![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1530576%2F6a0f95c5788eeed5cd6476273ce65528%2Fdeadline_7days.png?generation=1660838300893808&alt=media\")
Update 1: The inference will be done on .wav files instead of .mp3 files.
\nUpdate 2: Do note that inference notebooks should have internet disabled like mentioned In the Rules.
After 2 long months, Phase 1 of DL Sprint is ending this 25th of August. There has been a slight change of plans: the good news is your job here is nearly done. The only thing left is to properly submit your solution so we can validate your submission. Please go through the following carefully:
\n# EVENTS
\n (There are changes from the original timeline, please go through carefully)
\nPhase 1 Kaggle Submission: The current Kaggle competition will end on 25th August at 12:00 am. Participants are advised to submit their final predictions on Kaggle before 24th August 11:55 pm to avoid late submissions. This will mark the end of Phase 1. (There will be no Phase 1 Part 2)
All participants will have to submit their solutions to dlsprint2022@gmail.com. The submission format is given under the SUBMISSION FORMAT section (please go there for details). This submission deadline is on 26th August at 12:00 am. Participants are advised to submit their solutions before 25th August 11:55 pm to avoid any issues. Do note that late submission will result in disqualification. Furthermore, if you do not submit your solution in the described format, you will not be eligible for any Phase 2 prize (excluding the Best Notebooks Prize). You can always contact us if you have any questions regarding the format.
\nPhase 2 Offline Event: We will run your submission code on our hidden test dataset to generate a new leaderboard. The top 10 teams from this hidden dataset leaderboard will be invited to present their solutions at the Department of CSE, BUET premises on 2nd September at 9 am. The teams must report at 9 am to do a presentation before a panel of judges. The prize-giving ceremony will also be held on the same day. Do note that offline presentation from at least one member of the team is mandatory to be eligible for the prize.
\nP.S. The top 10 teams will also be asked to submit a simple 4-page paper based on the IEEE Conference paper template before the Phase 2 round and the paper should contain Introduction, Methodology, Results & Discussions, and Conclusion sections. This write-up is needed for a better understanding of your solution and thought process.
\n# SUBMISSION FORMAT
\nInclude the following information on a .txt / .pdf / .docx file and send it to dlsprint2022@gmail.com. Mention your Kaggle team name in the email subject.
If you have any questions you can comment on this discussion or reach out to us at dlsprint2022@gmail.com.
\n
Update 1: The inference will be done on .wav files instead of .mp3 files.
\nUpdate 2: Do note that inference notebooks should have internet disabled like mentioned In the Rules.
After 2 long months, Phase 1 of DL Sprint is ending this 25th of August. There has been a slight change of plans: the good news is your job here is nearly done. The only thing left is to properly submit your solution so we can validate your submission. Please go through the following carefully:
\n# EVENTS
\n (There are changes from the original timeline, please go through carefully)
\nPhase 1 Kaggle Submission: The current Kaggle competition will end on 25th August at 12:00 am. Participants are advised to submit their final predictions on Kaggle before 24th August 11:55 pm to avoid late submissions. This will mark the end of Phase 1. (There will be no Phase 1 Part 2)
All participants will have to submit their solutions to dlsprint2022@gmail.com. The submission format is given under the SUBMISSION FORMAT section (please go there for details). This submission deadline is on 26th August at 12:00 am. Participants are advised to submit their solutions before 25th August 11:55 pm to avoid any issues. Do note that late submission will result in disqualification. Furthermore, if you do not submit your solution in the described format, you will not be eligible for any Phase 2 prize (excluding the Best Notebooks Prize). You can always contact us if you have any questions regarding the format.
\nPhase 2 Offline Event: We will run your submission code on our hidden test dataset to generate a new leaderboard. The top 10 teams from this hidden dataset leaderboard will be invited to present their solutions at the Department of CSE, BUET premises on 2nd September at 9 am. The teams must report at 9 am to do a presentation before a panel of judges. The prize-giving ceremony will also be held on the same day. Do note that offline presentation from at least one member of the team is mandatory to be eligible for the prize.
\nP.S. The top 10 teams will also be asked to submit a simple 4-page paper based on the IEEE Conference paper template before the Phase 2 round and the paper should contain Introduction, Methodology, Results & Discussions, and Conclusion sections. This write-up is needed for a better understanding of your solution and thought process.
\n# SUBMISSION FORMAT
\nInclude the following information on a .txt / .pdf / .docx file and send it to dlsprint2022@gmail.com. Mention your Kaggle team name in the email subject.
If you have any questions you can comment on this discussion or reach out to us at dlsprint2022@gmail.com.
\n
Will you run the inference on mp3 files or wav?
