nogil
Pytorch
nogil | Pytorch | |
---|---|---|
31 | 340 | |
2,853 | 78,016 | |
- | 1.4% | |
5.7 | 10.0 | |
2 months ago | 4 days ago | |
Python | Python | |
GNU General Public License v3.0 or later | BSD 1-Clause License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
nogil
- Proof-of-Concept Multithreaded Python Without the GIL
-
Our Plan for Python 3.13
This might be a dumb question, but why would removing the GIL break FFI? Is it just that existing no-GIL implementations/proposals have discarded/ignored it, or is there a fundamental requirement, e.g. C programs unavoidably interact directly with the GIL? I know that the C-API is only stable between minor releases [0] compiled in the same manner [1], so it's not like the ecosystem is dependent upon it never changing.
I cannot seem to find much discussion about this. I have found a no-GIL interpreter that works with numpy, scikit, etc. [2][3] so it doesn't seem to be a hard limit. (That said, it was not stated if that particular no-GIL implementation requires specially built versions of C-API libs or if it's a drop-in replacement.)
[0]: https://docs.python.org/3/c-api/stable.html#c-api-stability
[1]: https://docs.python.org/3/c-api/stable.html#platform-conside...
[2]: https://github.com/colesbury/nogil
[3]: https://discuss.python.org/t/pep-703-making-the-global-inter...
-
Real Multithreading Is Coming to Python
https://github.com/colesbury/nogil does manage to get rid of the GIL, but it's not certain to make it into Python core. The main problem is the amount of existing libraries that depend on the existence of the GIL without realizing it - breaking those would be extremely disruptive.
-
[D] The hype around Mojo lang
CPython is also investigating the removal of the GIL (PEP703, nogil). I think requiring the GIL is a wider thing that libraries will need to address anyway. But also, for the same reason as above I'd be surprised if the Modular team thought that saying "you can run all your python code unchanged" was a good idea if there was a secret "except for code that uses numpy" muttered under the breath.
- PEP 684 was accepted – Per-interpreter GIL in Python 3.12
- PEP 703 – Making the Global Interpreter Lock Optional in CPython
-
Python 3.11.0 final is now available
I'm worried about the speedup
My understanding is that it's based on the most recent attempt to remove the GIL by Sam Gross
https://github.com/colesbury/nogil
In addition to some ways to try to not have nogil have as much overhead he added a lot of unrelated speed improvements so that python without the gil would still be faster not slower in single thread mode. They seem to have merged those performance patches first that means if they add his Gil removal patches in say python 3.12 it will still be substantially slower then 3.11 although faster then 3.10. I hope that doesn't stop them from removing the gil (at least by default)
- Removed the GIL back in 1996 from Python 1.4, primarily to create a re-entrant Python interpreter.
- I Tried Removing Python's GIL Back in 1996
-
Faster CPython 3.12 Plan
Looks like it's still active to me:
https://github.com/colesbury/nogil/
Pytorch
-
Clasificador de imágenes con una red neuronal convolucional (CNN)
PyTorch (https://pytorch.org/)
-
AI enthusiasm #9 - A multilingual chatbot📣🈸
torch is a package to manage tensors and dynamic neural networks in python (GitHub)
-
Einsum in 40 Lines of Python
PyTorch also has some support for them, but it's quite incomplete and has many issues so that it is basically unusable. And its future development is also unclear. https://github.com/pytorch/pytorch/issues/60832
-
Library for Machine learning and quantum computing
TensorFlow
-
My Favorite DevTools to Build AI/ML Applications!
TensorFlow, developed by Google, and PyTorch, developed by Facebook, are two of the most popular frameworks for building and training complex machine learning models. TensorFlow is known for its flexibility and robust scalability, making it suitable for both research prototypes and production deployments. PyTorch is praised for its ease of use, simplicity, and dynamic computational graph that allows for more intuitive coding of complex AI models. Both frameworks support a wide range of AI models, from simple linear regression to complex deep neural networks.
-
penzai: JAX research toolkit for building, editing, and visualizing neural nets
> does PyTorch have a similar concept
of course https://github.com/pytorch/pytorch/blob/main/torch/utils/_py...
-
Tinygrad: Hacked 4090 driver to enable P2P
fyi should work on most 40xx[1]
[1] https://github.com/pytorch/pytorch/issues/119638#issuecommen...
-
The Elements of Differentiable Programming
Sure, right here: https://github.com/pytorch/pytorch/blob/main/torch/autograd/...
Here's the documentation: https://pytorch.org/tutorials/intermediate/forward_ad_usage....
> When an input, which we call “primal”, is associated with a “direction” tensor, which we call “tangent”, the resultant new tensor object is called a “dual tensor” for its connection to dual numbers[0].
-
Functions and operators for Dot and Matrix multiplication and Element-wise calculation in PyTorch
*My post explains Dot, Matrix and Element-wise multiplication in PyTorch.
-
Dot vs Matrix vs Element-wise multiplication in PyTorch
In PyTorch with @, dot() or matmul():
What are some alternatives?
hpy - HPy: a better API for Python
Flux.jl - Relax! Flux is the ML library that doesn't make you tensor
mypyc - Compile type annotated Python to fast C extensions
mediapipe - Cross-platform, customizable ML solutions for live and streaming media.
numpy - The fundamental package for scientific computing with Python.
Apache Spark - Apache Spark - A unified analytics engine for large-scale data processing
python-feedstock - A conda-smithy repository for python.
flax - Flax is a neural network library for JAX that is designed for flexibility.
sbcl - Mirror of Steel Bank Common Lisp (SBCL)'s official repository
tinygrad - You like pytorch? You like micrograd? You love tinygrad! ❤️ [Moved to: https://github.com/tinygrad/tinygrad]
cosmopolitan - build-once run-anywhere c library
Pandas - Flexible and powerful data analysis / manipulation library for Python, providing labeled data structures similar to R data.frame objects, statistical functions, and much more