File size: 8,461 Bytes
cbf16d4
 
 
 
 
 
 
 
7037bb1
 
693d55a
cbf16d4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
693d55a
cbf16d4
 
 
693d55a
 
 
 
 
 
 
 
 
 
 
7037bb1
cbf16d4
693d55a
cbf16d4
 
 
 
 
 
 
 
693d55a
cbf16d4
 
 
 
 
 
 
 
 
 
 
 
7037bb1
 
 
 
693d55a
7037bb1
 
 
cbf16d4
 
693d55a
 
 
cbf16d4
 
693d55a
 
cbf16d4
693d55a
 
 
 
 
 
 
cbf16d4
693d55a
cbf16d4
 
 
 
 
 
693d55a
cbf16d4
 
 
 
 
 
 
 
693d55a
cbf16d4
 
 
693d55a
 
cbf16d4
7037bb1
 
693d55a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7037bb1
 
693d55a
 
 
 
62421f1
693d55a
 
 
 
 
 
 
 
 
 
62421f1
7037bb1
 
f1de868
693d55a
 
 
 
f1de868
 
693d55a
 
 
 
f1de868
7037bb1
62421f1
cbf16d4
 
 
 
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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
import gradio as gr
import pixeltable as pxt
from pixeltable.functions.mistralai import chat_completions
from datetime import datetime
from textblob import TextBlob
import nltk
from nltk.tokenize import word_tokenize
from nltk.corpus import stopwords
import os
import getpass
import re

# Ensure necessary NLTK data is downloaded
nltk.download('punkt', quiet=True)
nltk.download('stopwords', quiet=True)

# Set up Mistral API key
if 'MISTRAL_API_KEY' not in os.environ:
    os.environ['MISTRAL_API_KEY'] = getpass.getpass('Mistral AI API Key:')

# Define UDFs
@pxt.udf
def get_sentiment_score(text: str) -> float:
    return TextBlob(text).sentiment.polarity

@pxt.udf
def extract_keywords(text: str, num_keywords: int = 5) -> list:
    stop_words = set(stopwords.words('english'))
    words = word_tokenize(text.lower())
    keywords = [word for word in words if word.isalnum() and word not in stop_words]
    return sorted(set(keywords), key=keywords.count, reverse=True)[:num_keywords]

@pxt.udf
def calculate_readability(text: str) -> float:
    words = len(re.findall(r'\w+', text))
    sentences = len(re.findall(r'\w+[.!?]', text)) or 1
    average_words_per_sentence = words / sentences
    return 206.835 - 1.015 * average_words_per_sentence

def run_inference_and_analysis(task, system_prompt, input_text, temperature, top_p, max_tokens, min_tokens, stop, random_seed, safe_prompt):
    # Initialize Pixeltable
    pxt.drop_table('mistral_prompts', ignore_errors=True)
    t = pxt.create_table('mistral_prompts', {
        'task': pxt.StringType(),
        'system': pxt.StringType(),
        'input_text': pxt.StringType(),
        'timestamp': pxt.TimestampType(),
        'temperature': pxt.FloatType(),
        'top_p': pxt.FloatType(),
        'max_tokens': pxt.IntType(),
        'min_tokens': pxt.IntType(),
        'stop': pxt.StringType(),
        'random_seed': pxt.IntType(),
        'safe_prompt': pxt.BoolType()
    })
    
    # Insert new row
    t.insert([{
        'task': task,
        'system': system_prompt,
        'input_text': input_text,
        'timestamp': datetime.now(),
        'temperature': temperature,
        'top_p': top_p,
        'max_tokens': max_tokens,
        'min_tokens': min_tokens,
        'stop': stop,
        'random_seed': random_seed,
        'safe_prompt': safe_prompt
    }])
    
    # Define messages for chat completion
    msgs = [
        {'role': 'system', 'content': t.system},
        {'role': 'user', 'content': t.input_text}
    ]

    common_params = {
        'messages': msgs,
        'temperature': temperature,
        'top_p': top_p,
        'max_tokens': max_tokens if max_tokens is not None else 300,
        'min_tokens': min_tokens,
        'stop': stop.split(',') if stop else None,
        'random_seed': random_seed,
        'safe_prompt': safe_prompt
    }
    
    # Run inference with both models
    t['open_mistral_nemo'] = chat_completions(model='open-mistral-nemo', **common_params)
    t['mistral_medium'] = chat_completions(model='mistral-medium', **common_params)
    
    # Extract responses
    t['omn_response'] = t.open_mistral_nemo.choices[0].message.content
    t['ml_response'] = t.mistral_medium.choices[0].message.content
    
