bartowski's picture
Llamacpp quants
4686e82 verified
|
raw
history blame
3.77 kB
metadata
license: apache-2.0
base_model: []
library_name: transformers
tags:
  - mergekit
  - merge
pipeline_tag: text-generation
quantized_by: bartowski

Llamacpp Quantizations of Tess-10.7B-v2.0

Using llama.cpp release b2536 for quantization.

Original model: https://huggingface.co/Joseph717171/Tess-10.7B-v2.0

Download a file (not the whole branch) from below:

Filename Quant type File Size Description
Tess-10.7B-v2.0-Q8_0.gguf Q8_0 11.40GB Extremely high quality, generally unneeded but max available quant.
Tess-10.7B-v2.0-Q6_K.gguf Q6_K 8.80GB Very high quality, near perfect, recommended.
Tess-10.7B-v2.0-Q5_K_M.gguf Q5_K_M 7.59GB High quality, very usable.
Tess-10.7B-v2.0-Q5_K_S.gguf Q5_K_S 7.39GB High quality, very usable.
Tess-10.7B-v2.0-Q5_0.gguf Q5_0 7.39GB High quality, older format, generally not recommended.
Tess-10.7B-v2.0-Q4_K_M.gguf Q4_K_M 6.46GB Good quality, uses about 4.83 bits per weight.
Tess-10.7B-v2.0-Q4_K_S.gguf Q4_K_S 6.11GB Slightly lower quality with small space savings.
Tess-10.7B-v2.0-IQ4_NL.gguf IQ4_NL 6.14GB Decent quality, similar to Q4_K_S, new method of quanting,
Tess-10.7B-v2.0-IQ4_XS.gguf IQ4_XS 5.82GB Decent quality, new method with similar performance to Q4.
Tess-10.7B-v2.0-Q4_0.gguf Q4_0 6.07GB Decent quality, older format, generally not recommended.
Tess-10.7B-v2.0-Q3_K_L.gguf Q3_K_L 5.65GB Lower quality but usable, good for low RAM availability.
Tess-10.7B-v2.0-Q3_K_M.gguf Q3_K_M 5.19GB Even lower quality.
Tess-10.7B-v2.0-IQ3_M.gguf IQ3_M 4.84GB Medium-low quality, new method with decent performance.
Tess-10.7B-v2.0-IQ3_S.gguf IQ3_S 4.69GB Lower quality, new method with decent performance, recommended over Q3 quants.
Tess-10.7B-v2.0-Q3_K_S.gguf Q3_K_S 4.66GB Low quality, not recommended.
Tess-10.7B-v2.0-Q2_K.gguf Q2_K 4.00GB Extremely low quality, not recommended.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski