File size: 2,899 Bytes
18fa92b |
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 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 |
import gc
import psutil
import numpy as np
import torch
import logging
class MemoryManager:
def __init__(self):
self.allocated_memory = 0
self.max_memory_limit = psutil.virtual_memory().total * 0.8 # 80% of total RAM
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
def optimize_allocation(self):
"""Free up unused memory."""
gc.collect()
if torch.cuda.is_available():
torch.cuda.empty_cache()
logging.info("Memory optimization completed.")
def allocate_memory(self, size, dtype=np.float32):
"""Allocate memory for a given size and data type."""
if self.allocated_memory + size > self.max_memory_limit:
logging.warning("Memory limit exceeded. Optimizing memory...")
self.optimize_allocation()
self.allocated_memory += size
logging.info(f"Allocated {size} bytes of memory. Total allocated: {self.allocated_memory} bytes.")
return np.zeros(size, dtype=dtype)
def deallocate_memory(self, obj):
"""Deallocate memory for a given object."""
del obj
self.optimize_allocation()
logging.info("Object deallocated and memory optimized.")
def monitor_usage(self):
"""Monitor system memory usage and optimize if necessary."""
usage = psutil.virtual_memory().percent
if usage > 85:
logging.warning(f"High memory usage detected: {usage}%. Optimizing memory...")
self.optimize_allocation()
def adaptive_caching(self, model):
"""Disable gradients for large models to reduce memory usage."""
model_size = sum(p.numel() for p in model.parameters() if p.requires_grad)
if model_size * 4 > self.max_memory_limit: # 4 bytes per float32
for param in model.parameters():
param.requires_grad = False
logging.info("Adaptive caching applied: Gradients disabled for large model.")
return model
def __enter__(self):
"""Context manager entry point."""
return self
def __exit__(self, exc_type, exc_val, exc_tb):
"""Context manager exit point."""
self.optimize_allocation()
logging.info("MemoryManager context exited. Memory optimized.")
# Example usage
if __name__ == "__main__":
memory_manager = MemoryManager()
# Allocate memory
array = memory_manager.allocate_memory(1000000) # 1 million elements
print(array.shape)
# Deallocate memory
memory_manager.deallocate_memory(array)
# Monitor usage
memory_manager.monitor_usage()
# Adaptive caching for a model
model = torch.nn.Linear(1000, 1000)
model = memory_manager.adaptive_caching(model)
# Using context manager
with MemoryManager() as mm:
array = mm.allocate_memory(1000000)
print(array.shape) |