    # Run analysis
    t['large_sentiment_score'] = get_sentiment_score(t.ml_response)
    t['large_keywords'] = extract_keywords(t.ml_response)
    t['large_readability_score'] = calculate_readability(t.ml_response)
    t['open_sentiment_score'] = get_sentiment_score(t.omn_response)
    t['open_keywords'] = extract_keywords(t.omn_response)
    t['open_readability_score'] = calculate_readability(t.omn_response)
    
    # Get results
    results = t.select(
        t.omn_response, t.ml_response,
        t.large_sentiment_score, t.open_sentiment_score,
        t.large_keywords, t.open_keywords,
        t.large_readability_score, t.open_readability_score
    ).tail(1)
    
    return (
        results['omn_response'][0],
        results['ml_response'][0],
        results['large_sentiment_score'][0],
        results['open_sentiment_score'][0],
        results['large_keywords'][0],
        results['open_keywords'][0],
        results['large_readability_score'][0],
        results['open_readability_score'][0]
    )

def gradio_interface():
    with gr.Blocks() as demo:
        gr.Markdown("# LLM Prompt Studio")

        with gr.Row():
            with gr.Column():
                # Input components
                task = gr.Textbox(label="Task")
                system_prompt = gr.Textbox(label="System Prompt", lines=3)
                input_text = gr.Textbox(label="Input Text", lines=3)

                with gr.Accordion("Advanced Settings", open=False):
                    temperature = gr.Slider(minimum=0, maximum=1, value=0.7, step=0.1, label="Temperature")
                    top_p = gr.Slider(minimum=0, maximum=1, value=0.9, step=0.1, label="Top P")
                    max_tokens = gr.Number(label="Max Tokens", value=300)
                    min_tokens = gr.Number(label="Min Tokens", value=None)
                    stop = gr.Textbox(label="Stop Sequences (comma-separated)")
                    random_seed = gr.Number(label="Random Seed", value=None)
                    safe_prompt = gr.Checkbox(label="Safe Prompt", value=False)

                # Example prompts
                examples = [
                    ["Sentiment Analysis", 
                    "You are an AI trained to analyze the sentiment of text. Provide a detailed analysis of the emotional tone, highlighting key phrases that indicate sentiment.",
                    "The new restaurant downtown exceeded all my expectations. The food was exquisite, the service impeccable, and the ambiance was perfect for a romantic evening. I can't wait to go back!",
                    0.3, 0.95, 200, None, "", None, False],
                    
                    ["Story Generation",
                    "You are a creative writer. Generate a short, engaging story based on the given prompt. Include vivid descriptions and an unexpected twist.",
                    "In a world where dreams are shared, a young girl discovers she can manipulate other people's dreams.",
                    0.9, 0.8, 500, 300, "The end", None, False]
                ]

                gr.Examples(
                    examples=examples,
                    inputs=[
                        task, system_prompt, input_text,
                        temperature, top_p, max_tokens,
                        min_tokens, stop, random_seed,
                        safe_prompt
                    ],
                    outputs=[
                        omn_response, ml_response,
                        large_sentiment, open_sentiment,
                        large_keywords, open_keywords,
                        large_readability, open_readability
                    ],
                    fn=run_inference_and_analysis
                )

                submit_btn = gr.Button("Run Analysis")

            with gr.Column():
                # Output components
                omn_response = gr.Textbox(label="Open-Mistral-Nemo Response")
                ml_response = gr.Textbox(label="Mistral-Medium Response")
                
                with gr.Row():
                    large_sentiment = gr.Number(label="Mistral-Medium Sentiment")
                    open_sentiment = gr.Number(label="Open-Mistral-Nemo Sentiment")
                
                with gr.Row():
                    large_keywords = gr.Textbox(label="Mistral-Medium Keywords")
                    open_keywords = gr.Textbox(label="Open-Mistral-Nemo Keywords")
                
                with gr.Row():
                    large_readability = gr.Number(label="Mistral-Medium Readability")
                    open_readability = gr.Number(label="Open-Mistral-Nemo Readability")

        submit_btn.click(
            run_inference_and_analysis,
            inputs=[
                task, system_prompt, input_text,
                temperature, top_p, max_tokens,
                min_tokens, stop, random_seed,
                safe_prompt
            ],
            outputs=[
                omn_response, ml_response,
                large_sentiment, open_sentiment,
                large_keywords, open_keywords,
                large_readability, open_readability
            ]
        )

    return demo

if __name__ == "__main__":
    gradio_interface().launch()