File size: 1,655 Bytes
901fa6c
 
 
 
 
 
 
 
 
b77ba7a
901fa6c
 
e442590
901fa6c
 
 
 
 
846668f
 
e442590
1f81cc4
901fa6c
 
 
6e0fe75
901fa6c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
---
license: mit
widget:
- text: |-
    <SC1>- как ты?
    - <extra_id_0>
  example_title: how r u
language:
- ru
pipeline_tag: text2text-generation
---

# Usage

```python
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

device = "cuda"

tokenizer = AutoTokenizer.from_pretrained('TeraSpace/dialofred')
model = AutoModelForSeq2SeqLM.from_pretrained('TeraSpace/dialofred', device_map=device)# Add torch_dtype=torch.bfloat16 to use less memory
while True:
    text_inp = input("=>")
    lm_text=f'<SC1>- {text_inp}\n- <extra_id_0>'
    input_ids=torch.tensor([tokenizer.encode(lm_text)]).to(model.device)
    # outputs=model.generate(input_ids=input_ids,
    #                                 max_length=200,
    #                                 eos_token_id=tokenizer.eos_token_id,
    #                                 early_stopping=True,
    #                                 do_sample=True,
    #                                 temperature=1.0,
    #                                 top_k=0,
    #                                 top_p=0.85)
    # outputs=model.generate(input_ids,eos_token_id=tokenizer.eos_token_id,early_stopping=True)
    outputs=model.generate(input_ids=input_ids,
                                    max_length=200,
                                    eos_token_id=tokenizer.eos_token_id,
                                    early_stopping=True,
                                    do_sample=True,
                                    temperature=0.7,
                                    top_k=0,
                                    top_p=0.8)
    
    print(tokenizer.decode(outputs[0][1:]))
```