s3nh

/

mamba-gpt-3b-GGML

like 3

Thanks. Yup it is pretty good despite of quantizatiob method. I thought that there ll be lack of accuracy on more compressed models, but it is solid. Ill try tommorow with cerebras one.

You can also test it here https://huggingface.co/spaces/s3nh/mamba-gpt-3b

Pretty good model, can you also quantize this new model: https://huggingface.co/cerebras/btlm-3b-8k-base

thanks in advance

Try that one at night but it seems that btml is not support at the moment.