Greg Tatum
Greg Tatum
This looks like it's still valid. I guess we should try both directions in the pipeline script.
We're in a bit of a different state now. This is a great example of what information to include in our model registry. However the specific recommendation is a bit...
Our current setup uses "nmt_nfkc" for normalization, which is "Compatibility Decomposition, followed by Canonical Composition". That looks like a good normalization strategy for ensuring consistent byte representations of different Unicode...
This doesn't affect Taskcluster anymore due to our internal caching. Feel free to re-open if anyone hits this again in Snakemake.
Taskcluster caching is pretty robust these days for this type of issue.
If anyone is looking for a workaround, I was able to use stemming by: * Copying the repo locally (I used a git submodule) * Run `npm install` and `npm...
I sent the invites out, let me know if there is any other action on my part. I think fluent-langneg is blocked on Zibi.
If the idea is that the user should implement these, then it needs to be documented as such.
An example of this API appears to be faked in: https://github.com/projectfluent/fluent-rs/blob/9e62af6f18f5efed4ad8faaca2c6c0ceb14d5bde/fluent-fallback/tests/localization_test.rs#L129-L142
I'm planning on monitoring this through teacher training to see how it performs. If it looks good I'll let it proceed through the pipeline: https://firefox-ci-tc.services.mozilla.com/tasks/groups/DAWUQ580SXeggMPwzd0BHg