Koshkasa/sophosympatheia_Magistry-24B-v1.1-IQ5_KS-GGUF overview

---

license: other

base_model:

• sophosympatheia/Magistry-24B-v1.1

library_name: ik_llama.cpp

pipeline_tag: text-generation

tags:

• gguf
• quantized
• ik_llama.cpp
• iq5_ks
• 5 bit
• roleplay
• mixed precision

quantized_by: Koshkasa

base_model_relation: quantized

---

What's that?

The goal: Make a high quality quant of sophosympatheia/Magistry-24B-v1.1 using SOTA quant types from ik_llama.cpp, allowing the resulting gguf to fit into 16gb VRAM with KV buffer, without KVO, with system overhead.

The result: Mixed precision quantization of sophosympatheia/Magistry-24B-v1.1

quantized with ik_llama.cpp build: 9d07d868

incompatible with mainline llama.cpp

Layout

| Layer | Dims | Dims | Quant |

| --- | --- | --- | --- |

| token\_embd | 5120 | 131072.0 | iq5\_ks |

| | | | |

| | blk| 40| |

| attn\_k | 5120 | 1024 | iq6\_k |

| attn\_norm | 5120 | 1 | f32 |

| attn\_q | 5120 | 4096 | iq6\_k |

| attn\_v | 5120 | 1024 | iq6\_k |

| ffn\_down | 32768 | 5120 | iq5\_ks |

| ffn\_gate | 5120 | 32768 | iq4\_k |

| ffn\_norm | 5120 | 1 | f32 |

| ffn\_up | 5120 | 32768 | iq5\_ks |

| attn\_output | 4096 | 5120 | iq6\_k |

| | | | |

| output | 5120 | 131072 | iq6\_k |

| output\_norm | 5120 | 1 | f32 |

Cheers

MistralAI - the beloved base model(s).

ikawrakow and contributors of ik_llama.cpp - I probably misused your wonderful creation.

sophosympatheia - for the merge effort.

Everyone whose finetunes were included in the merge!

bartowski - for the imatrix + the myriad of quants we all benefit from.