Mistral-Large-Instruct-2407-v16-k65536-16384-woft

Disclaimer:

The model is reproduced based on the paper VPTQ: Extreme Low-bit Vector Post-Training Quantization for Large Language Models github and arXiv

The model itself is sourced from a community release.

It is intended only for experimental purposes.

Users are responsible for any consequences arising from the use of this model.

Note:

The PPL test results are for reference only and were collected using GPTQ testing script.

{
    "ctx_2048": {
        "wikitext2": 4.797906398773193,
        "c4": 7.611056804656982,
        "c4-new": 8.51098346710205
    },
    "ctx_4096": {
        "wikitext2": 4.509005069732666,
        "c4": 7.219871997833252,
        "c4-new": 8.102221488952637
    },
    "ctx_8192": {
        "wikitext2": 4.364211559295654,
        "c4": 5.560217380523682,
        "c4-new": 8.000030517578125
    }
}