John

Results 41 issues of John

I've not digged deep into this yet but my whole CPU utilization is only at 50%. I've compiled it with current VS build tools, all default, release mode of course....

It's only psychological nice to have the program jump to the inference part directly after mmap load. The entire model is loaded during first inference which means it influences our...

When static linking is selected the CUDA::cublas_static target is not found. Dynamic binary compilation works.

Given that the tensor struct uses padding it's not nice to add any more information into it. It currently has a static 8 byte padding at the end, that's perfect...

enhancement
high priority

Performance with cuBLAS isn't there yet, it is more a burden than a speedup with llama eval in my tests. In a simple benchmark case it is absolutely amazing, getting...

I changed the memory management The current variant only supports 16 allocated free buffers and it uses the first free one even if a better size is available. The new...

1) ggml tensors need a layer number parameter I'd use layer 0 for global and 1+ (could also be -1 and 0+ of course) 2) When a ggml tensor is...

enhancement
good first issue

### Question **Question 1)** I'm curious about the use of newline embedding, from what I can see it is appended after each of the subimage features post "unpad" then the...

I have a problem with my plugin stopping to work for some reason, it might be LivePlugin related or something else I can't figure out. However, a click on "reload"...

The timer stays loaded in the context of the LivePlugin. Even when you reload (or stop) the plugin, it will still continue in the background. If you start it several...