File size: 2,372 Bytes
8296d63 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 | # dynamic_gpu_predictor.py
from dynamic_predictor import DynamicPredictor
from batch_size_optimizer import BatchSizeOptimizer
from workload_scheduler import WorkloadScheduler
from performance_monitor import PerformanceMonitor
class DynamicGPUPredictor:
def __init__(self, model_path='models/gradient_boosting_model.joblib',
memory_limit_mb=8000, num_gpus=1):
# Initialize components
self.predictor = DynamicPredictor(model_path)
self.batch_optimizer = BatchSizeOptimizer(self.predictor, memory_limit_mb)
self.scheduler = WorkloadScheduler(self.predictor, num_gpus)
self.monitor = PerformanceMonitor(self.predictor)
def predict_and_optimize(self, model_features):
"""Predict execution time and optimize batch size"""
# Find optimal batch size
optimal_batch = self.batch_optimizer.find_optimal_batch_size(model_features)
# Update features with optimal batch size
optimized_features = model_features.copy()
optimized_features['batch_size'] = optimal_batch
# Make prediction
predicted_time = self.predictor.predict(optimized_features)
return {
'optimal_batch_size': optimal_batch,
'predicted_execution_time': predicted_time,
'estimated_memory_usage': self.batch_optimizer.estimate_memory_usage(
model_features, optimal_batch)
}
def schedule_model(self, model_features, job_id, priority=1):
"""Schedule a model execution on available GPUs"""
return self.scheduler.schedule_job(model_features, job_id, priority)
def record_actual_performance(self, model_name, batch_size, predicted_time, actual_time):
"""Record actual performance and detect anomalies"""
is_anomaly = self.monitor.record_performance(
model_name, batch_size, predicted_time, actual_time)
if is_anomaly:
# Rebalance workload if anomaly detected
self.scheduler.rebalance_workload()
return is_anomaly
def get_performance_insights(self, model_name, batch_size):
"""Get performance insights for a specific model"""
return self.monitor.get_performance_trend(model_name, batch_size)
|