Hugging Face's logo

Spaces:
disi98
/
model_validation
Runtime error

App Files Community
model_validation
File size: 4,993 Bytes 
51830d8
 
 
 
 
 
 
 
 
 
b1742e6
 
51830d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b1742e6
 
51830d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b1742e6
 
51830d8
 
 
 
 
 
 
 
 
 
 
b1742e6
 
 
51830d8
 
 
 
 
 
b1742e6
 
51830d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b1742e6
 
 
51830d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7023cf6
51830d8
 
 
 
 
 
 
 
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
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
 
# Import Libraries
import gradio as gr
import google.generativeai as genai



# using the gemini model
# api
genai.configure(api_key="AIzaSyCmdsQeYMapI6l3maUpAkrE_GkT348w5UA")

# model_1 = genai.GenerativeModel('gemini-pro')
# model_2 = genai.GenerativeModel('gemini-pro')

model = genai.GenerativeModel('gemini-pro')

# assign score function
def assign_score(response):
    if 'score=1' in response:
        score = 1
    elif 'score=2' in response:
        score = 2
    elif 'score=3' in response:
        score = 3
    elif 'score=4' in response:
        score = 4
    elif 'score=5' in response:
        score = 5
    else:
        score = 0
    return score

def model_prompt_response(prompt):
# using the gemini model
  
    prompt_response_1 = model.generate_content(prompt + "in 3 sentences or less")
    prompt_response_2 = model.generate_content(prompt + "in 3 sentences or less")
    response1_model = prompt_response_1.text
    response2_model = prompt_response_2.text

    prompt_to_judge_1 = f'Given the question, {prompt} and answer {response1_model}, decide whether: score = 1 when answer is strongly irrelevant to the question, score = 2 when the answer is slitely relevant, score = 3 when the answer is relevant, score = 4 when it is slitely relevant and score = 5 for strongly relevant answer. Give reason for your answer and start with scores and write the score as [score=1], [score=2], [score=3], [score=4], or [score=5].'

  
    prompt_to_judge_2 = f'Given the question, {prompt} and answer {response2_model}, decide whether: score = 1 when answer is strongly irrelevant to the question, score = 2 when the answer is slitely relevant, score = 3 when the answer is relevant, score = 4 when it is slitely relevant and score = 5 for strongly relevant answer. Give reason for your answer and start with scores and write the score as [score=1], [score=2], [score=3], [score=4], or [score=5]. '

    judge_1_response_1 = model.generate_content(prompt_to_judge_1)
    judge_2_response_2 = model.generate_content(prompt_to_judge_2)
    judge_1 = judge_1_response_1.text
    judge_2 = judge_2_response_2.text


    score1_judger = assign_score(judge_1)
    score2_judger = assign_score(judge_2) 


  # return values

    return response1_model, judge_1, score1_judger, response2_model, judge_2, score2_judger



# anonators judgement
def judge_response(score1_judger, score2_judger, judge_1, judge_2):
  score1_anonator = score1_judger
  score2_anonator = score2_judger
  response1_judger = judge_1
  response2_judger = judge_2

  return response1_judger, score1_anonator, response2_judger, score2_anonator

demo = gr.Blocks(theme=gr.themes.Monochrome(), title='Group1_Subtask2')




with demo:

  gr.Markdown(
      """
      # Group1_Subtask2
      ## The Model takes a request from the user and returns a response.
      #### A judger is asigned the response from the model and outputs corresponding judgement.
      #### The user asses the judgement and alters the responds to correspond to their views or objective truth
      """
  )
  prompt = gr.Textbox(label='Question to the model')
  with gr.Row():
    response1_model = gr.Textbox(label='Model Response 1')

    judge_1 = gr.Textbox(label='Judger Model Response 1/User Editable', interactive=True)

    score1_judger = gr.Number(value=2, label='Judger Model Score 1/User Editable', interactive=True)
  with gr.Row():
    response2_model = gr.Textbox(label='Model Response 2')
    judge_2 = gr.Textbox(label='Judger Model Response 2/User Editable', interactive=True)
    score2_judger = gr.Number(value=2, label='Judger Model Score 2/User Editable', interactive=True)


  response_model = gr.Button("Model response")





# user/judger edited response
  with gr.Row():
    response1_judger = gr.Textbox(label='Corrected Response 1')
    score1_anonator = gr.Number(label='Corrected score 1')
  
    response2_judger = gr.Textbox(label='Corrected Response 2')
    score2_anonator = gr.Number(label='Corrected Score 2')


  judge_model = gr.Button("Judge-Annonator")

  response_model.click(model_prompt_response, inputs=prompt, outputs=[response1_model,
                                                                      judge_1,
                                                                      score1_judger,
                                                                      response2_model,
                                                                      judge_2,
                                                                      score2_judger,
                                                                      ])
  judge_model.click(judge_response, inputs=[score1_judger, score2_judger, judge_1, judge_2] , outputs=[response1_judger, score1_anonator, response2_judger, score2_anonator])




demo.launch(share=True)

# Team Members
'''
 Collins Okoth: collins.okech@digitaldividedata.com
 Albert Disi: albert.disi@digitaldividedata.com
 Joseph Amadi: joseph.amadi@digitaldividedata.com
'''