aya-23-8B-IMat-GGUF / README.md
legraphista's picture
Upload README.md with huggingface_hub
3191314 verified
|
raw
history blame
5.42 kB
metadata
base_model: CohereForAI/aya-23-8B
inference: false
language:
  - en
  - fr
  - de
  - es
  - it
  - pt
  - ja
  - ko
  - zh
  - ar
  - el
  - fa
  - pl
  - id
  - cs
  - he
  - hi
  - nl
  - ro
  - ru
  - tr
  - uk
  - vi
library_name: gguf
license: cc-by-nc-4.0
pipeline_tag: text-generation
quantized_by: legraphista
tags:
  - quantized
  - GGUF
  - imatrix
  - quantization

aya-23-8B-IMat-GGUF

Llama.cpp imatrix quantization of aya-23-8B-IMat-GGUF

Original Model: CohereForAI/aya-23-8B
Original dtype: FP16 (float16)
Quantized by: llama.cpp b2998
IMatrix dataset: here

Files

IMatrix

Status: βœ… Available
Link: here

Common Quants

Filename Quant type File Size Status Uses IMatrix Is Split
aya-23-8B.Q8_0.gguf Q8_0 8.54GB βœ… Available βšͺ No πŸ“¦ No
aya-23-8B.Q6_K.gguf Q6_K 6.60GB βœ… Available βšͺ No πŸ“¦ No
aya-23-8B.Q4_K.gguf Q4_K 5.06GB βœ… Available 🟒 Yes πŸ“¦ No
aya-23-8B.Q3_K.gguf Q3_K 4.22GB βœ… Available 🟒 Yes πŸ“¦ No
aya-23-8B.Q2_K.gguf Q2_K 3.44GB βœ… Available 🟒 Yes πŸ“¦ No

All Quants

Filename Quant type File Size Status Uses IMatrix Is Split
aya-23-8B.FP16.gguf F16 16.07GB βœ… Available βšͺ No πŸ“¦ No
aya-23-8B.Q5_K.gguf Q5_K 5.80GB βœ… Available βšͺ No πŸ“¦ No
aya-23-8B.Q5_K_S.gguf Q5_K_S 5.67GB βœ… Available βšͺ No πŸ“¦ No
aya-23-8B.Q4_K_S.gguf Q4_K_S 4.83GB βœ… Available 🟒 Yes πŸ“¦ No
aya-23-8B.Q3_K_L.gguf Q3_K_L 4.53GB βœ… Available 🟒 Yes πŸ“¦ No
aya-23-8B.Q3_K_S.gguf Q3_K_S 3.87GB βœ… Available 🟒 Yes πŸ“¦ No
aya-23-8B.Q2_K_S Q2_K_S - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ4_NL IQ4_NL - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ4_XS IQ4_XS - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ3_M IQ3_M - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ3_S IQ3_S - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ3_XS IQ3_XS - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ3_XXS IQ3_XXS - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ2_M IQ2_M - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ2_S IQ2_S - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ2_XS IQ2_XS - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ2_XXS IQ2_XXS - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ1_M IQ1_M - ⏳ Processing 🟒 Yes -
aya-23-8B.IQ1_S IQ1_S - ⏳ Processing 🟒 Yes -

Downloading using huggingface-cli

First, make sure you have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download legraphista/aya-23-8B-IMat-GGUF --include "aya-23-8B.Q8_0.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download legraphista/aya-23-8B-IMat-GGUF --include "aya-23-8B.Q8_0/*" --local-dir aya-23-8B.Q8_0
# see FAQ for merging GGUF's

FAQ

Why is the IMatrix not applied everywhere?

According to this investigation, it appears that lower quantizations are the only ones that benefit from the imatrix input (as per hellaswag results).

How do I merge a split GGUF?

  1. Make sure you have gguf-split available
  2. Locate your GGUF chunks folder (ex: aya-23-8B.Q8_0)
  3. Run gguf-split --merge aya-23-8B.Q8_0/aya-23-8B.Q8_0-00001-of-XXXXX.gguf aya-23-8B.Q8_0.gguf
    • Make sure to point gguf-split to the first chunk of the split.

Got a suggestion? Ping me @legraphista!