Hi !
\nIn this competition I used only Kaggle compute resources for my experiments. I dont own a computer with graphic card hence can't train deep learning models locally. I would be up for colab pro but that option is unavailable where I live.
I wanted to know if any of the top teams (top 100) developed models solely using Kaggle compute resources. This will help me set realistic expectations about my standing in future competition.
", "messages": [ { "id": 983529, "postDate": "2020-08-24T11:35:08.810Z", "content": "Hi everybody! I exclusively used Kaggle TPUs and GPUs.
", "rawMarkdown": "Hi everybody! I exclusively used Kaggle TPUs and GPUs.", "votes": 5, "replies": [ { "id": 983584, "postDate": "2020-08-24T12:21:57.990Z", "content": "Congratulations on amazing rank ! That's great to know.
", "rawMarkdown": "Congratulations on amazing rank ! That's great to know. ", "votes": 1 } ] }, { "id": 979772, "postDate": "2020-08-21T05:37:54.650Z", "content": "Hi !
\nIn this competition I used only Kaggle compute resources for my experiments. I dont own a computer with graphic card hence can't train deep learning models locally. I would be up for colab pro but that option is unavailable where I live.
I wanted to know if any of the top teams (top 100) developed models solely using Kaggle compute resources. This will help me set realistic expectations about my standing in future competition.
", "rawMarkdown": "Hi ! \nIn this competition I used only Kaggle compute resources for my experiments. I dont own a computer with graphic card hence can't train deep learning models locally. I would be up for colab pro but that option is unavailable where I live. \n\nI wanted to know if any of the top teams (top 100) developed models solely using Kaggle compute resources. This will help me set realistic expectations about my standing in future competition. ", "votes": 5 }, { "id": 980994, "postDate": "2020-08-22T04:58:35.873Z", "content": "I exclusively used kaggle TPUs and GPUs. I've only learnt that colab TPU existed by reading this discussion, so thank you for that. Also, my best submission did not have any blending with public notebooks.
", "rawMarkdown": "I exclusively used kaggle TPUs and GPUs. I've only learnt that colab TPU existed by reading this discussion, so thank you for that. Also, my best submission did not have any blending with public notebooks.", "votes": 3, "replies": [ { "id": 981099, "postDate": "2020-08-22T06:57:54.137Z", "content": "Congratulations on 85th position ! Even I got to know that TPU is available in google colab ! Will definitely try out.
", "rawMarkdown": "Congratulations on 85th position ! Even I got to know that TPU is available in google colab ! Will definitely try out. " } ] }, { "id": 980522, "postDate": "2020-08-21T16:57:04.850Z", "content": "I used Kaggle TPU mostly. Towards the last week, I went started using colab TPU because it seemed that others were using it. I'm glad I used the colab TPU to run experiments because I hit my TPU quota very early on. I wish I had started using colab TPU earlier. Lesson learned for me.
", "rawMarkdown": "I used Kaggle TPU mostly. Towards the last week, I went started using colab TPU because it seemed that others were using it. I'm glad I used the colab TPU to run experiments because I hit my TPU quota very early on. I wish I had started using colab TPU earlier. Lesson learned for me.", "votes": 1, "replies": [ { "id": 981097, "postDate": "2020-08-22T06:56:53.063Z", "content": "Congratulations on the 74th rank ! Used TPU's alot in this competition. Will surely setup colab for the next competition ! See you on the LB ! ๐
", "rawMarkdown": "Congratulations on the 74th rank ! Used TPU's alot in this competition. Will surely setup colab for the next competition ! See you on the LB ! ๐" } ] }, { "id": 980085, "postDate": "2020-08-21T10:21:37.020Z", "content": "Yes u can still upgrade to Colab pro even outside US
\nbut for me it is not very stable and sometime just freeze for no reason (maybe my problem)
Thanks for pitching in ! Umm signing up for colab pro will be falsifying information ! But let's see, it's for the greater good ๐
", "rawMarkdown": "Thanks for pitching in ! Umm signing up for colab pro will be falsifying information ! But let's see, it's for the greater good ๐" } ] }, { "id": 981102, "postDate": "2020-08-22T06:59:36.333Z", "content": "Update: I think I have my answers. The Kaggle competitions are quiet democratic and there are free resources available for people to train their models, compete and win a competition ! Required investment is your own time. Thanks all. ๐๐
", "rawMarkdown": "Update: I think I have my answers. The Kaggle competitions are quiet democratic and there are free resources available for people to train their models, compete and win a competition ! Required investment is your own time. Thanks all. ๐๐", "votes": 2 }, { "id": 979882, "postDate": "2020-08-21T07:14:47.590Z", "content": "I used Colab Pro for this competition. It's quite enough to check the majority of ideas and models.
", "rawMarkdown": "I used Colab Pro for this competition. It's quite enough to check the majority of ideas and models.", "votes": 2, "replies": [ { "id": 979933, "postDate": "2020-08-21T08:05:01.247Z", "content": "Thanks for responding, I would be going for google colab for my experiments now.
