell you that version 4 succeded in showing that a submission.csv file was created, only I had called it submission1.csv and submission2.csv, and thus it could not be used for anything. BUT IT COULD TELL THAT IT WAS THERE. However, all other versions fail. \n\nI asked Kaggle for the difference, between e.g. version 6 and version 4, and this is what it shows:\n\n\nThe ONLY difference was the name. And the fact that I used a maximum of 5 files in version 4 and 1500 in all others.\n\n\n\n\n\nI know that I'm saving the submissions.csv file two times, to different paths, that resolve to the same place. But as far as I can tell, it makes no difference, whether I do both of them or only one of them.\n\nIf you make a commit and run all version of the latest version and try to submit it to the competition, what then? Does the submissions file appear for you? If not, can you spot why it wouldn't!?\n\nhttps://www.kaggle.com/tobiasbonnesen/birdclef2021-model-training-thesis-baseline\n\nI'm training my real model on my own computer, so I'm not super dependent on this for further scoring. It just annoys me so much, that it doesn't work, when as far as I can tell, it should.\n\nAnd no, it's not because of training time. This takes no time to train. You can even change the max audio files settings to allow for super-duper fast training time. 🤓\n\nThank you 🙏🙏🙏🙏🙏🙏🙏", "votes": null }, { "id": "1341132", "postDate": "06/08/2021 12:55:51", "content": "Delete all unneccesary output except submission.csv. That might help.
", "rawMarkdown": "Delete all unneccesary output except submission.csv. That might help.", "votes": null }, { "id": "1341144", "postDate": "06/08/2021 13:02:04", "content": "Thanks for the reply :-)
\nI tried that, it didn't do any diffeference. Interestingly enough, for version 4, where it actually detects some files as outputted, it is the case that several submissions files were outputted.
\nIf you go to my latest version where I'm only outputting one csv file
\nhttps://www.kaggle.com/tobiasbonnesen/birdclef2021-model-training-thesis-baseline
\nAnd try to submit it it for evaluation, you'll see it doesn't work.
I'm pretty sure this is a kaggle bug. <.< Though what triggers it I don't know. But maybe I'm wrong.
", "rawMarkdown": "Thanks for the reply :-) \n\nI tried that, it didn't do any diffeference. Interestingly enough, for version 4, where it actually detects some files as outputted, it is the case that several submissions files were outputted. \n\nIf you go to my latest version where I'm only outputting one csv file\nhttps://www.kaggle.com/tobiasbonnesen/birdclef2021-model-training-thesis-baseline\nAnd try to submit it it for evaluation, you'll see it doesn't work. \n\nI'm pretty sure this is a kaggle bug. <.< Though what triggers it I don't know. But maybe I'm wrong.", "votes": null }, { "id": "1341177", "postDate": "06/08/2021 13:33:58", "content": "Output Size 20.4 MB (on version 9/9)
\nThat is not \"one csv file\". Delete all your spectrograms that you generated during training.
", "rawMarkdown": "Output Size 20.4 MB (on version 9/9)\n\nThat is not \"one csv file\". Delete all your spectrograms that you generated during training.", "votes": null }, { "id": "1341361", "postDate": "06/08/2021 15:54:05", "content": "There is another difference. In the left code you output submission2.csv, and you don't output it in the right code. Anyway, the submission file must be named submission.csv.
\nAlso why are you using a path for saving your sub? Here is what I used:
\ndf[['row_id', 'birds']].to_csv('submission.csv', index=False)\nDelete all unneccesary output except submission.csv. That might help.
", "votes": null, "replies": [ { "id": 1341144, "author_name": "tobiasbonnesen", "author_url": "", "post_date": "06/08/2021 13:02:04", "content": "Thanks for the reply :-)
\nI tried that, it didn't do any diffeference. Interestingly enough, for version 4, where it actually detects some files as outputted, it is the case that several submissions files were outputted.
\nIf you go to my latest version where I'm only outputting one csv file
\nhttps://www.kaggle.com/tobiasbonnesen/birdclef2021-model-training-thesis-baseline
\nAnd try to submit it it for evaluation, you'll see it doesn't work.
I'm pretty sure this is a kaggle bug. <.< Though what triggers it I don't know. But maybe I'm wrong.
", "votes": null, "replies": [] }, { "id": 1341177, "author_name": "fffrrt", "author_url": "", "post_date": "06/08/2021 13:33:58", "content": "Output Size 20.4 MB (on version 9/9)
\nThat is not \"one csv file\". Delete all your spectrograms that you generated during training.
", "votes": null, "replies": [] } ] }, { "id": 1341361, "author_name": "cpmpml", "author_url": "", "post_date": "06/08/2021 15:54:05", "content": "There is another difference. In the left code you output submission2.csv, and you don't output it in the right code. Anyway, the submission file must be named submission.csv.
\nAlso why are you using a path for saving your sub? Here is what I used:
\ndf[['row_id', 'birds']].to_csv('submission.csv', index=False)\n