Jelmer

Results 74 issues of Jelmer

I'm trying to use the matmul primitive to multiply two quantized matrices (A:int8 & B:int8) into a C:float32, and want to add an unquantized bias:float32 to the output. To compensate...

enhancement

### Description This is a quick fix for marian-decoder always aborting on an uncaught `YAML::InvalidNode` exception when marian is compiled without CPU support. It is being thrown from the `get("model-mmap")`...

### Bug description Marian is build without AVX2 support on macOS even though the cpu (in this case [Intel(R) Core(TM) i9-9880H](https://ark.intel.com/content/www/us/en/ark/products/192987/intel-core-i9-9880h-processor-16m-cache-up-to-4-80-ghz.html)) does support the instructions. It seems to be due...

bug

Replacing the WASM demo page with code that can be copied into other projects so people can easily embed bergamot translator on their own website. I'm tempted to also add...

This looked like the easiest way to provide some documentation & examples for the WASM API. I've upgraded to node 18 because that allows us to use `fetch()` which matches...

When `responseOptions.qualityScores:true` is used with a model that has `skip-cost:true`, the quality scores output is still there and there is no error message. However, they average around 20 instead of...

low-priority

See https://developer.mozilla.org/en-US/docs/Web/HTML/Global_attributes/translate I can make elements with this attribute skip into the skip-element function, which is used for ignoring `` etc. But that might mess up sentences, as this attribute...

enhancement
help wanted
mod: html

Wikipedia uses [`0xad` unicode characters](https://en.wikipedia.org/wiki/Soft_hyphen) inside words to help word wrapping, but our translation models generally don't like that. For de->en it results in a lot of seemingly random `Ã3`...

mod: html

``` [IPKernelApp] ERROR | Exception in message handler: TypeError: Unregistered type : std::__1::vector The above exception was the direct cause of the following exception: Traceback (most recent call last): File...

bug
mod: python

From the last call: > Should we be dying on HTML parse errors and such? These are query-level errors that should be handled better with an error response. I proposed...