Date: 20210707
Start time: TBD
Zoom: TBD
- Ricky might not be able to make it, attending Rust Meetup Edinburg
- neuronika
- Enzyme updates: Manuel's first rustc PR got merged :) Update on next steps (updating rustup)
- Linfa OPTICS clustering PR (feels close to merging)
- Moving towards replacing reqwest with a combination of curl + pbr crates for MNIST download
- chrism
- Manuel
- xd009642
- david sjaustirni
-
Saw neuronika on Reddit]
- https://github.com/neuronika/neuronika
- Working on becoming similar to a Rust version of PyTorch, etc. (in the auto-diff area)
-
Daniel is working on Linfa OPTICS algorithm
- Is getting feedback on a little bit of
- Uses hnsw, works to update MSRV to ~1.50
- Based on supporting const generics
- Have a conversation with Lorenz about changing the Linfa CI away from the pinned nightly version up to the latest?
- Yuhan thinks that nightly in the CI might not be necessary at all
- Has a small lifetime issue, would be grateful is anyone could find some time to take a look
- Otherwise, seems close to merging
-
David has opened an issue on linfa for making pca work with sparse components
- rust-ml/linfa#143
- Lorenz suggested taking a look a some parameters in order to judge which particularly type of algorithm use
- Should be
-
David is interested in using a large external dataset to test some code (Yale Face Database B)
- Asked about including that directly in linfa-datasets
- Appx. 2GB in size
- Chris thinks breaking that out into a function for pulling that data in as needed from an external cache would make more sense, rather than directly including it
- Asked about including that directly in linfa-datasets
-
Daniel's group is doing speech recognition in Rust
- Needs to be able to handle at least 1000 voices
- Using the
- One researcher has had some issues with the Kodama/Linfa APIs; will ask them to do a small write-up about what those were
- Currently using Kaldi, also looking into https://speechbrain.github.io/