Lightning-AI / Lightning-AI/lightning-thunder
Dyanmic Shape TensorProxy
@jjsjann123 is already working on this.
Since Aug 23, 2024.
- Dominant language
- Python
- Stars
- 1.5k
- Forks
- 121
- PR merge metrics
- No merged PRs in 30d
Description
## 🚀 Feature
This is a continuation of the work in #262 . The phase 2 [proposal](https://docs.google.com/document/d/1jRizLZawjJptN_Qq7vHnqwK9vhlYc_woAwauGQQcqA8/edit#heading=h.yx2ge23f1e7) targets the re-use of compiled thunder program by supporting TensorProxy with dynamic shape and adjusting caching policy accordingly.
This is an umbrella issue to track progress. I would like to leave detailed technical discussion in separate issues and link them. This section will be gradually populated as we start picking up speed:
- shape propagation modeling in thunder
- [x] Enabling NumberProxy in TensorProxy.shape (toy PR #1027)
- [x] Adding shape prim (PR #1113)
- [x] Issue that blocks the #1182 . Should be resolved in #1201
- [ ] shape logic in prim needs to be modeled in trace Issue: #471
- [x] modeling shape queries: #1253 . Will need two PR to resolve, PR 1 #1260
- Static constraints and prologue checks
- [ ] #1133 We are ready to push towards this issue now. Since current shape queries ended up with duplicated shape access. We would want to clean it up.
- Handling of existing shape checks
- [ ] broadcast semantic handling in dynamic shape is broken in symbolic values cache. existing contraints doesn't work on NumberProxy in TensorProxy.shape. Issue opened #1525
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.