meta-pytorch / meta-pytorch/torchcodec
More robust metadata serialization
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.2k
- Forks
- 125
- Avg merge
- 22h 47m
- Merged PRs (30d)
- 54
Description
To serialize the container/stream metadata to go from C++ to Python, we manually create a JSON string in the C++ code and parse that in Python to get a dict.
We should try to find a more robust way to do that. There are a few alternatives, but ultimately we are limited to what custom ops allow us to return. Some potential alternatives:
- use a proper JSON library (in C++)
- pack all the info into tensors.
- If we're OK to drop the compile/export support for metadata, we could just bind the C++ structs to Python with pybind11.
Contributor guide
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.
Research direction
Read the metadata construction in src/torchcodec/decoders/_core/VideoDecoderOps.cpp around line 196 and its parsing in src/torchcodec/decoders/_core/_metadata.py around line 81. Compare a C++ JSON library, tensor packing, and pybind11 bindings against the custom-op and compile/export constraints described in the issue. Done means metadata crosses the C++/Python boundary robustly with the required support preserved.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp, python
- Domain
- backend-api-design
- Issue type
- Refactor
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100