Spaces:
Running
Running
Update main.py
Browse files
main.py
CHANGED
@@ -10,19 +10,23 @@ from fastapi.middleware.cors import CORSMiddleware
|
|
10 |
from fastapi.responses import HTMLResponse
|
11 |
import uvicorn
|
12 |
|
|
|
13 |
from pydantic import BaseModel
|
14 |
from pymongo import MongoClient
|
15 |
import jwt
|
16 |
from jwt import encode as jwt_encode
|
17 |
-
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
|
18 |
from bson import ObjectId
|
19 |
|
|
|
20 |
import ctranslate2
|
21 |
import sentencepiece as spm
|
22 |
import fasttext
|
|
|
23 |
|
24 |
-
import pytz
|
25 |
from datetime import datetime
|
|
|
|
|
|
|
26 |
import os
|
27 |
|
28 |
app = FastAPI()
|
@@ -47,9 +51,9 @@ templates_folder = os.path.join(os.path.dirname(__file__), "templates")
|
|
47 |
|
48 |
# Authentication
|
49 |
class User(BaseModel):
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
|
54 |
# Connect to the MongoDB database
|
55 |
client = MongoClient("mongodb://localhost:27017")
|
@@ -64,63 +68,63 @@ security = HTTPBearer()
|
|
64 |
|
65 |
@app.post("/login")
|
66 |
def login(user: User):
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
|
81 |
#Implement the registration route:
|
82 |
@app.post("/register")
|
83 |
def register(user: User):
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
return {"message": "User already exists"}
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
|
99 |
|
100 |
#Implement the `/api/user` route to fetch user data based on the JWT token
|
101 |
@app.get("/api/user")
|
102 |
def get_user(credentials: HTTPAuthorizationCredentials = Depends(security)):
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
if user_data["username"] and user_data["email"]:
|
116 |
-
|
117 |
-
raise HTTPException(status_code=401, detail="Invalid token")
|
118 |
|
119 |
#Define a helper function to generate a JWT token
|
120 |
def generate_token(email: str) -> str:
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
|
125 |
|
126 |
# Get time of request
|
@@ -135,6 +139,29 @@ def get_time():
|
|
135 |
|
136 |
full_date = f"{curr_day} | {curr_date} | {curr_time}"
|
137 |
return full_date, curr_time
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
138 |
|
139 |
# Load the model and tokenizer ..... only once!
|
140 |
beam_size = 1 # change to a smaller value for faster inference
|
@@ -154,13 +181,17 @@ sp_model_full_path = os.path.join(os.path.dirname(__file__), sp_model_file)
|
|
154 |
sp = spm.SentencePieceProcessor()
|
155 |
sp.load(sp_model_full_path)
|
156 |
|
|
|
157 |
# Import The Translator model
|
158 |
print("\nimporting Translator model")
|
159 |
ct_model_file = "sematrans-3.3B"
|
160 |
ct_model_full_path = os.path.join(os.path.dirname(__file__), ct_model_file)
|
161 |
translator = ctranslate2.Translator(ct_model_full_path, device)
|
|
|
|
|
|
|
162 |
|
163 |
-
print('\nDone importing models\n')
|
164 |
|
165 |
|
166 |
def translate_detect(userinput: str, target_lang: str):
|
@@ -213,6 +244,25 @@ def translate_enter(userinput: str, source_lang: str, target_lang: str):
|
|
213 |
# Return the source language and the translated text
|
214 |
return translations_desubword[0]
|
215 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
216 |
|
217 |
@app.get("/", response_class=HTMLResponse)
|
218 |
async def read_root(request: Request):
|
@@ -258,5 +308,23 @@ async def translate_enter_endpoint(request: Request):
|
|
258 |
"translated_text": translated_text_e,
|
259 |
}
|
260 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
261 |
|
262 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
from fastapi.responses import HTMLResponse
|
11 |
import uvicorn
|
12 |
|
13 |
+
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
|
14 |
from pydantic import BaseModel
|
15 |
from pymongo import MongoClient
|
16 |
import jwt
|
17 |
from jwt import encode as jwt_encode
|
|
|
18 |
from bson import ObjectId
|
19 |
|
20 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
21 |
import ctranslate2
|
22 |
import sentencepiece as spm