", "rawMarkdown": "Will you run the inference on mp3 files or wav?", "votes": null }, { "id": "1908010", "postDate": "08/21/2022 09:25:51", "content": "It will be run on .mp3 files.
", "rawMarkdown": "It will be run on .mp3 files.", "votes": null }, { "id": "1908104", "postDate": "08/21/2022 11:20:51", "content": "It takes 6:30 hours to infer 7747 test dataset with the inference notebook, whereas it takes 2:30 hours to infer on wav files.
", "rawMarkdown": "It takes 6:30 hours to infer 7747 test dataset with the inference notebook, whereas it takes 2:30 hours to infer on wav files.", "votes": null }, { "id": "1908166", "postDate": "08/21/2022 12:09:20", "content": "You have raised a valid point. Thank you! We had an internal discussion and have decided to use .wav files for inference. The announcement will be updated accordingly.
", "rawMarkdown": "You have raised a valid point. Thank you! We had an internal discussion and have decided to use .wav files for inference. The announcement will be updated accordingly.", "votes": null }, { "id": "1908409", "postDate": "08/21/2022 16:27:46", "content": "Do the notebooks have to be kaggle notebooks? We ran some of the training notebooks in Colab since it provides a longer run time. Can we submit those?
\nAnother Question is will internet be enabled or disabled while Inference? Can the inference notebook load the weights from hugging face like this notebook?
\nThanks.
The notebooks do have to be Kaggle notebooks. It shouldn't be much work refactoring the code from Colab to Kaggle. You can additionally share your colab codes but Kaggle notebooks are required for standardising the submissions.
\nAlso internet will be disabled for inference. You can save the weights as a Kaggle dataset and load the weights from there. Also don't forget to share the dataset with the mentioned people!
\nGood luck!
", "rawMarkdown": "The notebooks do have to be Kaggle notebooks. It shouldn't be much work refactoring the code from Colab to Kaggle. You can additionally share your colab codes but Kaggle notebooks are required for standardising the submissions.\n\nAlso internet will be disabled for inference. You can save the weights as a Kaggle dataset and load the weights from there. Also don't forget to share the dataset with the mentioned people!\n\nGood luck!", "votes": null }, { "id": "1908422", "postDate": "08/21/2022 16:56:39", "content": "Got it. Thanks for the detailed answer.
", "rawMarkdown": "Got it. Thanks for the detailed answer.", "votes": null } ], "comments": [ { "id": 1907969, "author_name": "sirajissalakeen", "author_url": "", "post_date": "08/21/2022 08:41:47", "content": "Will you run the inference on mp3 files or wav?
", "votes": null, "replies": [ { "id": 1908010, "author_name": "nexh98", "author_url": "", "post_date": "08/21/2022 09:25:51", "content": "It will be run on .mp3 files.
", "votes": null, "replies": [] }, { "id": 1908104, "author_name": "sirajissalakeen", "author_url": "", "post_date": "08/21/2022 11:20:51", "content": "It takes 6:30 hours to infer 7747 test dataset with the inference notebook, whereas it takes 2:30 hours to infer on wav files.
", "votes": null, "replies": [] }, { "id": 1908166, "author_name": "nexh98", "author_url": "", "post_date": "08/21/2022 12:09:20", "content": "You have raised a valid point. Thank you! We had an internal discussion and have decided to use .wav files for inference. The announcement will be updated accordingly.
", "votes": null, "replies": [] } ] }, { "id": 1908409, "author_name": "mbmmurad", "author_url": "", "post_date": "08/21/2022 16:27:46", "content": "Do the notebooks have to be kaggle notebooks? We ran some of the training notebooks in Colab since it provides a longer run time. Can we submit those?
\nAnother Question is will internet be enabled or disabled while Inference? Can the inference notebook load the weights from hugging face like this notebook?
\nThanks.
The notebooks do have to be Kaggle notebooks. It shouldn't be much work refactoring the code from Colab to Kaggle. You can additionally share your colab codes but Kaggle notebooks are required for standardising the submissions.
\nAlso internet will be disabled for inference. You can save the weights as a Kaggle dataset and load the weights from there. Also don't forget to share the dataset with the mentioned people!
\nGood luck!
", "votes": null, "replies": [] }, { "id": 1908422, "author_name": "mbmmurad", "author_url": "", "post_date": "08/21/2022 16:56:39", "content": "Got it. Thanks for the detailed answer.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1904944": "**Update 1: The inference will be done on .wav files instead of .mp3 files.**\n**Update 2: Do note that inference notebooks should have internet disabled like mentioned In the Rules.**\n\nAfter 2 long months, Phase 1 of DL Sprint is ending this 25th of August. There has been a slight change of plans: the good news is your job here is nearly done. The only thing left is to properly submit your solution so we can validate your submission. **Please go through the following carefully:**\n\n**# EVENTS