Jacob Kahn

@trias702 — you'd have to modify both the C++ source and bindings to get those values. If they're not bound now, there isn't a way to get them. cc @tlikhomanenko.

@loretoparisi — we probably don't have the bandwidth or knowhow to create NodeJS bindings ourselves, but we'd welcome a PR and would absolutely be interested in discussing this further. Would...

@Tom-Mirowski — if you're still willing, can you try running MKL's `setvars.sh` and see if that resolves the issue? SEe https://software.intel.com/content/www/us/en/develop/documentation/get-started-with-intel-oneapi-base-linux/top/before-you-begin.html under `Set Environment Variables for CLI Development`.

Yeah, this is a known issue with DNNL and vcpkg it looks like. Can you try building DNNL from source instead of using vcpkg? I hope to update these package...

@Anyueanne can you please paste the full trace if you're still at it? Also, if you're using the beam search decoder, that has been moved to [Flashlight Text](https://github.com/flashlight/text) — bindings...

@gmurthy — if you built Flashlight from source, can you run `make test` in your Flashlight build directory and ensure all of your tests are passing with the CPU backend?

@fann1993814 thanks for the PR -- I think we need to make this configurable and default to no scaling. If you're able to add that/get our existing tests passing, we...

@fangzhou-xie — great question, and sorry for the delay — we're still in the early stages of making Flashlight even more portable. R bindings aren't on the radar right now,...

@opentld @kommander — we don't have bandwidth to support Windows compilation at this point, but we'd welcome PRs. Most of the repo should compile with MSVC just fine.

@parthopdas — as with other bindings, we probably don't have the bandwidth or knowhow to create robust Rust bindings ourselves, but we'd welcome PRs and would absolutely be interested in...