|
23 |
import fasttext
|
24 |
+
import torch
|
25 |
|
|
|
26 |
from datetime import datetime
|
27 |
+
import gradio as gr
|
28 |
+
import pytz
|
29 |
+
import time
|
30 |
import os
|
31 |
|
32 |
app = FastAPI()
|
|
|
51 |
|
52 |
# Authentication
|
53 |
class User(BaseModel):
|
54 |
+
username: str = None # Make the username field optional
|
55 |
+
email: str
|
56 |
+
password: str
|
57 |
|
58 |
# Connect to the MongoDB database
|
59 |
client = MongoClient("mongodb://localhost:27017")
|
|
|
68 |
|
69 |
@app.post("/login")
|
70 |
def login(user: User):
|
71 |
+
# Check if user exists in the database
|
72 |
+
user_data = users_collection.find_one(
|
73 |
+
{"email": user.email, "password": user.password}
|
74 |
+
)
|
75 |
+
if user_data:
|
76 |
+
# Generate a token
|
77 |
+
token = generate_token(user.email)
|
78 |
+
# Convert ObjectId to string
|
79 |
+
user_data["_id"] = str(user_data["_id"])
|
80 |
+
# Store user details and token in local storage
|
81 |
+
user_data["token"] = token
|
82 |
+
return user_data
|
83 |
+
return {"message": "Invalid email or password"}
|
84 |
|
85 |
#Implement the registration route:
|
86 |
@app.post("/register")
|
87 |
def register(user: User):
|
88 |
+
# Check if user already exists in the database
|
89 |
+
existing_user = users_collection.find_one({"email": user.email})
|
90 |
+
if existing_user:
|
91 |
return {"message": "User already exists"}
|
92 |
+
#Insert the new user into the database
|
93 |
+
user_dict = user.dict()
|
94 |
+
users_collection.insert_one(user_dict)
|
95 |
+
# Generate a token
|
96 |
+
token = generate_token(user.email)
|
97 |
+
# Convert ObjectId to string
|
98 |
+
user_dict["_id"] = str(user_dict["_id"])
|
99 |
+
# Store user details and token in local storage
|
100 |
+
user_dict["token"] = token
|
101 |
+
return user_dict
|
102 |
|
103 |
|
104 |
#Implement the `/api/user` route to fetch user data based on the JWT token
|
105 |
@app.get("/api/user")
|
106 |
def get_user(credentials: HTTPAuthorizationCredentials = Depends(security)):
|
107 |
+
# Extract the token from the Authorization header
|
108 |
+
token = credentials.credentials
|
109 |
+
# Authenticate and retrieve the user data from the database based on the token
|
110 |
+
# Here, you would implement the authentication logic and fetch user details
|
111 |
+
# based on the token from the database or any other authentication mechanism
|
112 |
+
# For demonstration purposes, assuming the user data is stored in local storage
|
113 |
+
# Note: Local storage is not accessible from server-side code
|
114 |
+
# This is just a placeholder to demonstrate the concept
|
115 |
+
user_data = {
|
116 |
+
"username": "John Doe",
|
117 |
+
"email": "johndoe@example.com"
|
118 |
+
}
|
119 |
+
if user_data["username"] and user_data["email"]:
|
120 |
+
return user_data
|
121 |
+
raise HTTPException(status_code=401, detail="Invalid token")
|
122 |
|
123 |
#Define a helper function to generate a JWT token
|
124 |
def generate_token(email: str) -> str:
|
125 |
+
payload = {"email": email}
|
126 |
+
token = jwt_encode(payload, SECRET_KEY, algorithm="HS256")
|
127 |
+
return token
|
128 |
|
129 |
|
130 |
# Get time of request
|
|
|
139 |
|
140 |
full_date = f"{curr_day} | {curr_date} | {curr_time}"
|
141 |
return full_date, curr_time
|
142 |
+
|
143 |
+
|
144 |
+
def load_models():
|
145 |
+
# build model and tokenizer
|
146 |
+
model_name_dict = {
|
147 |
+
#'nllb-distilled-600M': 'facebook/nllb-200-distilled-600M',
|
148 |
+
#'nllb-1.3B': 'facebook/nllb-200-1.3B',
|
149 |
+
#'nllb-distilled-1.3B': 'facebook/nllb-200-distilled-1.3B',
|
150 |
+
#'nllb-3.3B': 'facebook/nllb-200-3.3B',
|
151 |
+
'nllb-moe-54b': 'facebook/nllb-moe-54b',
|
152 |
+
}
|
153 |
+
|
154 |
+
model_dict = {}
|
155 |
+
|
156 |
+
for call_name, real_name in model_name_dict.items():
|
157 |
+
print('\tLoading model: %s' % call_name)
|
158 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(real_name)
|
159 |
+
tokenizer = AutoTokenizer.from_pretrained(real_name)
