Comments (1)
Marking as needs design, since we haven't reached a conclusion yet on the right API. Some discussion from triage review:
alban: We could add an out= variant for matmul, would need extra work in autograd to handle it. Also, "have a performant kernel" is a separate step from "expose an API for this".
Richard: there should not be a separate way to do this in eager vs compile (compile has a custom dtype HOP)
nikita: we have a mixed_mm already? can maybe better advertise that this is the right API to use, there have been other similar issues, still needs design. Also from nikita: can support mixed dtypes in compile, but blas requires us to upcast for cpu (no performant kernel)
from pytorch.
Related Issues (20)
- [TensorDict - compile] Dynamo: set(arbitrary_iterable_object)` does not work but `{k for k in arbitrary_iterable_object}` does
- [TensorDict - compile] compile seems to struggle with `key in dict`
- [TensorDict - compile] Custom `with stuff(param)` support in dynamo
- [TensorDict - compile] `DictGuardManager does not support a leaf_guard` raised during execution of functionalized module
- TensorDict - torch.compile issue tracker
- Libtorch does not support SM90 HOT 2
- Distributed training with NCCL hangs on init_process_group when debug is DETAIL HOT 2
- torch compile error with `torch.Tensor.unsqueeze_`
- `max_entries` parameter of `torch.cuda.memory._record_memory_history()`
- float8 tensor subclass + export issues HOT 3
- `KeyError: 'ac_graph_id'` when using compile+SAC+reparameterization with `inline_inbuilt_nn_modules=False`
- [torch.compile] Tried to erase Node div but it still had 3 users in the graph HOT 3
- [torch.export] Simple model fails to export with dynamic shapes HOT 3
- DISABLED test_tracker_non_root_forward_backward (__main__.TestTrackerFullyShard1DTrainingCore) HOT 4
- Pytorch nightly aarch64 build is failing HOT 1
- Investigate getting all globals that need to be added to `torch.serialization.add_safe_globals` in one call rather than iteratively
- UNSTABLE pull / before-test / llm-retrieval HOT 2
- UNSTABLE trunk / before-test / llm-retrieval HOT 4
- aot_compile unintuitive error with make_fx graphs HOT 3
- Torch Distributed Pipelinging API datatype mismatch HOT 5
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.