Update README.md
Browse files
README.md
CHANGED
@@ -9,6 +9,25 @@ language:
|
|
9 |
|
10 |
This repository contains GGUF-v3 version (llama.cpp compatible) of **Chinese-Alpaca-2-7B-RLHF**, which is tuned on Chinese-Alpaca-2-7B with RLHF using DeepSpeed-Chat.
|
11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
For full model in HuggingFace format, please see: https://huggingface.co/hfl/chinese-alpaca-2-7b-rlhf
|
13 |
|
14 |
Please refer to [https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/) for more details.
|
|
|
9 |
|
10 |
This repository contains GGUF-v3 version (llama.cpp compatible) of **Chinese-Alpaca-2-7B-RLHF**, which is tuned on Chinese-Alpaca-2-7B with RLHF using DeepSpeed-Chat.
|
11 |
|
12 |
+
|
13 |
+
## Performance
|
14 |
+
|
15 |
+
Metric: PPL, lower is better
|
16 |
+
|
17 |
+
| Quant | original | imatrix (`-im`) |
|
18 |
+
|-----|------|------|
|
19 |
+
| Q2_K | 10.5211 +/- 0.14139 | 11.9331 +/- 0.16168 |
|
20 |
+
| Q3_K | 8.9748 +/- 0.12043 | 8.8238 +/- 0.11850 |
|
21 |
+
| Q4_K | 8.4643 +/- 0.11341 | 8.4226 +/- 0.11302 |
|
22 |
+
| Q5_K | 8.3722 +/- 0.11236 | 8.3336 +/- 0.11192 |
|
23 |
+
| Q6_K | 8.3207 +/- 0.11184 | 8.3047 +/- 0.11159 |
|
24 |
+
|
25 |
+
*The model with `-im` suffix is generated with important matrix, which has generally better performance (not always though).*
|
26 |
+
|
27 |
+
|
28 |
+
## Others
|
29 |
+
|
30 |
+
|
31 |
For full model in HuggingFace format, please see: https://huggingface.co/hfl/chinese-alpaca-2-7b-rlhf
|
32 |
|
33 |
Please refer to [https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2/) for more details.
|