MonteXiaofeng commited on
Commit
4a49479
1 Parent(s): 7d3f15f

Upload 3 files

Browse files
Files changed (3) hide show
  1. README.md +86 -0
  2. img/eval-result.jpeg +0 -0
  3. img/pipeline.png +0 -0
README.md CHANGED
@@ -1,3 +1,89 @@
1
  ---
2
  license: apache-2.0
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
  ---
4
+
5
+ ## Introduction
6
+
7
+ Aquila is a large language model trained by BAAI, and AquilaMed-RL is an industry model from Aquila language model. Based on the Aquila general pre-trained model, we continued pre-training , SFT and RL in the medical domain and obtained our AquilaMed-RL model.
8
+
9
+ ## Model Details
10
+
11
+ The pipeline of the training procedure is bellow, for more details you can read our technical report: https://github.com/FlagAI-Open/industry-application/blob/main/Aquila_med_tech-report.pdf
12
+
13
+ ![pipeline](./img/pipeline.png)
14
+
15
+ ## Evaluation
16
+
17
+ ![pipeline](./img/pipeline.png)
18
+
19
+ ## usage
20
+
21
+ when you have downloaded the model, you can use the bellow code to run the model
22
+
23
+ ```python
24
+
25
+ import torch
26
+ from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig
27
+
28
+
29
+ model_dir = "xxx"
30
+
31
+ tokenizer = AutoTokenizer.from_pretrained(model_dir, trust_remote_code=True)
32
+
33
+ config = AutoConfig.from_pretrained(model_dir, trust_remote_code=True)
34
+ model = AutoModelForCausalLM.from_pretrained(
35
+ model_dir, config=config, trust_remote_code=True
36
+ )
37
+ model.cuda()
38
+ model.eval()
39
+
40
+ template = "<|im_start|>system\nYou are a helpful assistant in medical domain.<|im_end|>\n<|im_start|>user\n{question}<|im_end|>\n<|im_start|>assistant\n"
41
+
42
+ text = "我肚子疼怎么办?"
43
+
44
+ item_instruction = template.format(question=text)
45
+
46
+ inputs = tokenizer(item_instruction, return_tensors="pt").to("cuda")
47
+ input_ids = inputs["input_ids"]
48
+ prompt_length = len(input_ids[0])
49
+ generate_output = model.generate(
50
+ input_ids=input_ids, do_sample=False, max_length=1024, return_dict_in_generate=True
51
+ )
52
+
53
+ response_ids = generate_output.sequences[0][prompt_length:]
54
+ predicts = tokenizer.decode(
55
+ response_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True
56
+ )
57
+
58
+ print("predict:", predicts)
59
+
60
+
61
+ """
62
+ predict: 肚子疼可能是多种原因引起的,例如消化不良、胃炎、胃溃疡、胆囊炎、胰腺炎、肠道感染等。如果疼痛持续或加重,或者伴随有呕吐、腹泻、发热等症状,建议尽快就医。如果疼痛轻微,可以尝试以下方法缓解:
63
+
64
+ 1. 饮食调整:避免油腻、辛辣、刺激性食物,多喝水,多吃易消化的食物,如米粥、面条、饼干等。
65
+
66
+ 2. 休息:避免剧烈运动,保持充足的睡眠。
67
+
68
+ 3. 热敷:用热水袋或毛巾敷在肚子上,可以缓解疼痛。
69
+
70
+ 4. 药物:可以尝试一些非处方药,如布洛芬、阿司匹林等,但请务必在医生的指导下使用。
71
+
72
+ 如果疼痛持续或加重,或者伴随有其他症状,建议尽快就医。
73
+
74
+ 希望我的回答对您有所帮助。如果您还有其他问题,欢迎随时向我提问。
75
+ """
76
+ ```
77
+
78
+
79
+
80
+ ## Citation
81
+
82
+ If you find our work helpful, feel free to give us a cite.
83
+
84
+ ```
85
+ @article{AquilaMed,
86
+ title={AquilaMed Technical Report},
87
+ year={2024}
88
+ }
89
+ ```
img/eval-result.jpeg ADDED
img/pipeline.png ADDED