Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,150 +1,96 @@
|
|
1 |
-
import os
|
2 |
from transformers import pipeline
|
3 |
-
|
4 |
-
from queue import Queue
|
5 |
import openai
|
6 |
-
import gc
|
7 |
-
import psutil
|
8 |
-
|
9 |
-
# تعریف مدلها
|
10 |
-
MODEL_CONFIG = {
|
11 |
-
"translation": "PontifexMaximus/opus-mt-iir-en-finetuned-fa-to-en", # ترجمه
|
12 |
-
"qa": "HooshvareLab/bert-fa-base-uncased", # پاسخ به سوال
|
13 |
-
"math": "OpenAI", # ریاضی با OpenAI
|
14 |
-
"persian_nlp": "HooshvareLab/bert-fa-zwnj-base", # پردازش زبان فارسی
|
15 |
-
"custom_ai": "universitytehran/PersianMind-v1.0", # شخصیسازی شده
|
16 |
-
}
|
17 |
-
|
18 |
-
# بارگذاری توکن از متغیر محیطی
|
19 |
-
TOKEN = os.getenv("Passsssssss")
|
20 |
-
if not TOKEN:
|
21 |
-
raise ValueError("API token is missing! Set it as 'Passsssssss' in environment variables.")
|
22 |
-
|
23 |
-
# تنظیم کلید API OpenAI
|
24 |
-
openai.api_key = os.getenv("OPENAI_API_KEY")
|
25 |
-
if not openai.api_key:
|
26 |
-
raise ValueError("OpenAI API key is missing! Set it as 'OPENAI_API_KEY' in environment variables.")
|
27 |
-
|
28 |
-
# محدودیت حافظه (برحسب گیگابایت)
|
29 |
-
MEMORY_LIMIT_GB = 15
|
30 |
-
|
31 |
-
class MemoryManager:
|
32 |
-
@staticmethod
|
33 |
-
def check_memory_usage():
|
34 |
-
"""بررسی مصرف حافظه و خروج در صورت تجاوز از محدودیت."""
|
35 |
-
process = psutil.Process(os.getpid())
|
36 |
-
memory_usage = process.memory_info().rss / (1024**3) # تبدیل به گیگابایت
|
37 |
-
if memory_usage > MEMORY_LIMIT_GB:
|
38 |
-
raise MemoryError(f"Memory usage exceeded {MEMORY_LIMIT_GB} GB! Current usage: {memory_usage:.2f} GB.")
|
39 |
|
40 |
class MultiModelSystem:
|
41 |
def __init__(self):
|
42 |
-
"""مدیریت مدلهای مختلف و پردازش موازی وظایف."""
|
43 |
self.models = {}
|
44 |
-
self.queue = Queue()
|
45 |
-
self.executor = ThreadPoolExecutor(max_workers=3) # محدود کردن وظایف همزمان
|
46 |
-
self.load_models()
|
47 |
-
|
48 |
-
def load_models(self):
|
49 |
-
"""بارگذاری مدلها به صورت دینامیک."""
|
50 |
-
try:
|
51 |
-
for task, model_id in MODEL_CONFIG.items():
|
52 |
-
if model_id == "OpenAI":
|
53 |
-
self.models[task] = None # مدل OpenAI نیازی به بارگذاری ندارد
|
54 |
-
else:
|
55 |
-
self.models[task] = model_id # ذخیره نام مدل، نه بارگذاری کامل
|
56 |
-
except Exception as e:
|
57 |
-
print(f"Error loading models: {e}")
|
58 |
-
raise
|
59 |
|
60 |
-
def
|
61 |
-
"""
|
62 |
-
model_id
|
63 |
-
|
64 |
-
|
65 |
-
try:
|
66 |
-
MemoryManager.check_memory_usage()
|
67 |
return pipeline(
|
68 |
task=self.get_task_type(task),
|
69 |
model=model_id,
|
70 |
-
use_auth_token=
|
71 |
-
device=0 if psutil.cpu_count(logical=False) > 1 else -1 # استفاده از GPU در صورت وجود
|
72 |
)
|
73 |
-
|
74 |
-
|
75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
|
77 |
@staticmethod
|
78 |
def get_task_type(task):
|
79 |
-
"""بازگرداندن نوع وظیفه براساس مدل."""