", "rawMarkdown": "Thanks for responding, I would be going for google colab for my experiments now. " } ] }, { "id": 979796, "postDate": "2020-08-21T05:57:24.997Z", "content": "We used Google Colab quite extensively. Some of our teammates have Pro accounts, but I managed to train quite a few models with a regular account. The TPU quotas are not explicitly disclosed, but I had at least about 5-6 hours every other day to perform experiments. This was a very nice addition to the Kaggle notebooks.
", "rawMarkdown": "We used Google Colab quite extensively. Some of our teammates have Pro accounts, but I managed to train quite a few models with a regular account. The TPU quotas are not explicitly disclosed, but I had at least about 5-6 hours every other day to perform experiments. This was a very nice addition to the Kaggle notebooks.", "votes": 2, "replies": [ { "id": 979806, "postDate": "2020-08-21T06:05:06.320Z", "content": "Congratulations on getting the 13th rank ! I will surely use google colab for my next competition. Before that I think I will be buying more space on Google drive for my datasets.
", "rawMarkdown": "Congratulations on getting the 13th rank ! I will surely use google colab for my next competition. Before that I think I will be buying more space on Google drive for my datasets. " }, { "id": 980526, "postDate": "2020-08-21T17:00:30.357Z", "content": "If you use TFRecords (namely the one's made by Chris, then you just needed to grab the gcs addresses, or use Kaggle API to get the datasets)
", "rawMarkdown": "If you use TFRecords (namely the one's made by Chris, then you just needed to grab the [gcs addresses](https://www.kaggle.com/teeyee314/melanoma-gcs-paths), or use Kaggle API to get the datasets)" } ] }, { "id": 979778, "postDate": "2020-08-21T05:44:07.887Z", "content": "Hi there,
\nNot quite the top 100 team here, but we used Google Colab for the entire course of the competition. While it is supposed to be only restricted to residents of the US, none of us actually are from there and we were still able to buy Pro plans.
", "rawMarkdown": "Hi there,\n\nNot quite the top 100 team here, but we used Google Colab for the entire course of the competition. While it is supposed to be only restricted to residents of the US, none of us actually are from there and we were still able to buy Pro plans.\n\n", "votes": 2, "replies": [ { "id": 979804, "postDate": "2020-08-21T06:02:56.280Z", "content": "@group16 , congratulations on your rank ! Thanks for shedding light. I think in the next competition I will try this out.
", "rawMarkdown": "@group16 , congratulations on your rank ! Thanks for shedding light. I think in the next competition I will try this out.", "votes": 1 }, { "id": 989332, "postDate": "2020-08-28T18:11:59.607Z", "content": "Is there not any problem if they detect that you purchase pro plan from outside the usa? I want to purchase access from Argentina but I am afraid of loosing my gmail accountโฆ
", "rawMarkdown": "\nIs there not any problem if they detect that you purchase pro plan from outside the usa? I want to purchase access from Argentina but I am afraid of loosing my gmail account...\n" } ] }, { "id": 980246, "postDate": "2020-08-21T12:41:38.803Z", "content": "I would love to know about it too.
", "rawMarkdown": "I would love to know about it too." }, { "id": 989084, "postDate": "2020-08-28T14:32:14.403Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 983529, "author_name": "Masdevallia", "author_url": "", "post_date": "2020-08-24T11:35:08.810000", "content": "Hi everybody! I exclusively used Kaggle TPUs and GPUs.
", "votes": 5, "replies": [ { "id": 983584, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-24T12:21:57.990000", "content": "Congratulations on amazing rank ! That's great to know.
", "votes": 1, "replies": [] } ] }, { "id": 980994, "author_name": "Sam Klein", "author_url": "", "post_date": "2020-08-22T04:58:35.873000", "content": "I exclusively used kaggle TPUs and GPUs. I've only learnt that colab TPU existed by reading this discussion, so thank you for that. Also, my best submission did not have any blending with public notebooks.
", "votes": 3, "replies": [ { "id": 981099, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-22T06:57:54.137000", "content": "Congratulations on 85th position ! Even I got to know that TPU is available in google colab ! Will definitely try out.
", "votes": 0, "replies": [] } ] }, { "id": 980522, "author_name": "Tim Yee", "author_url": "", "post_date": "2020-08-21T16:57:04.850000", "content": "I used Kaggle TPU mostly. Towards the last week, I went started using colab TPU because it seemed that others were using it. I'm glad I used the colab TPU to run experiments because I hit my TPU quota very early on. I wish I had started using colab TPU earlier. Lesson learned for me.
", "votes": 1, "replies": [ { "id": 981097, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-22T06:56:53.063000", "content": "Congratulations on the 74th rank ! Used TPU's alot in this competition. Will surely setup colab for the next competition ! See you on the LB ! ๐
", "votes": 0, "replies": [] } ] }, { "id": 980085, "author_name": "Fivestomars", "author_url": "", "post_date": "2020-08-21T10:21:37.020000", "content": "Yes u can still upgrade to Colab pro even outside US
\nbut for me it is not very stable and sometime just freeze for no reason (maybe my problem)
Thanks for pitching in ! Umm signing up for colab pro will be falsifying information ! But let's see, it's for the greater good ๐
", "votes": 0, "replies": [] } ] }, { "id": 981102, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-22T06:59:36.333000", "content": "Update: I think I have my answers. The Kaggle competitions are quiet democratic and there are free resources available for people to train their models, compete and win a competition ! Required investment is your own time. Thanks all. ๐๐
", "votes": 2, "replies": [] }, { "id": 979882, "author_name": "Maksim Zhdanov", "author_url": "", "post_date": "2020-08-21T07:14:47.590000", "content": "I used Colab Pro for this competition. It's quite enough to check the majority of ideas and models.
", "votes": 2, "replies": [ { "id": 979933, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-21T08:05:01.247000", "content": "Thanks for responding, I would be going for google colab for my experiments now.
", "votes": 0, "replies": [] } ] }, { "id": 979796, "author_name": "Nikita Kozodoi", "author_url": "", "post_date": "2020-08-21T05:57:24.997000", "content": "We used Google Colab quite extensively. Some of our teammates have Pro accounts, but I managed to train quite a few models with a regular account. The TPU quotas are not explicitly disclosed, but I had at least about 5-6 hours every other day to perform experiments. This was a very nice addition to the Kaggle notebooks.
", "votes": 2, "replies": [ { "id": 979806, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-21T06:05:06.320000", "content": "Congratulations on getting the 13th rank ! I will surely use google colab for my next competition. Before that I think I will be buying more space on Google drive for my datasets.
", "votes": 0, "replies": [] }, { "id": 980526, "author_name": "Tim Yee", "author_url": "", "post_date": "2020-08-21T17:00:30.357000", "content": "If you use TFRecords (namely the one's made by Chris, then you just needed to grab the gcs addresses, or use Kaggle API to get the datasets)
", "votes": 0, "replies": [] } ] }, { "id": 979778, "author_name": "Gilles Vandewiele", "author_url": "", "post_date": "2020-08-21T05:44:07.887000", "content": "Hi there,
\nNot quite the top 100 team here, but we used Google Colab for the entire course of the competition. While it is supposed to be only restricted to residents of the US, none of us actually are from there and we were still able to buy Pro plans.
", "votes": 2, "replies": [ { "id": 979804, "author_name": "RealSid", "author_url": "", "post_date": "2020-08-21T06:02:56.280000", "content": "@group16 , congratulations on your rank ! Thanks for shedding light. I think in the next competition I will try this out.
", "votes": 1, "replies": [] }, { "id": 989332, "author_name": "Marcelo Kittlein", "author_url": "", "post_date": "2020-08-28T18:11:59.607000", "content": "Is there not any problem if they detect that you purchase pro plan from outside the usa? I want to purchase access from Argentina but I am afraid of loosing my gmail accountโฆ
", "votes": 0, "replies": [] } ] }, { "id": 980246, "author_name": "Akshay M Naik", "author_url": "", "post_date": "2020-08-21T12:41:38.803000", "content": "I would love to know about it too.
", "votes": 0, "replies": [] }, { "id": 989084, "author_name": "", "author_url": "", "post_date": "2020-08-28T14:32:14.403000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "983529": "Hi everybody! I exclusively used Kaggle TPUs and GPUs.", "979772": "Hi ! \nIn this competition I used only Kaggle compute resources for my experiments. I dont own a computer with graphic card hence can't train deep learning models locally. I would be up for colab pro but that option is unavailable where I live. \n\nI wanted to know if any of the top teams (top 100) developed models solely using Kaggle compute resources. This will help me set realistic expectations about my standing in future competition. ", "980994": "I exclusively used kaggle TPUs and GPUs. I've only learnt that colab TPU existed by reading this discussion, so thank you for that. Also, my best submission did not have any blending with public notebooks.", "980522": "I used Kaggle TPU mostly. Towards the last week, I went started using colab TPU because it seemed that others were using it. I'm glad I used the colab TPU to run experiments because I hit my TPU quota very early on. I wish I had started using colab TPU earlier. Lesson learned for me.", "980085": "Yes u can still upgrade to Colab pro even outside US\nbut for me it is not very stable and sometime just freeze for no reason (maybe my problem)\n", "981102": "Update: I think I have my answers. The Kaggle competitions are quiet democratic and there are free resources available for people to train their models, compete and win a competition ! Required investment is your own time. Thanks all. ๐๐", "979882": "I used Colab Pro for this competition. It's quite enough to check the majority of ideas and models.", "979796": "We used Google Colab quite extensively. Some of our teammates have Pro accounts, but I managed to train quite a few models with a regular account. The TPU quotas are not explicitly disclosed, but I had at least about 5-6 hours every other day to perform experiments. This was a very nice addition to the Kaggle notebooks.", "979778": "Hi there,\n\nNot quite the top 100 team here, but we used Google Colab for the entire course of the competition. While it is supposed to be only restricted to residents of the US, none of us actually are from there and we were still able to buy Pro plans.\n\n", "980246": "I would love to know about it too.", "989084": "" } }