Comments (2)
The overall proposal looks good to me. I will let @kurman @d4l3k comment on the feasibility and/or detailed planning (in particular, what API definition would work best between torchrun and the plugins).
Minor comment re the UX:
The "plugin dir" concept seems to imply that the users may have multiple plugins, such as plugin_dir/redis
, plugin_dir/mongodb
, etc. In reality, would a single user have multiple plugins? If the user has only one plugin, would the UI below look a little more straightforward:
torchrun --rdzv-backend=/path/to/plugin_name.py ...
Then we can use importlib to dynamically load the module from the source file, e.g.: https://docs.python.org/3/library/importlib.html#importing-a-source-file-directly
from pytorch.
I'm not a huge fan of the PLUGINS_DIRECTORY
environment variable. Can we instead use the existing python entrypoints mechanism that's used by torchx, fsspec, etc?
from pytorch.
Related Issues (20)
- `test_dummy_mha_with_nt_cuda` fails on `sm70`, `sm75` HOT 4
- Dynamo export: unsupported FX node 'aten.expm1.default' HOT 4
- Error: CUDA error: CUDA_ERROR_INVALID_VALUE cuMemcpyDtoH failed with Halide GPU Backend HOT 2
- Dynamo export: Fake tensor broadcast error HOT 2
- UNSTABLE trunk / linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build-test / test (distributed) HOT 2
- UNSTABLE trunk / linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build-test / test (distributed) HOT 1
- UNSTABLE trunk / linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build-test / test (distributed) HOT 1
- How to Convert pytorch qat model to tensorrt HOT 1
- [AOTI] AttributeError: 'ShapeAsConstantBuffer' object has no attribute 'dtype'
- RVV support in PyTorch HOT 2
- Libtorch 2.3.1 requires Glibc_2.28 and can't be used on ubuntu 18. HOT 4
- my accuracy is not increasing HOT 1
- Support for 5-D output image tensors in `Col2Im`. HOT 1
- ChainedScheduler fail on CosineAnnealingWarmRestarts
- OSError: [WinError 126] The specified module could not be found. Error loading "\.venv\Lib\site-packages\torch\lib\fbgemm.dll" or one of its dependencies. HOT 1
- scaled_dot_product_attention fails on Ampere arch with head_dim > 128 HOT 2
- dynamo should recompile when a tensor subclass's inner tensor changes HOT 1
- torch._dynamo.exc.Unsupported: call_method NNModuleVariable() state_dict [] {}
- [custom_op] torch.library.define should be able to auto-infer schema.
- DynamicInt helper structure that is equivalent to mark_dynamic on an int HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from pytorch.