|
80 |
task_map = {
|
81 |
"translation": "translation",
|
82 |
"qa": "question-answering",
|
83 |
"persian_nlp": "text-classification",
|
84 |
"custom_ai": "text-generation",
|
85 |
-
"math": "text-generation"
|
86 |
}
|
87 |
return task_map.get(task, "text-generation")
|
88 |
|
89 |
-
def process_task(self, task, **kwargs):
|
90 |
-
"""
|
91 |
if task not in self.models:
|
92 |
-
|
93 |
-
|
94 |
-
def task_handler():
|
95 |
-
try:
|
96 |
-
if task == "math":
|
97 |
-
result = self.process_math_task(kwargs.get("text"))
|
98 |
-
else:
|
99 |
-
model = self.load_model_for_task(task)
|
100 |
-
if model:
|
101 |
-
result = model(**kwargs)
|
102 |
-
else:
|
103 |
-
result = None
|
104 |
-
self.queue.put(result)
|
105 |
-
except Exception as e:
|
106 |
-
print(f"Error processing task '{task}': {e}")
|
107 |
-
self.queue.put(None)
|
108 |
-
finally:
|
109 |
-
gc.collect() # آزادسازی حافظه پس از هر وظیفه
|
110 |
|
111 |
-
|
|
|
|
|
|
|
|
|
112 |
|
113 |
def process_math_task(self, text):
|
114 |
-
"""
|
115 |
try:
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
prompt=text,
|
120 |
-
max_tokens=100
|
121 |
)
|
122 |
-
return response
|
123 |
except Exception as e:
|
124 |
-
print(f"Error processing math task
|
125 |
return None
|
126 |
|
127 |
-
|
128 |
-
|
129 |
-
|
|
|
130 |
|
131 |
# نمونه استفاده
|
132 |
if __name__ == "__main__":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
133 |
system = MultiModelSystem()
|
134 |
|
135 |
-
# وظایف مختلف
|
136 |
tasks = [
|
137 |
-
{"task": "translation", "kwargs": {"text": "سلام دنیا!", "src_lang": "fa", "tgt_lang": "en"}},
|
138 |
-
{"task": "qa", "kwargs": {"question": "پایتخت ایران چیست؟", "context": "ایران کشوری در خاورمیانه است و پایتخت آن تهران است."}},
|
139 |
-
{"task": "math", "kwargs": {"text": "What is the integral of x^2?"}},
|
140 |
-
{"task": "persian_nlp", "kwargs": {"text": "این یک جمله فارسی است."}},
|
141 |
-
{"task": "custom_ai", "kwargs": {"text": "تحلیل متنهای تاریخی طبری."}}
|
142 |
]
|
143 |
|
144 |
-
# ارسال وظایف برای پردازش
|
145 |
for task_info in tasks:
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
|
|
|
|
1 |
from transformers import pipeline
|
2 |
+
import os
|
|
|
3 |
import openai
|
4 |
+
import gc # برای آزاد کردن حافظه
|
5 |
+
import psutil # برای مانیتور کردن حافظه
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
class MultiModelSystem:
|
8 |
def __init__(self):
|
|
|
9 |
self.models = {}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
+
def load_model(self, task, model_id, use_auth_token=None):
|
12 |
+
"""مدلها را به صورت lazy بارگذاری میکند."""
|
13 |
+
if model_id == "OpenAI":
|
14 |
+
return self.load_openai_model()
|
15 |
+
else:
|
|
|
|
|
16 |
return pipeline(
|
17 |
task=self.get_task_type(task),
|
18 |
model=model_id,
|
19 |
+
use_auth_token=use_auth_token
|
|
|
20 |
)
|
21 |
+
|
22 |
+
def unload_model(self, task):
|
23 |
+
"""مدل بارگذاری شده را از حافظه پاک میکند."""
|
24 |
+
if task in self.models:
|
25 |
+
del self.models[task]
|
26 |
+
gc.collect() # جمعآوری حافظه
|
27 |
+
|
28 |
+
@staticmethod
|
29 |
+
def load_openai_model():
|
30 |
+
"""مدل ریاضی OpenAI."""
|
31 |
+
return "OpenAI (Math)"
|
32 |
|
33 |
@staticmethod
|
34 |
def get_task_type(task):
|
|
|
35 |
task_map = {
|
36 |
"translation": "translation",
|
37 |
"qa": "question-answering",
|
38 |
"persian_nlp": "text-classification",
|
39 |
"custom_ai": "text-generation",
|
40 |
+
"math": "text-generation",
|
41 |
}
|
42 |
return task_map.get(task, "text-generation")
|
43 |
|
44 |
+
def process_task(self, task, model_id, **kwargs):
|
45 |
+
"""مدیریت وظایف."""
|
46 |
if task not in self.models:
|
47 |
+
self.models[task] = self.load_model(task, model_id, use_auth_token=os.getenv("Passsssssss"))
|
48 |
+
model = self.models[task]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
+
# پردازش وظیفه
|
51 |
+
if task == "math":
|
52 |
+
return self.process_math_task(kwargs.get("text"))
|
53 |
+
else:
|
54 |
+
return model(**kwargs)
|
55 |
|
56 |
def process_math_task(self, text):
|
57 |
+
"""مدیریت وظایف ریاضی OpenAI."""
|
58 |
try:
|
59 |
+
response = openai.ChatCompletion.create(
|
60 |
+
model="gpt-4",
|
61 |
+
messages=[{"role": "user", "content": text}]
|
|
|
|
|
62 |
)
|
63 |
+
return response['choices'][0]['message']['content'].strip()
|
64 |
except Exception as e:
|
65 |
+
print(f"Error processing math task: {e}")
|
66 |
return None
|
67 |
|
68 |
+
# مانیتورینگ حافظه
|
69 |
+
def check_memory_usage():
|
70 |
+
mem = psutil.virtual_memory()
|
71 |
+
print(f"Memory usage: {mem.percent}% ({mem.used / (1024 ** 3):.2f} GB used)")
|
72 |
|
73 |
# نمونه استفاده
|
74 |
if __name__ == "__main__":
|
75 |
+
MODEL_CONFIG = {
|
76 |
+
"translation": "PontifexMaximus/opus-mt-iir-en-finetuned-fa-to-en",
|
77 |
+
"qa": "HooshvareLab/bert-fa-base-uncased",
|
78 |
+
"math": "OpenAI",
|
79 |
+
"persian_nlp": "HooshvareLab/bert-fa-zwnj-base",
|
80 |
+
"custom_ai": "universitytehran/PersianMind-v1.0",
|
81 |
+
}
|
82 |
+
|
83 |
system = MultiModelSystem()
|
84 |
|
|
|
85 |
tasks = [
|
86 |
+
{"task": "translation", "model_id": MODEL_CONFIG["translation"], "kwargs": {"text": "سلام دنیا!", "src_lang": "fa", "tgt_lang": "en"}},
|
87 |
+
{"task": "qa", "model_id": MODEL_CONFIG["qa"], "kwargs": {"question": "پایتخت ایران چیست؟", "context": "ایران کشوری در خاورمیانه است و پایتخت آن تهران است."}},
|
88 |
+
{"task": "math", "model_id": MODEL_CONFIG["math"], "kwargs": {"text": "What is the integral of x^2?"}},
|
|
|
|
|
89 |
]
|
90 |
|
|
|
91 |
for task_info in tasks:
|
92 |
+
check_memory_usage() # نمایش میزان حافظه
|
93 |
+
result = system.process_task(task_info["task"], task_info["model_id"], **task_info["kwargs"])
|
94 |
+
print(f"Result for task '{task_info['task']}':", result)
|
95 |
+
system.unload_model(task_info["task"]) # آزادسازی حافظه
|
96 |
+
check_memory_usage() # دوباره بررسی حافظه
|