Hello friends.
\nDo you know good alternatives (cost/benefit) to Google Colab? I have Pro+ but every day I'm maxing out the limit they give (they force to finish the training of my models in the middle, which makes it a little difficult).
\nI saw the GCP prices, but the amount of experiments that are in my queue would be prohibitive $$$
", "messages": [ { "id": 1652672, "postDate": "2022-01-16T21:53:09.100Z", "content": "Hello friends.
\nDo you know good alternatives (cost/benefit) to Google Colab? I have Pro+ but every day I'm maxing out the limit they give (they force to finish the training of my models in the middle, which makes it a little difficult).
\nI saw the GCP prices, but the amount of experiments that are in my queue would be prohibitive $$$
", "rawMarkdown": "Hello friends. \n\nDo you know good alternatives (cost/benefit) to Google Colab? I have Pro+ but every day I'm maxing out the limit they give (they force to finish the training of my models in the middle, which makes it a little difficult).\n\nI saw the GCP prices, but the amount of experiments that are in my queue would be prohibitive $$$", "votes": 6 }, { "id": 1652704, "postDate": "2022-01-16T22:41:55.587Z", "content": "I was thinking about Azure but …. price is mind blowing … I asked Microsoft about support and they gave me info that could help but in the last days of competition (renting machine with 2xGPUs is expensive).
\nSince we have GPU war :) I was looking for solution but …. I can not afford HP Z (it is my dream) or Nvidia DGX …. so I still use Colab … and trying to find …. solution for the poor 😂😂😂😭😜
", "rawMarkdown": "I was thinking about Azure but .... price is mind blowing ... I asked Microsoft about support and they gave me info that could help but in the last days of competition (renting machine with 2xGPUs is expensive).\n\nSince we have GPU war :) I was looking for solution but .... I can not afford HP Z (it is my dream) or Nvidia DGX .... so I still use Colab ... and trying to find .... solution for the poor 😂😂😂😭😜", "votes": 5, "replies": [ { "id": 1652705, "postDate": "2022-01-16T22:48:26.670Z", "content": "We are poor wanting to participate in the rich man's game 😂😂😂😂😂😂
", "rawMarkdown": "We are poor wanting to participate in the rich man's game 😂😂😂😂😂😂", "votes": 6 }, { "id": 1652706, "postDate": "2022-01-16T22:51:04.607Z", "content": "😄😄😄😄 sometimes limitations bring fantastic solutions :)
\nYou asked very good question. Waiting for solutions.
Batch Size = 0 😬😬😬😂😂😂
", "rawMarkdown": "Batch Size = 0 😬😬😬😂😂😂", "votes": 2 }, { "id": 1653668, "postDate": "2022-01-17T19:18:56.243Z", "content": "Batch Size = 0 could be great as well in this case. Probably you can randomly generate bboxes and achieve better score then training most of NN 😂😂🤣🤣
", "rawMarkdown": "Batch Size = 0 could be great as well in this case. Probably you can randomly generate bboxes and achieve better score then training most of NN 😂😂🤣🤣", "votes": 2 }, { "id": 1653674, "postDate": "2022-01-17T19:25:47.907Z", "content": "Trueee 😂😂😂😂😂😂😂
", "rawMarkdown": "Trueee 😂😂😂😂😂😂😂", "votes": 1 }, { "id": 1668842, "postDate": "2022-01-30T05:19:29.320Z", "content": "Hey @remakkinas , can I ask how you are storing your data? standard gdrive space is very small so I'm considering the 100gb plan for storage
", "rawMarkdown": "Hey @remakkinas , can I ask how you are storing your data? standard gdrive space is very small so I'm considering the 100gb plan for storage" }, { "id": 1668934, "postDate": "2022-01-30T07:21:13.643Z", "content": "I can provide 3090
", "rawMarkdown": "I can provide 3090" }, { "id": 1668936, "postDate": "2022-01-30T07:23:10.683Z", "content": "@kennyxie gDrive 200GB plan
", "rawMarkdown": "@kennyxie gDrive 200GB plan", "votes": 1 }, { "id": 1669027, "postDate": "2022-01-30T09:24:00.640Z", "content": "i am using 100gb storage plan, its good to store all models while training
", "rawMarkdown": "i am using 100gb storage plan, its good to store all models while training" } ] }, { "id": 1652769, "postDate": "2022-01-17T01:11:13.217Z", "content": "Hello @robsonsan, I don't think there is anything else price effective as Colab Pro+. Another option will be GCP, AWS, or AZURE, but it can get costly, especially with distributed training. Even if you can afford your DGX system, you will have to use it so much to justify the investment; probably, I will buy a pair of GTX3090TI if I have the money but under current conditions and scalpers taking all of them and reselling is impossible. :(
", "rawMarkdown": "Hello @robsonsan, I don't think there is anything else price effective as Colab Pro+. Another option will be GCP, AWS, or AZURE, but it can get costly, especially with distributed training. Even if you can afford your DGX system, you will have to use it so much to justify the investment; probably, I will buy a pair of GTX3090TI if I have the money but under current conditions and scalpers taking all of them and reselling is impossible. :(", "votes": 2 }, { "id": 1653037, "postDate": "2022-01-17T07:25:12.800Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1652704, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-16T22:41:55.587000", "content": "I was thinking about Azure but …. price is mind blowing … I asked Microsoft about support and they gave me info that could help but in the last days of competition (renting machine with 2xGPUs is expensive).
