|
--- |
|
base_model: alpindale/WizardLM-2-8x22B |
|
tags: |
|
- wizardlm |
|
- microsoft |
|
- instruct |
|
- finetune |
|
- gguf |
|
- importance matrix |
|
- imatrix |
|
model-index: |
|
- name: Not-WizardLM-2-8x22B-iMat-GGUF |
|
results: [] |
|
license: apache-2.0 |
|
--- |
|
|
|
# WizardLM-2-8x22B GGUF quants based on reupload at alpindale/WizardLM-2-8x22B |
|
|
|
## GGUFs created with an importance matrix (details below) |
|
|
|
This is based on a reupload by an alternate source as microsoft deleted the model shortly after release, I will validate checksums after it is released again, to see if MS did any changes. |
|
|
|
Source Model: [alpindale/WizardLM-2-8x22B](https://huggingface.co/lucyknada/microsoft_WizardLM-2-7B) |
|
|
|
Quantized with [llama.cpp](https://github.com/ggerganov/llama.cpp) commit [5dc9dd7152dedc6046b646855585bd070c91e8c8](https://github.com/ggerganov/llama.cpp/commit/5dc9dd7152dedc6046b646855585bd070c91e8c8) (master from 2024-04-09) |
|
|
|
Imatrix was generated from the f16 gguf via this command: |
|
|
|
./imatrix -c 512 -m $out_path/$base_quant_name -f $llama_cpp_path/groups_merged.txt -o $out_path/imat-f16-gmerged.dat |
|
|
|
Using the dataset from [here](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8395384) |
