PyTorch

“PyTorch in C++ naturally gets little use, because most people don't want to use C++”

“for the good they did releasing Torch, FastText vectors, and the LLaMa models among others for free”

“you have to install the same package, torch, but from different Python indexes in order to install the cpu version or gpu version, on Linux”

“most have decent support for AMD, it's not that different from CUDA once you get it working”

“when using PyTorch I wanted to try to speed up my NN kernel by 2x by just using half precision and haven't noticed any speedup at all”

“Meta initiated both pytorch and llama which have had profoundly positive effects on AI research”

“It has, along with Python, become something of a linga franca for research in the field.”

“PyTorch is obviously a huge (huge undersells it) contribution that shouldn’t be forgotten.”

“you have to remember to install dependencies like pytorch from an alternate repo”

“Pytorch is widely accepted as the de facto ML framework in both research and industry.”

“The value of PyTorch lies more in utilizing accelerators like GPUs while offering a nice abstraction.”

“It captures the complexity to allow you to make a whole GPT model with it, running on your GPU, and keep the code readable.”

“The pytorch ecosystem and initial push to open weight models, I consider a pretty good thing for the world”

“Meta assigned hundreds of engineers to work PyTorch. That is only going to happen at a handful of companies.”

“a lot of ML frameworks (Tensorflow, PyTorch, etc) by default use the GPU relatively suboptimally”

“Throwing away libraries like ablumentations and pytorch because you move to a different ecosystem is a real loss.”

“Pytorch is such a maddening mess of half implemented research features in a state of Heisen-deprecation”

“To increase your chances of anyone actually clicking on your GH link, use good old Pytorch.”

“low level libraries like pytorch as well as higher level tools like Ollama all seamlessly support multiple GPUs”

“NumPy, PyTorch, JAX have non-trivial overhead, and might even be slower than vanilla Python”

“You can just take Pytorch, JAX, CuPy or whatnot and accelerate your numpy code”

“Facebook was the primary contributor behind Pytorch, which basically set the stage for early GPT implementations.”

“via torch or arrayfire or whatever, you could dispatch it to both CPU and GPU at will”

“If you are doing any serious work with torch or local LLMs, you still run into issues where wheels aren't available”

“Pytorch is fairly efficient, but python as an ecosystem can have some blind spots or bottlenecks”

“A simple kernel in torch is hundreds to thousands of LOCs, with much worse style.”

“magic error telling that tensorflow version they are installing is not compatible with pytorch version”

“where you can add two tensors by hand, or have your whole network as a single nn.Module”

“I ended up sticking with PT and am extremely grateful for all the work put into it”

“It is hard to overstate how much Pytorch has accelerated ML/AI development, across the board.”

“PyTorch felt friendlier and more Pythonic, and it came with a comprehensive library of prebuilt components for deep learning in torch.nn”

“PT has a unique level of broad support that few other open source technology can reach.”

“it has made it possible for me to play with ideas that I would have only dreamed of”

“the PyTorch approach was such a huge improvement that there had been an immediate shift”

“It is a wildly successful project which has acquired ton of code over the years.”

“PyTorch offered a more reasonable level of abstraction along with excellent API documentation and tutorials”

“eager evaluation that made it possible to debug every step in the network forward/backward passes”

“I switched to PyTorch for my next project which was more "batteries included".”

“PyTorch is one of those tools that’s so simple and easy to take apart that you feel like you might’ve been able to make it yourself.”

“Move your performance critical kernels outside Python. Numpy, Pytorch, external databases, Golang microservice, etc.”