I plan to use this competition as my class project. I see the datasets are very large. What machine(GPU, CPU, Memory) required for this project if I want to train a model using training sets, and predict result in test set? I have Macbook Pro 2018 16GB. Should I looking into using virtual machine like GCP VM? The training speed is very important so that I can study the results, re-tune the models, learn how different models works. Thanks for the help.
", "messages": [ { "id": "552423", "postDate": "06/13/2019 23:59:33", "content": "I plan to use this competition as my class project. I see the datasets are very large. What machine(GPU, CPU, Memory) required for this project if I want to train a model using training sets, and predict result in test set? I have Macbook Pro 2018 16GB. Should I looking into using virtual machine like GCP VM? The training speed is very important so that I can study the results, re-tune the models, learn how different models works. Thanks for the help.
", "rawMarkdown": "I plan to use this competition as my class project. I see the datasets are very large. What machine(GPU, CPU, Memory) required for this project if I want to train a model using training sets, and predict result in test set? I have Macbook Pro 2018 16GB. Should I looking into using virtual machine like GCP VM? The training speed is very important so that I can study the results, re-tune the models, learn how different models works. Thanks for the help.", "votes": null }, { "id": "552471", "postDate": "06/14/2019 02:46:18", "content": "The same question
", "rawMarkdown": "The same question", "votes": null }, { "id": "552509", "postDate": "06/14/2019 04:23:52", "content": "Yes. You would definitely require a GPU for such a competition. And as you said that the training speed is very important for you, I would suggest opting for a VM on the cloud like GCP/AWS, etc which give you free credits for initial signup to try out their services. Hope this helps
", "rawMarkdown": "Yes. You would definitely require a GPU for such a competition. And as you said that the training speed is very important for you, I would suggest opting for a VM on the cloud like GCP/AWS, etc which give you free credits for initial signup to try out their services. Hope this helps", "votes": null }, { "id": "557674", "postDate": "06/21/2019 11:31:45", "content": "You absolutely need to get the biggest GPU e.g V100 to run on cloud. Everything is huge in this competition, the dataset, the image size, even inferencing on the test dataset is taking me over 3 hours (it is still running as I write this). I would suggest beginner to work on other easier dataset first and only come back to this when you have more confidence and experience.
", "rawMarkdown": "You absolutely need to get the biggest GPU e.g V100 to run on cloud. Everything is huge in this competition, the dataset, the image size, even inferencing on the test dataset is taking me over 3 hours (it is still running as I write this). I would suggest beginner to work on other easier dataset first and only come back to this when you have more confidence and experience.", "votes": null }, { "id": "558327", "postDate": "06/22/2019 06:29:32", "content": "You can always resize the images to make them fit in your memory. For instance I am resizing to 512
", "rawMarkdown": "You can always resize the images to make them fit in your memory. For instance I am resizing to 512", "votes": null } ], "comments": [ { "id": 552471, "author_name": "hesene", "author_url": "", "post_date": "06/14/2019 02:46:18", "content": "The same question
", "votes": null, "replies": [] }, { "id": 552509, "author_name": "chandanverma", "author_url": "", "post_date": "06/14/2019 04:23:52", "content": "Yes. You would definitely require a GPU for such a competition. And as you said that the training speed is very important for you, I would suggest opting for a VM on the cloud like GCP/AWS, etc which give you free credits for initial signup to try out their services. Hope this helps
", "votes": null, "replies": [] }, { "id": 557674, "author_name": "soonyau", "author_url": "", "post_date": "06/21/2019 11:31:45", "content": "You absolutely need to get the biggest GPU e.g V100 to run on cloud. Everything is huge in this competition, the dataset, the image size, even inferencing on the test dataset is taking me over 3 hours (it is still running as I write this). I would suggest beginner to work on other easier dataset first and only come back to this when you have more confidence and experience.
", "votes": null, "replies": [ { "id": 558327, "author_name": "stecasasso", "author_url": "", "post_date": "06/22/2019 06:29:32", "content": "You can always resize the images to make them fit in your memory. For instance I am resizing to 512
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "552423": "I plan to use this competition as my class project. I see the datasets are very large. What machine(GPU, CPU, Memory) required for this project if I want to train a model using training sets, and predict result in test set? I have Macbook Pro 2018 16GB. Should I looking into using virtual machine like GCP VM? The training speed is very important so that I can study the results, re-tune the models, learn how different models works. Thanks for the help.", "552471": "The same question", "552509": "Yes. You would definitely require a GPU for such a competition. And as you said that the training speed is very important for you, I would suggest opting for a VM on the cloud like GCP/AWS, etc which give you free credits for initial signup to try out their services. Hope this helps", "557674": "You absolutely need to get the biggest GPU e.g V100 to run on cloud. Everything is huge in this competition, the dataset, the image size, even inferencing on the test dataset is taking me over 3 hours (it is still running as I write this). I would suggest beginner to work on other easier dataset first and only come back to this when you have more confidence and experience.", "558327": "You can always resize the images to make them fit in your memory. For instance I am resizing to 512" }, "source": "meta" }