Jannis Schönleber

icon indicating a website link www.linkedin.com/in/jannis-schoenleber

icon company 44ai (former ETHZ, Huawei Research) icon location Bern, Switzerland icon location joennlae = jönnlä (swiss german)

Results 24 comments of Jannis Schönleber

[Bug]: Regression ~~for AWQ marlin kernels~~ from v0.6.2 to v0.6.3 when using CUDA Graphs

Maybe it is not a marlin issue (sorry for the call out): #9448 They report it with a non quantised model.

[Feat][whisper] add more sampling parameters to whisper endpoint

Added documentation and updated the example

[Feat][whisper] add more sampling parameters to whisper endpoint

I am currently not working on that. As the key feature for us would be beam search implementation. Which is currently in a heavy limbo state due to the migration...

Discussion: allow different placement margin per instance also in global placement

@maliberty The instances need a lower placement density, and this is the solution we could come up with. There may be a better solution that we could not find. We...