\nSince we have GPU war :) I was looking for solution but …. I can not afford HP Z (it is my dream) or Nvidia DGX …. so I still use Colab … and trying to find …. solution for the poor 😂😂😂😭😜
", "votes": 5, "replies": [ { "id": 1652705, "author_name": "Robson", "author_url": "", "post_date": "2022-01-16T22:48:26.670000", "content": "We are poor wanting to participate in the rich man's game 😂😂😂😂😂😂
", "votes": 6, "replies": [] }, { "id": 1652706, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-16T22:51:04.607000", "content": "😄😄😄😄 sometimes limitations bring fantastic solutions :)
\nYou asked very good question. Waiting for solutions.
Batch Size = 0 😬😬😬😂😂😂
", "votes": 2, "replies": [] }, { "id": 1653668, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-17T19:18:56.243000", "content": "Batch Size = 0 could be great as well in this case. Probably you can randomly generate bboxes and achieve better score then training most of NN 😂😂🤣🤣
", "votes": 2, "replies": [] }, { "id": 1653674, "author_name": "Shashwat Raman", "author_url": "", "post_date": "2022-01-17T19:25:47.907000", "content": "Trueee 😂😂😂😂😂😂😂
", "votes": 1, "replies": [] }, { "id": 1668842, "author_name": "aquaright", "author_url": "", "post_date": "2022-01-30T05:19:29.320000", "content": "Hey @remakkinas , can I ask how you are storing your data? standard gdrive space is very small so I'm considering the 100gb plan for storage
", "votes": 0, "replies": [] }, { "id": 1668934, "author_name": "zhiqiang he", "author_url": "", "post_date": "2022-01-30T07:21:13.643000", "content": "I can provide 3090
", "votes": 0, "replies": [] }, { "id": 1668936, "author_name": "Remek Kinas", "author_url": "", "post_date": "2022-01-30T07:23:10.683000", "content": "@kennyxie gDrive 200GB plan
", "votes": 1, "replies": [] }, { "id": 1669027, "author_name": "SeshuRaju 🧘♂️", "author_url": "", "post_date": "2022-01-30T09:24:00.640000", "content": "i am using 100gb storage plan, its good to store all models while training
", "votes": 0, "replies": [] } ] }, { "id": 1652769, "author_name": "C4rl05/V", "author_url": "", "post_date": "2022-01-17T01:11:13.217000", "content": "Hello @robsonsan, I don't think there is anything else price effective as Colab Pro+. Another option will be GCP, AWS, or AZURE, but it can get costly, especially with distributed training. Even if you can afford your DGX system, you will have to use it so much to justify the investment; probably, I will buy a pair of GTX3090TI if I have the money but under current conditions and scalpers taking all of them and reselling is impossible. :(
", "votes": 2, "replies": [] }, { "id": 1653037, "author_name": "", "author_url": "", "post_date": "2022-01-17T07:25:12.800000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1652672": "Hello friends. \n\nDo you know good alternatives (cost/benefit) to Google Colab? I have Pro+ but every day I'm maxing out the limit they give (they force to finish the training of my models in the middle, which makes it a little difficult).\n\nI saw the GCP prices, but the amount of experiments that are in my queue would be prohibitive $$$", "1652704": "I was thinking about Azure but .... price is mind blowing ... I asked Microsoft about support and they gave me info that could help but in the last days of competition (renting machine with 2xGPUs is expensive).\n\nSince we have GPU war :) I was looking for solution but .... I can not afford HP Z (it is my dream) or Nvidia DGX .... so I still use Colab ... and trying to find .... solution for the poor 😂😂😂😭😜", "1652769": "Hello @robsonsan, I don't think there is anything else price effective as Colab Pro+. Another option will be GCP, AWS, or AZURE, but it can get costly, especially with distributed training. Even if you can afford your DGX system, you will have to use it so much to justify the investment; probably, I will buy a pair of GTX3090TI if I have the money but under current conditions and scalpers taking all of them and reselling is impossible. :(", "1653037": "" } }