Mistral-Large-Instruct-2407-iMat-GGUF

<img src="https://i.imgur.com/P68dXux.png" width="400"/>

Mistral-Large-Instruct-2407-iMat-GGUF

[!WARNING] <b>Important Note:</b> Inferencing in llama.cpp has now been merged in PR #8604. Please ensure you are on release b3438 or newer. Text-generation-web-ui (Ooba) is also working as of 7/23. Official support for Kobold.cpp is still pending. </b>

Quantized from Mistral-Large-Instruct-2407 123B fp16

• Weighted quantizations were creating using fp16 GGUF and groups_merged.txt in 105 chunks and n_ctx=512
• For a brief rundown of iMatrix quant performance please see this PR
• <i>All quants are verified working prior to uploading to repo for your safety and convenience</i>

<b>KL-Divergence Reference Chart</b> (Click on image to view in full size) <img src="https://i.imgur.com/mV0nYdA.png" width="920"/>

[!TIP] <b>Quant-specific Tips:</b>

• If you are getting a cudaMalloc failed: out of memory error, try passing an argument for lower context in llama.cpp, e.g. for 8k: -c 8192
• If you have all ampere generation or newer cards, you can use flash attention like so: -fa
• Provided Flash Attention is enabled you can also use quantized cache to save on VRAM e.g. for 8-bit: -ctk q8_0 -ctv q8_0
• Files split with llama.cpp's gguf-split. No need to manually combine files - just download all files for a specific quant size and load the first file (labeled "00001-")

Original model card can be found here