Anhforth commited on
Commit
d6cf52d
1 Parent(s): 4fe88f4

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -2
README.md CHANGED
@@ -31,12 +31,22 @@ The additional details of the Aquila model will be presented in the official tec
31
  Aquila2-7B is a base model that can be used for continuation.
32
 
33
  ```python
34
- from transformers import AutoTokenizer, AutoModelForCausalLM
35
  import torch
 
 
 
36
  device = torch.device("cuda")
37
  model_info = "BAAI/Aquila2-7B"
38
  tokenizer = AutoTokenizer.from_pretrained(model_info, trust_remote_code=True)
39
- model = AutoModelForCausalLM.from_pretrained(model_info, trust_remote_code=True)
 
 
 
 
 
 
 
 
40
  model.eval()
41
  model.to(device)
42
  text = "杭州亚运会的亮点和期待 2023年9月23日至10月8日,杭州将举办第19届亚洲运动会"
 
31
  Aquila2-7B is a base model that can be used for continuation.
32
 
33
  ```python
 
34
  import torch
35
+ from transformers import AutoTokenizer, AutoModelForCausalLM
36
+ from transformers import BitsAndBytesConfig
37
+
38
  device = torch.device("cuda")
39
  model_info = "BAAI/Aquila2-7B"
40
  tokenizer = AutoTokenizer.from_pretrained(model_info, trust_remote_code=True)
41
+ quantization_config=BitsAndBytesConfig(
42
+ load_in_4bit=True,
43
+ bnb_4bit_use_double_quant=True,
44
+ bnb_4bit_quant_type="nf4",
45
+ bnb_4bit_compute_dtype=torch.bfloat16,
46
+ )
47
+ model = AutoModelForCausalLM.from_pretrained(model_info, trust_remote_code=True, torch_dtype=torch.float16,
48
+ # quantization_config=quantization_config, # Uncomment this line for 4bit quantization
49
+ )
50
  model.eval()
51
  model.to(device)
52
  text = "杭州亚运会的亮点和期待 2023年9月23日至10月8日,杭州将举办第19届亚洲运动会"