New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
large-v1 model support in CoreML #925
Comments
Yes, please add |
@abCods how fast does the large-v1 model compiles into cormel format for you? I have it running for hours, when I do force stop as suggested in other threads - I just get an error that it is not completed. hence can't compile it yet. |
On an M2 mini machine, it takes about a couple of seconds for me. It shouldn't even take minutes, let alone hours. I guess there could be an issue with any dependency either missing or non compatible. |
@abCods it really takes seconds on base/small models, but as I go to bigger ones it takes more time - medium already going some time, waiting until it is done.. |
What are your machine's specs? |
M1 Ultra / 128GB |
I'll get my hands on one of the M1s and share my experience. |
Hi @ggerganov
First of all, thank you for your amazing work with
Whisper.cpp
. I have been especially taking advantage of your work around the CoreML support for the Apple Silicon. It is truly a cost effective solution.In my case, I have been using the multilingual model including the
large
andlarge-v1
. Over the period of time the data suggests thatlarge-v1
is better off for the multilingual case as it offers better performance and it is less prone to the odd issue of repetitions in the transcriptions ( as the multilingual recordings are more prone to it. )When it comes to the CoreML, the
models/download-coreml-model.sh
has the option oflarge-v1
model.models=( "tiny.en" "tiny" "base.en" "base" "small.en" "small" "medium.en" "medium" "large-v1" "large" )
But the
models/convert-whisper-to-coreml.py
doesn't offer the further support on it.parser.add_argument("--model", type=str, help="model to convert (e.g. tiny, tiny.en, base, base.en, small, small.en, medium, medium.en, large)", required=True)
Am I missing something here or is the
large-v1
isn't supported yet? If so, I would be happy to contribute here.Thanks
The text was updated successfully, but these errors were encountered: