Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

support for the CUDA platform of NVIDIA graphics cards #209

Open
Ing-Brayan-Martinez opened this issue Apr 8, 2024 · 2 comments
Open

support for the CUDA platform of NVIDIA graphics cards #209

Ing-Brayan-Martinez opened this issue Apr 8, 2024 · 2 comments

Comments

@Ing-Brayan-Martinez
Copy link

Greetings community, I am a developer of the Dart language. Lately I have been making quite controversial proposals since I am very concerned about the future of this language. My main observation is that the Dart language has already reached a fairly important level of maturity. The problem with this language is the lack of frameworks and its main difficulty is developing the ecosystem

According to the feedback I have received, many of the things I have raised are not possible because the team that maintains the Dart language has no control because at Google things are handled in separate teams, which is why I am here requesting this.

The idea is to extend the functionalities of the Tensorflow library written in Dart to turn it into a complete framework. For this, support for CUDA cores is important because a developer of the Dart language could create generative AI models without the need to learn another language. The idea is to do all the development in Dart. Currently I think that Python has a monopoly in this area. I talked a little about this in the following proposal.

I don't know if it's too much to ask for a collaboration between the team that maintains Tensorflow and the team that maintains the Dart language to extend the functionalities of Tensorflow and turn it into the second most important framework, since the Dart language has Flutter, we really want to invest in this language but Sometimes we don't understand why the Google teams don't agree and work like united brothers to promote the programming languages made internally, the Go language ran with a little more luck.

I really think that the Dart language should have more attention since it is an internal language, for example Microsoft has done this with C# giving full compatibility to all the frameworks made in Microsoft, the same thing Apple does with Swift, that's why I ask for the support of the Tensorflow team to promote this framework on the Dart platform, many of us trust this technology and we want Google to develop it better by offering a variety of frameworks to find greater reasons to invest

@KernelInterrupt
Copy link

I think that maybe you can do this by using dart:ffi to create a wrapper for tensorflow cpp API,but it’s hard to do so because you have to deal with NVCC and you must write some scripts to automatically configure environment for different users to make it easy to use.
btw,I don’t think it’s necessary to use Dart to replace Python because Dart doesn’t have advantages over Python in terms of machine learning.Take me as an example:I’m currently reimplementing PyTorch high-level APIs by using dart:ffi to create wrappers for libtorch,and I found that sometimes it’s impossible to reimplement some operations which make PyTorch easy to use as Dart doesn’t have some magic functions(for example,radd).Also,I think sometimes directly train neural networks in Dart is not so important because you can train them in python(by using Tensorflow and PyTorch)and then output graphs,after that you can do inference in Dart easily.
In addition,I think that what Dart really need is a numpy-like tensor library.That’s why I’m reimplementing PyTorch high-level APIs in Dart.because tensor library for PyTorch is just like numpy.

@Ing-Brayan-Martinez
Copy link
Author

@KernelInterrupt excellent job, I'm going to learn Dart ffi and I myself will start creating my own tools I hope I can at least receive some support from Google to guarantee the success of the project I want an ecosystem made in Dart Google has everything to do it the idea is that Dart can Being the main language of any developer that is an economically profitable language, there is no point in investing time and money in learning this technology if you cannot monetize it in the labor market with any type of project, not only Apps, but any type of project is what it offers. language like C# or Swfit, those who invest in these languages always have several options at the framework level to participate in any project always using the main language that was chosen

All of us who are dedicated to using Dart and Flutter currently want this so as not to depend on other technological stacks, for example, some javascript or java framework, we want a 360 view of the Dart language, which is why it is necessary that the Google cloud offers support and go step by step gaining ground to see the growth of the platform, what they are doing is very good, although they must now look towards the back end and AI to develop the language

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants