Add subgraph optimization error detection tool
Open
TorchBlade
- Dominant language
- C++
- Stars
- 933
- Forks
- 169
- PR merge metrics
- No merged PRs in 30d
Description
It's hard to debug the root cause of the precision errors of the model/graph outputs after BladeDISC optimization. One way to nail down the debug region is to truncate the model and try to do BladeDISC optimization until there is no precision difference. Once the truncate phase terminated, we should have found the operation from which the BladeDISC optimization would introduce error results.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.