Update README.md
Browse files
README.md
CHANGED
@@ -5,17 +5,57 @@ thumbnail: "https://i.ibb.co/HBqvBFY/mountain-xianxia-chinese-scenic-landscape-c
|
|
5 |
tags:
|
6 |
- text generation
|
7 |
- pytorch
|
|
|
8 |
|
9 |
---
|
10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
## Qilin-lit-6b (V1.1.0)
|
12 |
|
13 |
-
Fine-tuned version of EleutherAI/gpt-j-6B on Coreweave's infrastructure (<https://www.coreweave.com/>) using an A40 over ~80 hours.
|
14 |
|
15 |
3150 steps, 1 epoch trained on 550 MB of primarily Xianxia genre Webnovels. (Translated to English)
|
16 |
|
17 |
---
|
18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
## Version History
|
20 |
|
21 |
1.1.0 - 550 MB Dataset 3150 steps epoch 1
|
|
|
5 |
tags:
|
6 |
- text generation
|
7 |
- pytorch
|
8 |
+
license: mit
|
9 |
|
10 |
---
|
11 |
|
12 |
+
#Qilin-lit-6b Description
|
13 |
+
|
14 |
+
Most updated version is V1.1.0 which is finetuned on 550 MB of webnovels found on the NovelUpdates website. (https://www.novelupdates.com/)
|
15 |
+
|
16 |
+
|
17 |
+
## Downstream Uses
|
18 |
+
|
19 |
+
This model can be used for entertainment purposes and as a creative writing assistant for fiction writers.
|
20 |
+
|
21 |
+
## Example Code
|
22 |
+
|
23 |
+
```
|
24 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
25 |
+
|
26 |
+
model = AutoModelForCausalLM.from_pretrained('rexwang8/qilin-lit-6b')
|
27 |
+
tokenizer = AutoTokenizer.from_pretrained('rexwang8/lit-6b')
|
28 |
+
|
29 |
+
prompt = '''I had eyes but couldn't see Mount Tai!'''
|
30 |
+
|
31 |
+
input_ids = tokenizer.encode(prompt, return_tensors='pt')
|
32 |
+
output = model.generate(input_ids, do_sample=True, temperature=1.0, top_p=0.9, repetition_penalty=1.2, max_length=len(input_ids[0])+100, pad_token_id=tokenizer.eos_token_id)
|
33 |
+
|
34 |
+
generated_text = tokenizer.decode(output[0])
|
35 |
+
print(generated_text)
|
36 |
+
```
|
37 |
+
|
38 |
+
---
|
39 |
## Qilin-lit-6b (V1.1.0)
|
40 |
|
41 |
+
Fine-tuned version of EleutherAI/gpt-j-6B (https://huggingface.co/EleutherAI/gpt-j-6B) on Coreweave's infrastructure (<https://www.coreweave.com/>) using an A40 over ~80 hours.
|
42 |
|
43 |
3150 steps, 1 epoch trained on 550 MB of primarily Xianxia genre Webnovels. (Translated to English)
|
44 |
|
45 |
---
|
46 |
|
47 |
+
## Team members and Acknowledgements
|
48 |
+
|
49 |
+
Rex Wang - Author
|
50 |
+
|
51 |
+
Coreweave - Computational materials
|
52 |
+
|
53 |
+
With help from:
|
54 |
+
|
55 |
+
Wes Brown, Anthony Mercurio
|
56 |
+
|
57 |
+
---
|
58 |
+
|
59 |
## Version History
|
60 |
|
61 |
1.1.0 - 550 MB Dataset 3150 steps epoch 1
|