Flashlight: Fast and flexible machine learning in C++ • Hacker News Board

Re: Flashlight: Fast and flexible machine learning in C++

#31

Post by snovv_crash » Fri, Apr 16, 2021, 8:58 PM UTC

angry_octet wrote: post #9

Narew wrote: post #5

Earlier quoted context omitted.

Flashlight vs torch (pytorch). ^^

Yes, I could see the construction, I just can't think why they wouldn't anticipate the problems it could cause down the track.

You spend too much time on the internet.

Re: Flashlight: Fast and flexible machine learning in C++

#32

Post by dynamite-ready » Fri, Apr 16, 2021, 9:03 PM UTC

Does anyone else think that C++ makes more sense for ML work than Python? I'd been thinking so for years. Both for deployment/performance and data wrangling purposes.

Re: Flashlight: Fast and flexible machine learning in C++

#33

Post by coliveira » Fri, Apr 16, 2021, 9:03 PM UTC

dwrodri wrote: post #21

This seems really cool, but I don't get why they would pour work into this while simultaneously building a C++ front-end for PyTorch[1]. Per the blog post, both frameworks have the goal of empowering ML researchers to iterate on ML models in such a way that it becomes easier to reason about performance than it would be talking to a bunch of dynamically linked object files behind an interpreter. Facebook is a huge com…

Have you ever thought that it is not good for a big company like FB to be dependent on a single language/technology? This is the reason why these companies will invest in multiple solutions in different languages.

Re: Flashlight: Fast and flexible machine learning in C++

#34

Post by whimsicalism » Fri, Apr 16, 2021, 9:05 PM UTC

dynamite-ready wrote: post #32

Does anyone else think that C++ makes more sense for ML work than Python? I'd been thinking so for years. Both for deployment/performance and data wrangling purposes.

In general, I think languages with static typing are preferable.

C++ seems ideal for me right now because it is the only other language with a somewhat mature stack (perhaps Julia as well, but I haven't played too much around with that).

Re: Flashlight: Fast and flexible machine learning in C++

#35

Post by suyash » Fri, Apr 16, 2021, 9:08 PM UTC

Wonder why they didn't leverage TorchScript ? However this is a welcome news in ML Community. Would love to see this library break away from NVIDIA CUDA dependency and optimize for Apple M1 chips

Re: Flashlight: Fast and flexible machine learning in C++

#36

Post by ipsum2 » Fri, Apr 16, 2021, 9:09 PM UTC

dwrodri wrote: post #25

whimsicalism wrote: post #23

Earlier quoted context omitted.

Flashlight is much lower level and gives more fine-grained performance control. For instance, I don't think there is really any way to do real-time speech recognition that is fast with PyTorch because of how it is architected. From my understanding, Tensor Comprehensions and Halide are both very tentative research projects. > not having it's direction beholden to the whims of Alphabet (see: Swift for Tensorflow). I d…

This is a very interesting claim. I find it credible because it stands to reason that projects like DeepSpeed[1] and TorchScript[2] wouldn't need to exist if inference performance of research PyTorch models was satisfactory for production, but often case it isn't. It appears as though Flashlight is built on ArrayFire. I haven't seen how gradients are managed in arrayfire-ml, but perhaps it is the case that the autogr…

PyTorch is based off of Torch, which was first released in 2002, predating TensorFlow by 13 years.

https://en.wikipedia.org/wiki/Torch_(machine_learning)

Re: Flashlight: Fast and flexible machine learning in C++

#37

Post by suyash » Fri, Apr 16, 2021, 9:10 PM UTC

whimsicalism wrote: post #14

I wish Flashlight had a way to easily export models to other frameworks.

that can be done is they support ONNX

Re: Flashlight: Fast and flexible machine learning in C++

#38

Post by tmotwu » Fri, Apr 16, 2021, 9:12 PM UTC

kyawzazaw wrote: post #18

flashlight is such a common word used for an appliance. I would never think of anything else upon hearing this. Millions of people I assume as well.

So is Torch? A much more common word to refer to a flashlight.

Re: Flashlight: Fast and flexible machine learning in C++

#39

Post by liuliu » Fri, Apr 16, 2021, 9:13 PM UTC

dwrodri wrote: post #25

whimsicalism wrote: post #23

Earlier quoted context omitted.

Flashlight is much lower level and gives more fine-grained performance control. For instance, I don't think there is really any way to do real-time speech recognition that is fast with PyTorch because of how it is architected. From my understanding, Tensor Comprehensions and Halide are both very tentative research projects. > not having it's direction beholden to the whims of Alphabet (see: Swift for Tensorflow). I d…

This is a very interesting claim. I find it credible because it stands to reason that projects like DeepSpeed[1] and TorchScript[2] wouldn't need to exist if inference performance of research PyTorch models was satisfactory for production, but often case it isn't. It appears as though Flashlight is built on ArrayFire. I haven't seen how gradients are managed in arrayfire-ml, but perhaps it is the case that the autogr…

Torch (Lua) predates TensorFlow and is Lecun's pet project for a few years at that point already. But Lua as a language is unpopular at a time. PyTorch would be a welcome addition then. But even if no PyTorch (nor Caffe2) in an alternative timeline, I would imagine FB would be stuck with Lua Torch for quite some time.

Re: Flashlight: Fast and flexible machine learning in C++

#40

Post by wdpk » Fri, Apr 16, 2021, 9:20 PM UTC

Anyone has experience with the JIT in arrayfire? https://arrayfire.com/performance-improvements-to-jit-in-arr...

Many interesting features in this framework, autograd looks neat too.

I guess the deployment in real world C++ apps will be easier than PyTorch or Tensorflow, especially at the edge in scenarios with little or no network access.