train-modle / database /database.py
fokan's picture
Force Space rebuild v2.1.0 with incremental training
5be0e59
"""
Database manager for the AI Knowledge Distillation Platform
"""
import sqlite3
import logging
from pathlib import Path
from typing import Dict, Any, List, Optional
from datetime import datetime
logger = logging.getLogger(__name__)
class DatabaseManager:
"""
Centralized database manager for all platform data
"""
def __init__(self, db_dir: str = "database"):
"""
Initialize database manager
Args:
db_dir: Directory for database files
"""
self.db_dir = Path(db_dir)
self.db_dir.mkdir(parents=True, exist_ok=True)
# Database file paths
self.tokens_db = self.db_dir / "tokens.db"
self.training_db = self.db_dir / "training_sessions.db"
self.performance_db = self.db_dir / "performance_metrics.db"
self.medical_db = self.db_dir / "medical_datasets.db"
# Initialize all databases
self._init_all_databases()
logger.info("Database Manager initialized")
def _init_all_databases(self):
"""Initialize all database schemas"""
self._init_tokens_database()
self._init_training_database()
self._init_performance_database()
self._init_medical_database()
def _init_tokens_database(self):
"""Initialize tokens database"""
with sqlite3.connect(self.tokens_db) as conn:
conn.execute('''
CREATE TABLE IF NOT EXISTS tokens (
id INTEGER PRIMARY KEY AUTOINCREMENT,
name TEXT UNIQUE NOT NULL,
token_type TEXT NOT NULL,
encrypted_token TEXT NOT NULL,
is_default BOOLEAN DEFAULT FALSE,
description TEXT,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
last_used TIMESTAMP,
usage_count INTEGER DEFAULT 0,
is_active BOOLEAN DEFAULT TRUE
)
''')
conn.execute('''
CREATE TABLE IF NOT EXISTS token_usage_log (
id INTEGER PRIMARY KEY AUTOINCREMENT,
token_name TEXT NOT NULL,
operation TEXT NOT NULL,
timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
success BOOLEAN,
error_message TEXT
)
''')
conn.commit()
def _init_training_database(self):
"""Initialize training sessions database"""
with sqlite3.connect(self.training_db) as conn:
conn.execute('''
CREATE TABLE IF NOT EXISTS training_sessions (
id INTEGER PRIMARY KEY AUTOINCREMENT,
session_id TEXT UNIQUE NOT NULL,
teacher_model TEXT NOT NULL,
student_model TEXT NOT NULL,
dataset_name TEXT,
training_type TEXT NOT NULL,
status TEXT DEFAULT 'initialized',
progress REAL DEFAULT 0.0,
current_step INTEGER DEFAULT 0,
total_steps INTEGER,
current_loss REAL,
best_loss REAL,
learning_rate REAL,
batch_size INTEGER,
temperature REAL,
alpha REAL,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
started_at TIMESTAMP,
completed_at TIMESTAMP,
error_message TEXT,
config_json TEXT
)
''')
conn.execute('''
CREATE TABLE IF NOT EXISTS training_logs (
id INTEGER PRIMARY KEY AUTOINCREMENT,
session_id TEXT NOT NULL,
step INTEGER NOT NULL,
loss REAL,
learning_rate REAL,
memory_usage_mb REAL,
timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
additional_metrics TEXT
)
''')
conn.commit()
def _init_performance_database(self):
"""Initialize performance metrics database"""
with sqlite3.connect(self.performance_db) as conn:
conn.execute('''
CREATE TABLE IF NOT EXISTS system_metrics (
id INTEGER PRIMARY KEY AUTOINCREMENT,
timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
cpu_usage_percent REAL,
memory_usage_mb REAL,
memory_usage_percent REAL,
available_memory_gb REAL,
disk_usage_percent REAL,
temperature_celsius REAL
)
''')
conn.execute('''
CREATE TABLE IF NOT EXISTS model_performance (
id INTEGER PRIMARY KEY AUTOINCREMENT,
model_name TEXT NOT NULL,
operation TEXT NOT NULL,
duration_seconds REAL,
memory_peak_mb REAL,
throughput_samples_per_second REAL,
accuracy REAL,
timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
additional_metrics TEXT
)
''')
conn.commit()
def _init_medical_database(self):
"""Initialize medical datasets database"""
with sqlite3.connect(self.medical_db) as conn:
conn.execute('''
CREATE TABLE IF NOT EXISTS medical_datasets (
id INTEGER PRIMARY KEY AUTOINCREMENT,
dataset_name TEXT UNIQUE NOT NULL,
repo_id TEXT NOT NULL,
description TEXT,
size_gb REAL,
num_samples INTEGER,
