Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,212 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import re
|
3 |
+
import json
|
4 |
+
import time
|
5 |
+
import requests
|
6 |
+
import anthropic
|
7 |
+
import google.auth
|
8 |
+
import gradio as gr
|
9 |
+
from uuid import uuid4
|
10 |
+
from dotenv import load_dotenv
|
11 |
+
from google.auth.transport.requests import Request
|
12 |
+
|
13 |
+
|
14 |
+
# Gemini
|
15 |
+
def get_google_token():
|
16 |
+
credentials, project = google.auth.load_credentials_from_dict(
|
17 |
+
json.loads(os.environ.get('GCP_FINETUNE_KEY')),
|
18 |
+
scopes=[
|
19 |
+
"https://www.googleapis.com/auth/cloud-platform",
|
20 |
+
"https://www.googleapis.com/auth/generative-language.tuning",
|
21 |
+
],
|
22 |
+
)
|
23 |
+
request = Request()
|
24 |
+
credentials.refresh(request)
|
25 |
+
access_token = credentials.token
|
26 |
+
return access_token
|
27 |
+
|
28 |
+
|
29 |
+
def dubpro_english_to_hindi(text):
|
30 |
+
API_URL = os.environ.get("GEMINI_FINETUNED_ENG_HINDI_API")
|
31 |
+
BEARER_TOKEN = get_google_token()
|
32 |
+
headers = {
|
33 |
+
"Authorization": f"Bearer {BEARER_TOKEN}",
|
34 |
+
"Content-Type": "application/json",
|
35 |
+
}
|
36 |
+
payload = {
|
37 |
+
"contents": [
|
38 |
+
{
|
39 |
+
"parts": [{"text": f"text: {text}"}],
|
40 |
+
"role": "user",
|
41 |
+
}
|
42 |
+
],
|
43 |
+
"generationConfig": {
|
44 |
+
"maxOutputTokens": 8192,
|
45 |
+
"temperature": 0.85,
|
46 |
+
},
|
47 |
+
"safetySettings": [
|
48 |
+
{"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_NONE"},
|
49 |
+
{"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_NONE"},
|
50 |
+
{"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_NONE"},
|
51 |
+
{"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_NONE"},
|
52 |
+
],
|
53 |
+
}
|
54 |
+
result = requests.post(
|
55 |
+
url=API_URL,
|
56 |
+
headers=headers,
|
57 |
+
json=payload
|
58 |
+
)
|
59 |
+
response = result.json()
|
60 |
+
response_content = response['candidates'][0]['content']['parts'][0]['text'].replace("translated:", "").strip()
|
61 |
+
return response_content
|
62 |
+
|
63 |
+
|
64 |
+
def gemini_english_to_hindi(text):
|
65 |
+
API_URL = os.environ.get("GEMINI_FINETUNED_ENG_HINDI_API")
|
66 |
+
BEARER_TOKEN = get_google_token()
|
67 |
+
headers = {
|
68 |
+
"Authorization": f"Bearer {BEARER_TOKEN}",
|
69 |
+
"Content-Type": "application/json",
|
70 |
+
}
|
71 |
+
payload = {
|
72 |
+
"contents": [
|
73 |
+
{
|
74 |
+
"parts": [{"text": f"Translate the following text to Hindi: `{text}` Output: "}],
|
75 |
+
"role": "user",
|
76 |
+
}
|
77 |
+
],
|
78 |
+
"generationConfig": {
|
79 |
+
"maxOutputTokens": 8192,
|
80 |
+
"temperature": 0.85,
|
81 |
+
},
|
82 |
+
"safetySettings": [
|
83 |
+
{"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_NONE"},
|
84 |
+
{"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_NONE"},
|
85 |
+
{"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_NONE"},
|
86 |
+
{"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_NONE"},
|
87 |
+
],
|
88 |
+
}
|
89 |
+
result = requests.post(
|
90 |
+
url=API_URL,
|
91 |
+
headers=headers,
|
92 |
+
json=payload
|
93 |
+
)
|
94 |
+
response = result.json()
|
95 |
+
response_content = response['candidates'][0]['content']['parts'][0]['text'].replace("translated:", "").replace("`", "").strip()
|
96 |
+
return response_content
|
97 |
+
|
98 |
+
|
99 |
+
# GPT models
|
100 |
+
def clean(result):
|
101 |
+
text = result["choices"][0]['message']["content"]
|
102 |
+
text = re.sub(r"\(.*?\)|\[.*?\]","", text)
|
103 |
+
text = text.strip("'").replace('"', "")
|
104 |
+
if "\n" in text.strip("\n"):
|
105 |
+
text = text.split("\n")[-1]
|
106 |
+
return text
|
107 |
+
|
108 |
+
def openai_english_to_hindi(text, model):
|
109 |
+
prompt = f"Translate the following English text into Hindi such that the meaning in unchanged. Return only the translated text: `{text}`. Output: "
|
110 |
+
|
111 |
+
headers = {
|
112 |
+
"Content-Type": "application/json",
|
113 |
+
"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"
|
114 |
+
}
|
115 |
+
|
116 |
+
messages = [
|
117 |
+
