aashish1904
commited on
Commit
•
0ee2ab6
1
Parent(s):
58462c2
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
---
|
3 |
+
|
4 |
+
license: apache-2.0
|
5 |
+
language:
|
6 |
+
- zh
|
7 |
+
- en
|
8 |
+
pipeline_tag: text-generation
|
9 |
+
tags:
|
10 |
+
- roleplay
|
11 |
+
- rp
|
12 |
+
- character
|
13 |
+
|
14 |
+
---
|
15 |
+
|
16 |
+
![](https://lh7-rt.googleusercontent.com/docsz/AD_4nXeiuCm7c8lEwEJuRey9kiVZsRn2W-b4pWlu3-X534V3YmVuVc2ZL-NXg2RkzSOOS2JXGHutDuyyNAUtdJI65jGTo8jT9Y99tMi4H4MqL44Uc5QKG77B0d6-JfIkZHFaUA71-RtjyYZWVIhqsNZcx8-OMaA?key=xt3VSDoCbmTY7o-cwwOFwQ)
|
17 |
+
|
18 |
+
# QuantFactory/Peach-9B-8k-Roleplay-GGUF
|
19 |
+
This is quantized version of [ClosedCharacter/Peach-9B-8k-Roleplay](https://huggingface.co/ClosedCharacter/Peach-9B-8k-Roleplay) created using llama.cpp
|
20 |
+
|
21 |
+
# Original Model Card
|
22 |
+
|
23 |
+
|
24 |
+
<!-- header start -->
|
25 |
+
<!-- 200823 -->
|
26 |
+
<div style="width: auto; margin-left: auto; margin-right: auto">
|
27 |
+
<img src="./PeachGirl.png" alt="Peach" style="width: 100%; min-width: 400px; display: block; margin: auto;">
|
28 |
+
</div>
|
29 |
+
<!-- header end -->
|
30 |
+
|
31 |
+
# Peach-9B-8k-Roleplay
|
32 |
+
|
33 |
+
Peach-9B-8k-Roleplay is a chat large language model obtained by finetuning [01-ai/Yi-1.5-9B](https://huggingface.co/01-ai/Yi-1.5-9B) model on more than 100K conversations created through our data synthesis approach.
|
34 |
+
|
35 |
+
**Maybe The Best LLM with Small Parameters under 34B**
|
36 |
+
|
37 |
+
## How to start
|
38 |
+
The version of Transformers we are using is as follows, but a newer version may be available.
|
39 |
+
```
|
40 |
+
torch==1.13.1
|
41 |
+
gradio==3.50.2
|
42 |
+
transformers==4.37.2
|
43 |
+
```
|
44 |
+
|
45 |
+
Then run the following code to infer.
|
46 |
+
|
47 |
+
```python
|
48 |
+
import torch
|
49 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
50 |
+
|
51 |
+
model_name_or_path = "ClosedCharacter/Peach-9B-8k-Roleplay"
|
52 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
|
53 |
+
model = AutoModelForCausalLM.from_pretrained(
|
54 |
+
model_name_or_path, torch_dtype=torch.bfloat16,
|
55 |
+
trust_remote_code=True, device_map="auto")
|
56 |
+
messages = [
|
57 |
+
{"role": "system", "content": "你是黑丝御姐"},
|
58 |
+
{"role": "user", "content": "你好,你是谁"},
|
59 |
+
]
|
60 |
+
input_ids = tokenizer.apply_chat_template(conversation=messages, tokenize=True, return_tensors="pt")
|
61 |
+
output = model.generate(
|
62 |
+
inputs=input_ids.to("cuda"),
|
63 |
+
temperature=0.3,
|
64 |
+
top_p=0.5,
|
65 |
+
no_repeat_ngram_size=6,
|
66 |
+
repetition_penalty=1.1,
|
67 |
+
max_new_tokens=512)
|
68 |
+
print(tokenizer.decode(output[0]))
|
69 |
+
|
70 |
+
```
|
71 |
+
|
72 |
+
Or you can just use below code to run web demo.
|
73 |
+
```
|
74 |
+
python demo.py
|
75 |
+
```
|
76 |
+
## Benchmark
|
77 |
+
| Metric | Value |
|
78 |
+
|----------------|-----------------|
|
79 |
+
| MMLU (5-shot) | 66.19 |
|
80 |
+
| CMMLU (5-shot) | 69.07 |
|
81 |
+
|
82 |
+
|
83 |
+
## Warning
|
84 |
+
All response are generated by AI and do not represent the views or opinions of the developers.
|
85 |
+
|
86 |
+
1. Despite having done rigorous filtering, due to the uncontrollability of LLM, our model may still generate **toxic, harmful, and NSFW** content.
|
87 |
+
|
88 |
+
2. Due to limitations in model parameters, the 9B model may perform poorly on mathematical tasks, coding tasks, and logical capabilities.
|
89 |
+
|
90 |
+
3. Our training data is capped at a maximum length of 8k, so excessively long conversation turns may result in a decline in the quality of responses.
|
91 |
+
|
92 |
+
4. We used bilingual Chinese-English data for training, so the model may not perform well on other low-resource languages.
|
93 |
+
|
94 |
+
5. The model may generate a significant amount of hallucinations, so it is recommended to use lower values for temperature and top_p parameters.
|
95 |
+
|
96 |
+
|
97 |
+
# Contact Us
|
98 |
+
|
99 |
+
**微信 / WeChat: Fungorum**
|
100 |
+
|
101 |
+
**邮箱 / E-mail: [email protected]**
|