Re:Show HN: Deta – A cloud platform for building and deployi...
by axlee on Wednesday June 24, 2020 @12:40PM
(#23630283)
> All I want are some blocks that I can glue together into a tensor graph and then deploy it for training and inference.
Something like that? https://peltarion.com/platform
(no relation)
Re:Show HN: Deta – A cloud platform for building and deployi...
by inetsee on Wednesday June 24, 2020 @02:42PM
(#23632116)
I looked at the website and I couldn't find a pricing page anywhere. I tried looking for a site map to see if that would lead to a pricing page, but there's no site map page either.
Re:Show HN: Deta – A cloud platform for building and deployi...
by memexy on Wednesday June 24, 2020 @12:42PM
(#23630325)
This is great. Thanks!
Re:Show HN: Deta – A cloud platform for building and deployi...
by abdelhai on Wednesday June 24, 2020 @12:35PM
(#23630166)
Sorry, this a poorly placed joke :/ No AI planned!
Re:Show HN: Deta – A cloud platform for building and deployi...
by staticassertion on Wednesday June 24, 2020 @12:46PM
(#23630417)
I'd definitely remove it, it does not come off as a joke.
Re:Show HN: Deta – A cloud platform for building and deployi...
by abdelhai on Wednesday June 24, 2020 @12:43PM
(#23630360)
Re: "All I want are some blocks that I can glue together..."
This is a bit over my head, any chance we could discuss this on Slack/email/call? I believe we might be able to help.
Re:Show HN: Deta – A cloud platform for building and deployi...
by memexy on Wednesday June 24, 2020 @12:50PM
(#23630477)
I just meant I want an editor that can help me express dataflow graphs but instead of the usual programming nodes I want to work with AI building blocks like convolutions and batch norm. This page on TensorFlow has some example graphs: https://www.tensorflow.org/tensorboard/graphs.
But I assumed the AI part was serious so if it's not then I think this currently would be out of the product scope so I wouldn't worry about it. Hope that helps.
> DetAI (pretty soon) – No-code neural nets to auto-fix bugs before they happen.
I (and I suspect others as well) don't really need no-code auto-fixes for code. What I want is simple workflows for training WYSIWYG AI models. I'm not interested in writing python to express the model's computation graph. All I want are some blocks that I can glue together into a tensor graph and then deploy it for training and inference.