Low performance using this library

[ohadkorenok]

Hey, i'm trying to run ggml-base-q5_1.bin on my mobile and its still very slow. Do you know how can it be improved?
I'm using trasncribeRealtime

[sirramin]

You can use onnx as an alternative

[ohadkorenok]

I don't think it's related to the model as much to the memory leak in the code....
Do you have a realtime example that works for you?

[sirramin]

https://github.com/microsoft/onnxruntime-inference-examples/tree/main/mobile/examples/React_Native/ort-rn-basic-usage

[ohadkorenok]

looking good bro but have you tried that? it has good results?

[sirramin]

Yes I implemented it in my app. It has much better performance and better accuracy.

…

On Sun, Aug 18, 2024, 12:26 AM Ohad Koren ***@***.***> wrote: looking good bro but have you tried that? it has good results? — Reply to this email directly, view it on GitHub <#263 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAYX67Q4Y6DTGLIRTNC7YB3ZR62GFAVCNFSM6AAAAABMOACEPWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOJUHE4DGNJVGY> . You are receiving this because you commented.Message ID: <mybigday/whisper .***@***.***>

[Martin000N]

Where exactly does the memory leak happen?

[ohadkorenok]

In realtime processing. I can't use it for too much long

…

On Sun, Feb 2, 2025, 10:30 AM Martin Novak ***@***.***> wrote: Where exactly does the memory leak happen? — Reply to this email directly, view it on GitHub <#263 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADIUSSQRLY6SAW4BQVOWU3D2NXJSJAVCNFSM6AAAAABMOACEPWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMRZGI4TOMZQGE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[Martin000N]

Have you done any research into seeing where the problem is exactly located (as in, if its part of whisper.cpp / is it part of the binding)? It would be really helpful if you made a minimal example program to see where in the code it happens. I cant replicate it myself.