|
|
|
@ -5,18 +5,17 @@ import torch.nn as nn
|
|
|
|
|
from typing import Optional |
|
|
|
|
from collections import defaultdict |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
LINE_WIDTH = 108 |
|
|
|
|
LINE = '-' * LINE_WIDTH + '\n' |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class TensorDetector(): |
|
|
|
|
|
|
|
|
|
def __init__(self, |
|
|
|
|
show_info: bool = True, |
|
|
|
|
log: str = None, |
|
|
|
|
include_cpu: bool = False, |
|
|
|
|
module: Optional[nn.Module] = None |
|
|
|
|
): |
|
|
|
|
module: Optional[nn.Module] = None): |
|
|
|
|
"""This class is a detector to detect tensor on different devices. |
|
|
|
|
|
|
|
|
|
Args: |
|
|
|
@ -57,12 +56,12 @@ class TensorDetector():
|
|
|
|
|
|
|
|
|
|
def mem_format(self, real_memory_size): |
|
|
|
|
# format the tensor memory into a reasonal magnitude |
|
|
|
|
if real_memory_size >= 2 ** 30: |
|
|
|
|
return str(real_memory_size / (2 ** 30)) + ' GB' |
|
|
|
|
if real_memory_size >= 2 ** 20: |
|
|
|
|
return str(real_memory_size / (2 ** 20)) + ' MB' |
|
|
|
|
if real_memory_size >= 2 ** 10: |
|
|
|
|
return str(real_memory_size / (2 ** 10)) + ' KB' |
|
|
|
|
if real_memory_size >= 2**30: |
|
|
|
|
return str(real_memory_size / (2**30)) + ' GB' |
|
|
|
|
if real_memory_size >= 2**20: |
|
|
|
|
return str(real_memory_size / (2**20)) + ' MB' |
|
|
|
|
if real_memory_size >= 2**10: |
|
|
|
|
return str(real_memory_size / (2**10)) + ' KB' |
|
|
|
|
return str(real_memory_size) + ' B' |
|
|
|
|
|
|
|
|
|
def collect_tensors_state(self): |
|
|
|
@ -125,8 +124,7 @@ class TensorDetector():
|
|
|
|
|
minus = outdated + minus |
|
|
|
|
if len(self.order) > 0: |
|
|
|
|
for tensor_id in self.order: |
|
|
|
|
self.info += template_format.format('+', |
|
|
|
|
str(self.tensor_info[tensor_id][0]), |
|
|
|
|
self.info += template_format.format('+', str(self.tensor_info[tensor_id][0]), |
|
|
|
|
str(self.tensor_info[tensor_id][1]), |
|
|
|
|
str(tuple(self.tensor_info[tensor_id][2])), |
|
|
|
|
str(self.tensor_info[tensor_id][3]), |
|
|
|
@ -137,8 +135,7 @@ class TensorDetector():
|
|
|
|
|
self.info += '\n' |
|
|
|
|
if len(minus) > 0: |
|
|
|
|
for tensor_id in minus: |
|
|
|
|
self.info += template_format.format('-', |
|
|
|
|
str(self.saved_tensor_info[tensor_id][0]), |
|
|
|
|
self.info += template_format.format('-', str(self.saved_tensor_info[tensor_id][0]), |
|
|
|
|
str(self.saved_tensor_info[tensor_id][1]), |
|
|
|
|
str(tuple(self.saved_tensor_info[tensor_id][2])), |
|
|
|
|
str(self.saved_tensor_info[tensor_id][3]), |
|
|
|
@ -148,7 +145,6 @@ class TensorDetector():
|
|
|
|
|
# deleted the updated tensor |
|
|
|
|
self.saved_tensor_info.pop(tensor_id) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# trace where is the detect() |
|
|
|
|
locate_info = inspect.stack()[2] |
|
|
|
|
locate_msg = '"' + locate_info.filename + '" line ' + str(locate_info.lineno) |
|
|
|
@ -168,7 +164,7 @@ class TensorDetector():
|
|
|
|
|
with open(self.log + '.log', 'a') as f: |
|
|
|
|
f.write(self.info) |
|
|
|
|
|
|
|
|
|
def detect(self, include_cpu = False): |
|
|
|
|
def detect(self, include_cpu=False): |
|
|
|
|
self.include_cpu = include_cpu |
|
|
|
|
self.collect_tensors_state() |
|
|
|
|
self.print_tensors_state() |
|
|
|
|