mradermacher/shqiponja-59b-v1-i1-GGUF

About

weighted/imatrix quants of https://huggingface.co/nisten/shqiponja-59b-v1

only the first 40k tokens of my 160k token training data is used as the model overflowed (likely a problem with the model weights)

static quants are available at https://huggingface.co/mradermacher/shqiponja-59b-v1-GGUF

Usage

If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.

Provided Quants

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Link	Type	Size/GB	Notes
GGUF	i1-IQ1_S	13.4	for the desperate
GGUF	i1-IQ1_M	14.5	mostly desperate
GGUF	i1-IQ2_XXS	16.5
GGUF	i1-IQ2_XS	18.2
GGUF	i1-IQ2_S	19.2
GGUF	i1-IQ2_M	20.8
GGUF	i1-Q2_K	22.5	IQ3_XXS probably better
GGUF	i1-IQ3_XXS	23.4	lower quality
GGUF	i1-IQ3_XS	24.9
GGUF	i1-Q3_K_S	26.2	IQ3_XS probably better
GGUF	i1-IQ3_S	26.3	beats Q3_K*
GGUF	i1-IQ3_M	27.2
GGUF	i1-Q3_K_M	29.1	IQ3_S probably better
GGUF	i1-Q3_K_L	31.7	IQ3_M probably better
GGUF	i1-IQ4_XS	32.2
GGUF	i1-Q4_0	34.0	fast, low quality
GGUF	i1-Q4_K_S	34.2	optimal size/speed/quality
GGUF	i1-Q4_K_M	36.0	fast, recommended
GGUF	i1-Q5_K_S	41.2
GGUF	i1-Q5_K_M	42.3
GGUF	i1-Q6_K	49.0	practically like static Q6_K

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):

And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9

FAQ / Model Request

See https://huggingface.co/mradermacher/model_requests for some answers to questions you might have and/or if you want some other model quantized.

Thanks

I thank my company, nethype GmbH, for letting me use its servers and providing upgrades to my workstation to enable this work in my free time.