The hybrid quant employs different quantization levels on a per layer basis to enable both high performance and small file size at the same time. The quants employed are all K to avoid slow CPU or older GPU processing of IQ quants.

Q4_K_H layer quants are as follows:

Q4_K_L : Q4_K_M + attn_o = q6_k
Q5_K_L : Q5_K_L : attn_v = q8_0 attn_o = q6_k ffn_d = q6_k

   LAYER_TYPES='[
   [0 ,"Q4_K_M"],[1 ,"Q4_K_S"],[2 ,"Q3_K_L"],[3 ,"Q3_K_M"],[4 ,"Q3_K_L"],[5 ,"Q3_K_M"],[6 ,"Q3_K_L"],[7 ,"Q3_K_M"],
   [8 ,"Q3_K_L"],[9 ,"Q3_K_M"],[10,"Q3_K_L"],[11,"Q3_K_M"],[12,"Q3_K_L"],[13,"Q3_K_L"],[14,"Q4_K_S"],[15,"Q3_K_L"],
   [16,"Q4_K_S"],[17,"Q3_K_L"],[18,"Q4_K_S"],[19,"Q3_K_L"],[20,"Q4_K_S"],[21,"Q3_K_L"],[22,"Q4_K_S"],[23,"Q3_K_L"],
   [24,"Q4_K_S"],[25,"Q4_K_S"],[26,"Q4_K_S"],[27,"Q4_K_S"],[28,"Q4_K_S"],[29,"Q4_K_S"],[30,"Q4_K_S"],[31,"Q4_K_S"],
   [32,"Q4_K_M"],[33,"Q4_K_S"],[34,"Q4_K_M"],[35,"Q4_K_S"],[36,"Q4_K_M"],[37,"Q4_K_S"],[38,"Q4_K_M"],[39,"Q4_K_S"],
   [40,"Q4_K_M"],[41,"Q4_K_M"],[42,"Q4_K_M"],[43,"Q4_K_M"],[44,"Q4_K_L"],[45,"Q4_K_M"],[46,"Q4_K_L"],[47,"Q5_K_L"]
   ]'
   FLAGS="--token-embedding-type Q4_K --output-tensor-type Q6_K --layer-types-high"

This quant was optimized over a small set of curated test prompts for code generation ability and then sanity checked for good performance on humaneval.

Comparison:

Quant	size	PPL	Comment
IQ4_XS	8.2e9	8.03	-
Q4_K_H	8.6e9	8.06	Hybrid quant with Q4_K embedding Q6_K output

Usage:

The model can be speculated with Qwen 2.5 Coder 0.5B Instruct with no vocab translation. It is trained at 32k context which can be extended to 128k using YARN:

-rope-scaling yarn --yarn-orig-ctx 32768 --rope_scale 4

For other than 128k context set rope_scale to the fraction of configured context size / 32768.0.

Approximate performance on 12G VRAM 4070 with weigths and context in VRAM:

Q	QKV	ND	NKV	gen tps	Comment
Q4_K_H	F16	0	16k	48	No draft
Q4_K_H	F16	8	12.5k	143	Spec 8
Q4_K_H	Q8_0	0	29.5k	48	No draft
Q4_K_H	Q8_0	8	22k	143	Spec 8

for speculation a fixed length ND=8 token draft was used with a custom downstream speculator.

Benchmarks:

A full set of code evals for the quant is given here: https://huggingface.co/spaces/steampunque/benchlm

Download the file from below:

Link	Type	Size/e9 B	Notes
Qwen2.5-Coder-14B-Instruct.Q4_K_H.gguf	Q4_K_H	8.6e9 B	better code gen performance than IQ4_XS

A discussion thread about the hybrid layer quant approach can be found here on the llama.cpp git repository:

https://github.com/ggml-org/llama.cpp/discussions/13040

steampunque/Qwen2.5-Coder-14B-Instruct-MP-GGUF

Mixed Precision GGUF layer quantization of Qwen2.5-Coder-14B-Instruct by Qwen

Download the file from below:

steampunque/
Qwen2.5-Coder-14B-Instruct-MP-GGUF