Japanese iMatrix dataset
First and foremost, I want to express my sincere gratitude to the person who quantized and published the model I recently created. Thank you so much for your effort and contribution.
I have a question about the imatrix dataset used for the quantization process. Could you please provide more information about your imatrix dataset? I'm curious to learn more about it.
I would also like to mention that there is an imatrix dataset specifically designed for Japanese LLMs.
TFMC/imatrix-dataset-for-japanese-llm
If you have the time and opportunity in the future, I kindly request that you consider trying out the quantization process using the Japanese LLM imatrix dataset as well.
Once again, I deeply appreciate your work in quantizing and publishing the model. When I release a new model, I will reach out and inform you.
Thank you for your time and consideration.
I write about it a little in my FAQ - it contains almost exclusively english and no japanese at all, which is why I usually avoid imatrix quants of primarily non-english models (if I can identify them :). I cannot easily incorporate other sets at the moment, but I will keep it in mind.