modalities TEXT,
specialties TEXT,
languages TEXT,
last_accessed TIMESTAMP,
access_count INTEGER DEFAULT 0,
is_cached BOOLEAN DEFAULT FALSE,
cache_path TEXT,
metadata_json TEXT
)
''')
conn.execute('''
CREATE TABLE IF NOT EXISTS dicom_files (
id INTEGER PRIMARY KEY AUTOINCREMENT,
file_path TEXT UNIQUE NOT NULL,
patient_id TEXT,
study_date TEXT,
modality TEXT,
file_size_mb REAL,
processed BOOLEAN DEFAULT FALSE,
processed_at TIMESTAMP,
metadata_json TEXT
)
''')
conn.commit()
def get_connection(self, db_name: str) -> sqlite3.Connection:
"""Get database connection"""
db_map = {
'tokens': self.tokens_db,
'training': self.training_db,
'performance': self.performance_db,
'medical': self.medical_db
}
if db_name not in db_map:
raise ValueError(f"Unknown database: {db_name}")
return sqlite3.connect(db_map[db_name])
def execute_query(self, db_name: str, query: str, params: tuple = ()) -> List[tuple]:
"""Execute query and return results"""
with self.get_connection(db_name) as conn:
cursor = conn.execute(query, params)
return cursor.fetchall()
def execute_update(self, db_name: str, query: str, params: tuple = ()) -> int:
"""Execute update query and return affected rows"""
with self.get_connection(db_name) as conn:
cursor = conn.execute(query, params)
conn.commit()
return cursor.rowcount
def backup_databases(self, backup_dir: str = "backups") -> Dict[str, str]:
"""Create backup of all databases"""
backup_path = Path(backup_dir)
backup_path.mkdir(parents=True, exist_ok=True)
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
backup_files = {}
db_files = {
'tokens': self.tokens_db,
'training': self.training_db,
'performance': self.performance_db,
'medical': self.medical_db
}
for db_name, db_file in db_files.items():
if db_file.exists():
backup_file = backup_path / f"{db_name}_{timestamp}.db"
# Copy database file
import shutil
shutil.copy2(db_file, backup_file)
backup_files[db_name] = str(backup_file)
logger.info(f"Backed up {db_name} database to {backup_file}")
return backup_files
def get_database_stats(self) -> Dict[str, Any]:
"""Get statistics about all databases"""
stats = {}
db_files = {
'tokens': self.tokens_db,
'training': self.training_db,
'performance': self.performance_db,
'medical': self.medical_db
}
for db_name, db_file in db_files.items():
if db_file.exists():
file_size_mb = db_file.stat().st_size / (1024**2)
# Get table counts
try:
with self.get_connection(db_name) as conn:
cursor = conn.execute(
"SELECT name FROM sqlite_master WHERE type='table'"
)
tables = [row[0] for row in cursor.fetchall()]
table_counts = {}
for table in tables:
cursor = conn.execute(f"SELECT COUNT(*) FROM {table}")
count = cursor.fetchone()[0]
table_counts[table] = count
stats[db_name] = {
'file_size_mb': file_size_mb,
'tables': table_counts,
'total_records': sum(table_counts.values())
}
except Exception as e:
stats[db_name] = {
'file_size_mb': file_size_mb,
'error': str(e)
}
else:
stats[db_name] = {
'file_size_mb': 0,
'status': 'not_created'
}
return stats
def cleanup_old_data(self, days_to_keep: int = 30) -> Dict[str, int]:
"""Cleanup old data from databases"""
cutoff_date = datetime.now().timestamp() - (days_to_keep * 24 * 3600)
cleanup_stats = {}
try:
# Cleanup old performance metrics
with self.get_connection('performance') as conn:
cursor = conn.execute(
"DELETE FROM system_metrics WHERE timestamp < ?",
(cutoff_date,)
)
cleanup_stats['system_metrics'] = cursor.rowcount
conn.commit()
# Cleanup old training logs
with self.get_connection('training') as conn:
cursor = conn.execute(
"DELETE FROM training_logs WHERE timestamp < ?",
(cutoff_date,)
)
cleanup_stats['training_logs'] = cursor.rowcount
conn.commit()
# Cleanup old token usage logs
with self.get_connection('tokens') as conn:
cursor = conn.execute(
"DELETE FROM token_usage_log WHERE timestamp < ?",
(cutoff_date,)
)
cleanup_stats['token_usage_log'] = cursor.rowcount
conn.commit()
logger.info(f"Cleaned up old data: {cleanup_stats}")
except Exception as e:
logger.error(f"Error cleaning up old data: {e}")
cleanup_stats['error'] = str(e)
return cleanup_stats