{"role": "system", "content": f"You are a language translation assistant."},
|
118 |
+
{"role": "user", "content": prompt}
|
119 |
+
]
|
120 |
+
|
121 |
+
resp = None
|
122 |
+
while resp is None:
|
123 |
+
resp = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json={
|
124 |
+
"model": model,
|
125 |
+
"messages": messages
|
126 |
+
})
|
127 |
+
if resp.status_code != 200:
|
128 |
+
print(resp.text)
|
129 |
+
time.sleep(0.5)
|
130 |
+
response_json = resp.json()
|
131 |
+
|
132 |
+
result_text = clean(response_json)
|
133 |
+
return result_text
|
134 |
+
|
135 |
+
|
136 |
+
# Azure translate
|
137 |
+
def azure_english_to_hindi(text):
|
138 |
+
headers = {
|
139 |
+
"Ocp-Apim-Subscription-Key": os.environ.get("AZURE_TRANSLATE_KEY"),
|
140 |
+
"Ocp-Apim-Subscription-Region": os.environ.get("AZURE_TRANSLATE_REGION"),
|
141 |
+
"Content-type": "application/json",
|
142 |
+
"X-ClientTraceId": str(uuid4()),
|
143 |
+
}
|
144 |
+
ENDPOINT = "https://api.cognitive.microsofttranslator.com/translate"
|
145 |
+
params = {
|
146 |
+
"api-version": "3.0",
|
147 |
+
"from": "en-US",
|
148 |
+
"to": "hi-IN",
|
149 |
+
}
|
150 |
+
texts = [{"text": text}]
|
151 |
+
request = requests.post(ENDPOINT, headers=headers, params=params, json=texts)
|
152 |
+
response = request.json()
|
153 |
+
return response[0]["translations"][0]["text"]
|
154 |
+
|
155 |
+
|
156 |
+
# Anthopic Claude 3 Haiku
|
157 |
+
def claude_english_to_hindi(text):
|
158 |
+
client = anthropic.Anthropic()
|
159 |
+
message = client.messages.create(
|
160 |
+
model="claude-3-haiku-20240307",
|
161 |
+
max_tokens=1000,
|
162 |
+
temperature=0.8,
|
163 |
+
system="You are an expert language translator.",
|
164 |
+
messages=[
|
165 |
+
{
|
166 |
+
"role": "user",
|
167 |
+
"content": [
|
168 |
+
{
|
169 |
+
"type": "text",
|
170 |
+
"text": "Translate the following English text into Hindi such that the meaning in unchanged. Return only the translated text: `The Electrons journey starts with Complex 1 or it can start with Complex 2 either of that.`. Output: "
|
171 |
+
}
|
172 |
+
]
|
173 |
+
}
|
174 |
+
]
|
175 |
+
)
|
176 |
+
return message.content[0].text
|
177 |
+
|
178 |
+
|
179 |
+
def render_translations(text):
|
180 |
+
dubpro = dubpro_english_to_hindi(text)
|
181 |
+
azure = azure_english_to_hindi(text)
|
182 |
+
gemini = gemini_english_to_hindi(text)
|
183 |
+
gpt_4 = openai_english_to_hindi(text, model="gpt-4")
|
184 |
+
claude_haiku = claude_english_to_hindi(text)
|
185 |
+
return gr.update(value=gpt_4), gr.update(value=dubpro), gr.update(value=gemini), gr.update(value=claude_haiku), gr.update(value=azure)
|
186 |
+
|
187 |
+
|
188 |
+
with gr.Blocks(title="English to Hindi Translation Tools", theme="gradio/monochrome") as demo:
|
189 |
+
gr.Markdown("# English to Hindi Translation Tools")
|
190 |
+
input_textbox = gr.Textbox(label="Input Text", info="Text to translate", value="When you did it, you must have attended your classes well or you must have done your daily revision. Now you feel scared.")
|
191 |
+
submit = gr.Button(label="Submit")
|
192 |
+
with gr.Row():
|
193 |
+
gr.Label(value="Dubpro's Model", scale=1)
|
194 |
+
dubpro_model_textbox = gr.Textbox(label="Translated Text", scale=2, interactive=False)
|
195 |
+
with gr.Row():
|
196 |
+
gr.Label(value="GPT 4", scale=1)
|
197 |
+
gpt_4_textbox = gr.Textbox(label="Translated Text", scale=2, interactive=False)
|
198 |
+
with gr.Row():
|
199 |
+
gr.Label(value="Google Gemini", scale=1)
|
200 |
+
google_textbox = gr.Textbox(label="Translated Text", scale=2, interactive=False)
|
201 |
+
with gr.Row():
|
202 |
+
gr.Label(value="Anthropic Claude 3", scale=1)
|
203 |
+
claude_textbox = gr.Textbox(label="Translated Text", scale=2, interactive=False)
|
204 |
+
with gr.Row():
|
205 |
+
gr.Label(value="Azure Translate", scale=1)
|
206 |
+
azure_textbox = gr.Textbox(label="Translated Text", scale=2, interactive=False)
|
207 |
+
|
208 |
+
submit.click(render_translations, input_textbox, [gpt_4_textbox, dubpro_model_textbox, google_textbox, claude_textbox, azure_textbox])
|
209 |
+
|
210 |
+
|
211 |
+
if __name__=="__main__":
|
212 |
+
demo.launch()
|