wangyulong
commited on
Commit
•
4d9849b
1
Parent(s):
0e617ca
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,33 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: apache-2.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
language:
|
3 |
+
- zh
|
4 |
+
tags:
|
5 |
+
- text generation
|
6 |
+
- pytorch
|
7 |
+
- causal-lm
|
8 |
license: apache-2.0
|
9 |
---
|
10 |
+
|
11 |
+
# Mengzi-GPT-neo model (Chinese)
|
12 |
+
Pretrained model on 300G Chinese corpus.
|
13 |
+
|
14 |
+
## Usage
|
15 |
+
```python
|
16 |
+
import torch
|
17 |
+
import sentencepiece as spm
|
18 |
+
from transformers import GPTNeoForCausalLM
|
19 |
+
tokenizer = spm.SentencePieceProcessor(model_file="mengzi_gpt.model")
|
20 |
+
model = GPTNeoForCausalLM.from_pretrained("Langboat/mengzi-gpt-neo-base")
|
21 |
+
|
22 |
+
def lm(prompt, top_k, top_p, max_length, repetition_penalty):
|
23 |
+
input_ids = torch.tensor(tokenizer.encode([prompt]), dtype=torch.long, device='cuda')
|
24 |
+
gen_tokens = model.generate(
|
25 |
+
input_ids,
|
26 |
+
do_sample=True,
|
27 |
+
top_k=top_k,
|
28 |
+
top_p=top_p,
|
29 |
+
max_length=max_length+len(prompt),
|
30 |
+
repetition_penalty=repetition_penalty)
|
31 |
+
result = tokenizer.decode(gen_tokens.tolist())[0]
|
32 |
+
return result
|
33 |
+
```
|