|
160 |
+
model_dict[call_name+'_model'] = model
|
161 |
+
model_dict[call_name+'_tokenizer'] = tokenizer
|
162 |
+
|
163 |
+
return model_dict
|
164 |
+
|
165 |
|
166 |
# Load the model and tokenizer ..... only once!
|
167 |
beam_size = 1 # change to a smaller value for faster inference
|
|
|
181 |
sp = spm.SentencePieceProcessor()
|
182 |
sp.load(sp_model_full_path)
|
183 |
|
184 |
+
'''
|
185 |
# Import The Translator model
|
186 |
print("\nimporting Translator model")
|
187 |
ct_model_file = "sematrans-3.3B"
|
188 |
ct_model_full_path = os.path.join(os.path.dirname(__file__), ct_model_file)
|
189 |
translator = ctranslate2.Translator(ct_model_full_path, device)
|
190 |
+
'''
|
191 |
+
print("\nimporting Translator model")
|
192 |
+
model_dict = load_models()
|
193 |
|
194 |
+
print('\nDone importing models\n')
|
195 |
|
196 |
|
197 |
def translate_detect(userinput: str, target_lang: str):
|
|
|
244 |
# Return the source language and the translated text
|
245 |
return translations_desubword[0]
|
246 |
|
247 |
+
def translate_faster(userinput3: str, source_lang3: str, target_lang3: str):
|
248 |
+
if len(model_dict) == 2:
|
249 |
+
model_name = 'nllb-moe-54b'
|
250 |
+
|
251 |
+
start_time = time.time()
|
252 |
+
|
253 |
+
model = model_dict[model_name + '_model']
|
254 |
+
tokenizer = model_dict[model_name + '_tokenizer']
|
255 |
+
|
256 |
+
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=source_lang3, tgt_lang=target_lang3)
|
257 |
+
output = translator(userinput3, max_length=400)
|
258 |
+
end_time = time.time()
|
259 |
+
|
260 |
+
output = output[0]['translation_text']
|
261 |
+
result = {'inference_time': end_time - start_time,
|
262 |
+
'source': source,
|
263 |
+
'target': target,
|
264 |
+
'result': output}
|
265 |
+
return result
|
266 |
|
267 |
@app.get("/", response_class=HTMLResponse)
|
268 |
async def read_root(request: Request):
|
|
|
308 |
"translated_text": translated_text_e,
|
309 |
}
|
310 |
|
311 |
+
@app.post("/translate_faster/")
|
312 |
+
async def translate_faster_endpoint(request: Request):
|
313 |
+
dataf = await request.json()
|
314 |
+
userinputf = datae.get("userinput")
|
315 |
+
source_langf = datae.get("source_lang")
|
316 |
+
target_langf = datae.get("target_lang")
|
317 |
+
ffull_date = get_time()[0]
|
318 |
+
print(f"\nrequest: {ffull_date}\nSource_language; {source_langf}, Target Language; {target_langf}, User Input: {userinputf}\n")
|
319 |
+
|
320 |
+
if not userinputf or not target_langf:
|
321 |
+
raise HTTPException(status_code=422, detail="'userinput' 'sourc_lang'and 'target_lang' are required.")
|
322 |
|
323 |
+
translated_text_f = translate_faster(userinputf, source_langf, target_langf)
|
324 |
+
fcurrent_time = get_time()[1]
|
325 |
+
print(f"\nresponse: {fcurrent_time}; ... Translated Text: {translated_text_f}\n\n")
|
326 |
+
return {
|
327 |
+
"translated_text": translated_text_f,
|
328 |
+
}
|
329 |
+
|
330 |
+
print("\nAPI started successfully .......\n")
|