multipy
multipy copied to clipboard
pytorch
torch::deploy (multipy for non-torch uses) is a system that lets you get around the GIL problem by running multiple Python interpreters in a single C++ process.
This issue can be reproduced using docker with https://github.com/pytorch/multipy/pull/233, otherwise, one a) needs to be running the unit test on a cuda enabled machine + on python 3.7/3.8. This error...
Summary: Add preprocessor flag to unit test for importlib.metadata; the latter was introduced in python3.8. Test Plan: Reviewers: Subscribers: Tasks: Tags:
Summary: As title. Test Plan: Reviewers: Subscribers: Tasks: Tags:
Cpython has had some changes from v3.10 onwards, for example https://fburl.com/diffing/lrcq4gej . We want to add some conditional inclusions to multipy cmake to support these changes. Step 1/n : Replace...
Currently deploy only supports ELFs for the custom linker logic. It would be great to support Mach-O so we can use multiple in mac environments for use during training/dataloading as...
Add support for pip installing pybind bindings Prototype code at https://github.com/pytorch/multipy/tree/pybind
Would be nice to have a runtime/deploy + dynamo test to ensure they work seamlessly out of the box
Right now we depend on gcc ld for packing the interpreter into the library -- this doesn't work when using lld so it would be good for us to rework...
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #265 I ran into a few bugs when trying to use deploy + dynamo. This PR adds some torchdynamo unit tests to...
### 📚 The doc issue Hi, I'm trying the tutorial example of deploy and aim to package a model and do inference in C++. But I ran into a problems...
