Update README.md
Browse files
README.md
CHANGED
@@ -7,8 +7,6 @@ language:
|
|
7 |
|
8 |
# Llama-3-Chinese-8B-Instruct-GGUF
|
9 |
|
10 |
-
## Warning: llama.cpp has [breaking changes on Llama-3 pre-tokenizer](https://github.com/ggerganov/llama.cpp/pull/6920), which significantly affect performance. We will update GGUF mdoels in the next few hours.
|
11 |
-
|
12 |
<p align="center">
|
13 |
<a href="https://github.com/ymcui/Chinese-LLaMA-Alpaca-3"><img src="https://ymcui.com/images/chinese-llama-alpaca-3-banner.png" width="600"/></a>
|
14 |
</p>
|
@@ -23,18 +21,19 @@ Further details (performance, usage, etc.) should refer to GitHub project page:
|
|
23 |
|
24 |
Metric: PPL, lower is better
|
25 |
|
26 |
-
|
27 |
-
| :---: | -------: | ------------------: | -----------------: |
|
28 |
-
| Q2_K | 2.96 GB | 10.3918 +/- 0.13288 | **9.1722 +/- 0.11502** |
|
29 |
-
| Q3_K | 3.74 GB | 6.3018 +/- 0.07849 | **6.1901 +/- 0.07734** |
|
30 |
-
| Q4_0 | 4.34 GB | 6.0628 +/- 0.07501 | **5.9623 +/- 0.07444** |
|
31 |
-
| Q4_K | 4.58 GB | 5.9066 +/- 0.07419 | **5.8847 +/- 0.07406** |
|
32 |
-
| Q5_0 | 5.21 GB | 5.8562 +/- 0.07355 | **5.8032 +/- 0.07287** |
|
33 |
-
| Q5_K | 5.34 GB | 5.8062 +/- 0.07331 | **5.8058 +/- 0.07329** |
|
34 |
-
| Q6_K | 6.14 GB | 5.7757 +/- 0.07298 | **5.7745 +/- 0.07287** |
|
35 |
-
| Q8_0 | 7.95 GB | 5.7626 +/- 0.07272 | 5.7626 +/- 0.07272 |
|
36 |
-
| F16 | 14.97 GB | 5.7628 +/- 0.07275 | N/A |
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
## Others
|
40 |
|
|
|
7 |
|
8 |
# Llama-3-Chinese-8B-Instruct-GGUF
|
9 |
|
|
|
|
|
10 |
<p align="center">
|
11 |
<a href="https://github.com/ymcui/Chinese-LLaMA-Alpaca-3"><img src="https://ymcui.com/images/chinese-llama-alpaca-3-banner.png" width="600"/></a>
|
12 |
</p>
|
|
|
21 |
|
22 |
Metric: PPL, lower is better
|
23 |
|
24 |
+
*Note: Old models have been removed due to its inferior performance (llama.cpp has breaking changes on pre-tokenizer).*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
+
| Quant | Size | PPL (new model) | PPL (old model) |
|
27 |
+
| :---: | -------: | ------------------: | -----------------: |
|
28 |
+
| Q2_K | 2.96 GB | 9.1168 +/- 0.10711 | 10.3918 +/- 0.13288 |
|
29 |
+
| Q3_K | 3.74 GB | 5.4082 +/- 0.05955 | 6.3018 +/- 0.07849 |
|
30 |
+
| Q4_0 | 4.34 GB | 5.2048 +/- 0.05725 | 6.0628 +/- 0.07501 |
|
31 |
+
| Q4_K | 4.58 GB | 5.0189 +/- 0.05520 | 5.9066 +/- 0.07419 |
|
32 |
+
| Q5_0 | 5.21 GB | 4.9803 +/- 0.05493 | 5.8562 +/- 0.07355 |
|
33 |
+
| Q5_K | 5.34 GB | 4.9195 +/- 0.05436 | 5.8062 +/- 0.07331 |
|
34 |
+
| Q6_K | 6.14 GB | 4.8966 +/- 0.05413 | 5.7757 +/- 0.07298 |
|
35 |
+
| Q8_0 | 7.95 GB | 4.8822 +/- 0.05396 | 5.7626 +/- 0.07272 |
|
36 |
+
| F16 | 14.97 GB | 4.8802 +/- 0.05392 | 5.7628 +/- 0.07275 |
|
37 |
|
38 |
## Others
|
39 |
|