Would you recommend this product?
1 Review5.0/5
Hi Hunters, I’m Sai, one of the cofounders at FloydHub. We're building FloydHub to be a “Heroku for deep learning”. We are in the current batch (W17) at YC. Thanks to @chrismessina for hunting us! 10 months ago, I was working at Microsoft and doing a lot of deep learning (DL) there. While the DL community is terrific, I was often frustrated by how difficult it was to get started and build upon others’ work. For example, running any popular Github project often started with an exercise in dependency hell. As I untangled these for myself, I wrote up some notes on setting up the top DL frameworks, which became super popular on Github (https://github.com/saiprashanths...). That's when I realized that engineering was a huge bottleneck in deep learning and a problem worth solving after all. I’ve since quit my job and have been working fulltime for the last 9 months on building FloydHub to make deep learning easier. Our goal is to let the data scientists focus on the science, while we handle the engineering grunt work (provisioning and scaling infra, running reproducible experiments, enabling sharing and collaboration, supporting DL frameworks with zero setup, shipping trained models to production easily, etc.) Lots of interesting challenges - happy to talk about them! We have a lot of work ahead, but we’re excited to share with you what we have so far! We have a neat Neural Style Transfer demo you can try out now (https://www.floydhub.com/#examples). Let us know what you think! Looking forward to your feedback and answering questions.
Upvote (23)Share
@saip @chrismessina any special deals for poor students? :)
@gopietz Not yet, but we've been thinking about setting up an academic discount. I'd like to learn more - shoot us an email: founders@floydhub.com
This is pretty awesome... the whole field of Deep Learning is becoming more and more accessible and easy to use. Here's the launch post on HN.
Upvote (11)Share
This is great @saip @narenst! Was trying to figure out what GPUs you guys are using (seems like K80). Definitely on par with all the cloud providers right now (GCP, AWS, Azure) but I'm frustrated that noone is offering the GTX 1080 or the new Titan X. Both of those cards are quite a bit faster than the K80s for Deep Learning. Would love it if you guys could incorporate one of them somehow :).
@samiur1204 @narenst Agreed! We're currently on AWS, so we're limited to Tesla K80s at the moment, unfortunately. But better GPUs and hosting our own infra is something we have been talking quite a bit recently. I'll keep you posted on that front!
Hello Hunters! I'm Naren, the other co-founder. We are very excited to see all the feedback so far, please keep them coming! One question for all: What Deep Learning frameworks / algorithms do you want to see on Floyd? Maybe something that you already use or always wanted to try but it has been difficult to setup / get started. And we will port them to Floyd :)
@narenst Keras, Chainer and PyTorch are what I use. Since Keras is part of TensorFlow now, I'm sure that's already supported, but Chainer and PyTorch handle dynamic networks much better.
@samiur1204 @narenst Feedback taken. Chainer and PyTorch coming up very shortly!
So I'm probably a bit sensitive when it comes to Heroku comparisons 😉 But... Nice DX, what appears at this stage to be pretty compelling pricing, and a marketplace in the works too?! 😱 Wow. This is shaping up to be something very impressive. Great job team. I'm not really in the target market as I'm fine running infrastructure but a massive n00b when it comes to tensorflow. I really want to up my TF game so I can take this for a serious spin though.
@glenngillen Thanks! Driving down price is something we're serious about and that's a good part of our tech challenge. GPUs on the cloud are ridiculously expensive and a big barrier for people to experiment with DL. Do take it for a spin and let us know what you think! We definitely want to make it easy to get started, even for beginners and enthusiasts like yourself. We have one full end-to-end tutorial (style transfer), but hoping to add more over the coming days - may be that's where the marketplace will help :)