Enable configuration/tags to turn off data introspection in SDK
- Dominant language
- Jupyter Notebook
- Stars
- 2.6k
- Forks
- 213
- PR merge metrics
- No merged PRs in 30d
Description
**Is your feature request related to a problem? Please describe.**
Currently the SDK introspects all outputs. Sometimes, especially in the case of Parallel, this could be a really long list of things; it'd be nice to turn it off selectively.
**Describe the solution you'd like**
Option 1: inspect tag to skip
```python
@tag(sdk=["skip.introspection"])
def my_func(...) -> ...:
```
Option 2: skip via configuration
```python
sdk_config = {"my_func": ["skip.introspection"]}
HamiltonTracker(..., tracker_config=sdk_config) # or some config file
```
---
We'd then want these to both output some type of result that the UI could then render.
Minimally based on the type returned it could be something basic about it -- so when you got to the UI you can see that this was skipped.
**Describe alternatives you've considered**
N/A
**Additional context**
This is an optimization.
Functionality to hide sensitive information from being logged could follow a similar path -- but likely a more explicit decorator in the code would be best in this situation.
Contributor guide
Assessment
This issue has not been assessed yet.