37 lines
1.0 KiB
Markdown
37 lines
1.0 KiB
Markdown
---
|
|
license: apache-2.0
|
|
language:
|
|
- zh
|
|
- en
|
|
---
|
|
|
|
# Chinese-LLaMA-2-7B-GGUF
|
|
|
|
This repository contains the GGUF-v3 models (llama.cpp compatible) for **Chinese-LLaMA-2-7B**.
|
|
|
|
|
|
## Performance
|
|
|
|
Metric: PPL, lower is better
|
|
|
|
| Quant | original | imatrix (`-im`) |
|
|
|-----|------|------|
|
|
| Q2_K | 15.1160 +/- 0.30469 | 12.7682 +/- 0.26022 |
|
|
| Q3_K | 9.9588 +/- 0.20549 | 9.8508 +/- 0.20484 |
|
|
| Q4_0 | 9.8085 +/- 0.20350 | - |
|
|
| Q4_K | 9.5802 +/- 0.20015 | 9.6327 +/- 0.20219 |
|
|
| Q5_0 | 9.4783 +/- 0.19622 | - |
|
|
| Q5_K | 9.5132 +/- 0.19989 | 9.4447 +/- 0.19772 |
|
|
| Q6_K | 9.4640 +/- 0.19909 | 9.4507 +/- 0.19849 |
|
|
| Q8_0 | 9.4659 +/- 0.19927 | - |
|
|
| F16 | 9.4627 +/- 0.19921 | - |
|
|
|
|
*The model with `-im` suffix is generated with important matrix, which has generally better performance (not always though).*
|
|
|
|
|
|
## Others
|
|
|
|
|
|
For Hugging Face version, please see: https://huggingface.co/hfl/chinese-llama-2-7b
|
|
|
|
Please refer to [https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/) for more details. |