Spaces:
Running
Running
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
| 3 |
+
import json
|
| 4 |
+
|
| 5 |
+
# Load words and languages from JSON files
|
| 6 |
+
with open("top_500_quran_lemmas_fixed.json", encoding="utf-8") as f:
|
| 7 |
+
word_list = json.load(f)
|
| 8 |
+
|
| 9 |
+
with open("language_list.json", encoding="utf-8") as f:
|
| 10 |
+
language_list = json.load(f)
|
| 11 |
+
|
| 12 |
+
# Format dropdown options
|
| 13 |
+
word_options = [f"{word['text']} ({word['english']})" for word in word_list]
|
| 14 |
+
language_options = [f"{lang['name']} ({lang['code']})" for lang in language_list]
|
| 15 |
+
|
| 16 |
+
# Load DeepSeek-V3 model
|
| 17 |
+
model_id = "deepseek-ai/DeepSeek-V3"
|
| 18 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 19 |
+
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto")
|
| 20 |
+
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
|
| 21 |
+
|
| 22 |
+
# Generate prompt
|
| 23 |
+
def create_prompt(word_entry, language_code):
|
| 24 |
+
prompt = f"""
|
| 25 |
+
You are a friendly Quranic AI assistant.
|
| 26 |
+
|
| 27 |
+
The word is: {word_entry['text']} ({word_entry['english']})
|
| 28 |
+
|
| 29 |
+
Please provide the following in simple, easy-to-understand language, translated into {language_code}:
|
| 30 |
+
1. Translation of the word.
|
| 31 |
+
2. Its root word and any related words (derivatives).
|
| 32 |
+
3. Where it appears in the Qur'an — list the Surah and Ayah numbers.
|
| 33 |
+
4. Give an explanation of each appearance (based on context), in {language_code}.
|
| 34 |
+
|
| 35 |
+
Avoid technical terms. Make it feel like a helpful teacher explaining to a student.
|
| 36 |
+
"""
|
| 37 |
+
return prompt.strip()
|
| 38 |
+
|
| 39 |
+
# Function to call the model
|
| 40 |
+
def process(word_label, lang_label):
|
| 41 |
+
# Extract selected word data
|
| 42 |
+
selected_word = next((w for w in word_list if w['text'] in word_label), None)
|
| 43 |
+
language_code = lang_label.split("(")[-1].strip(")")
|
| 44 |
+
|
| 45 |
+
if not selected_word:
|
| 46 |
+
return "Word not found in list."
|
| 47 |
+
|
| 48 |
+
prompt = create_prompt(selected_word, language_code)
|
| 49 |
+
result = generator(prompt, max_new_tokens=800, do_sample=True, temperature=0.7)[0]["generated_text"]
|
| 50 |
+
|
| 51 |
+
return result.replace(prompt, "").strip()
|
| 52 |
+
|
| 53 |
+
# Gradio UI
|
| 54 |
+
with gr.Blocks() as demo:
|
| 55 |
+
gr.Markdown("## 📖 Quran Word Explorer with DeepSeek-V3")
|
| 56 |
+
with gr.Row():
|
| 57 |
+
word_input = gr.Dropdown(choices=word_options, label="Select a Quran Word")
|
| 58 |
+
lang_input = gr.Dropdown(choices=language_options, label="Select Language")
|
| 59 |
+
|
| 60 |
+
output = gr.Textbox(label="DeepSeek-V3 Output", lines=20)
|
| 61 |
+
run_btn = gr.Button("Get Info")
|
| 62 |
+
|
| 63 |
+
run_btn.click(fn=process, inputs=[word_input, lang_input], outputs=output)
|
| 64 |
+
|
| 65 |
+
demo.launch()
|