Skip to content

P
pytensor
提交 7a3dbfd4 authored 作者: Roy Xue's avatar Roy Xue
浏览文件
操作

For a better output

上级 dc6847e9
显示空白字符变更
内嵌 并排
正在显示 包含 40 行增加 和 62 行删除
theano/compile/profiling.py
浏览文件 @ 7a3dbfd4
...@@ -636,6 +636,24 @@ class ProfileStats(object): ...@@ -636,6 +636,24 @@ class ProfileStats(object):
new_max_node_memory_saved_by_inplace = 0 new_max_node_memory_saved_by_inplace = 0
def count_running_memory(order, thunk_old_storage, nodes_mem): def count_running_memory(order, thunk_old_storage, nodes_mem):
"""
Calculate memory with specific node order
Return a list including the following values
1. node_memory_size
Sum of the size of all variables that actually allocate
memory (excluding views, and inplace);
2. running_memory_size
The memory allocated after the current apply node
3. running_max_memory_size
The maximum of running_memory_size during the function
4. node_memory_saved_by_view
The sum of memory saved by returning view instead of new
allocation
5. node_memory_saved_by_inplace
The sum of memory saved by reusing the input instead of
new allocation
"""
for node in order: for node in order:
val = nodes_mem[node] val = nodes_mem[node]
dmap = getattr(node.op, 'destroy_map', None) dmap = getattr(node.op, 'destroy_map', None)
...@@ -659,39 +677,12 @@ class ProfileStats(object): ...@@ -659,39 +677,12 @@ class ProfileStats(object):
if not isinstance(old_v, str): if not isinstance(old_v, str):
running_memory_size -= old_v running_memory_size -= old_v
return node_memory_size, running_memory_size, running_max_memory_size, node_memory_saved_by_inplace, node_memory_saved_by_view return [node_memory_size, running_memory_size, running_max_memory_size, node_memory_saved_by_inplace, node_memory_saved_by_view]
for fgraph, nodes_mem in fct_memory.iteritems(): for fgraph, nodes_mem in fct_memory.iteritems():
# Sum of the size of all variables in bytes # Sum of the size of all variables in bytes
sum_size = sum([sum([v for v in val if not isinstance(v, str)]) sum_size = sum([sum([v for v in val if not isinstance(v, str)])
for key, val in nodes_mem.iteritems()]) for key, val in nodes_mem.iteritems()])
# Sum of the size of all variables that actually allocate
# memory (excluding views, and inplace);
node_memory_size = 0
# The sum of memory saved by returning view instead of new
# allocation
node_memory_saved_by_view = 0
# The sum of memory saved by reusing the input instead of
# new allocation
node_memory_saved_by_inplace = 0
# The memory allocated after the current apply node
running_memory_size = 0
# The maximum of running_memory_size during the function
running_max_memory_size = 0
new_node_memory_size = 0
# The sum of memory saved by returning view instead of new
# allocation with new node executed order
new_node_memory_saved_by_view = 0
# The sum of memory saved by reusing the input instead of
# new allocation with new node executed order
new_node_memory_saved_by_inplace = 0
# The memory allocated after the current apply node with
# new node executed order
new_running_memory_size = 0
# The maximum of running_memory_size during the function
# with new node executed order
new_running_max_memory_size = 0
order = fgraph.toposort() order = fgraph.toposort()
# A list of intermediate variable that are not need # A list of intermediate variable that are not need
...@@ -708,33 +699,33 @@ class ProfileStats(object): ...@@ -708,33 +699,33 @@ class ProfileStats(object):
(input not in fgraph.outputs) and (input not in fgraph.outputs) and
node == last_user[input]]) node == last_user[input]])
node_memory_size, running_memory_size, running_max_memory_size, node_memory_saved_by_view, node_memory_saved_by_inplace = count_running_memory(order, post_thunk_old_storage, nodes_mem) old_running_memory = count_running_memory(order, post_thunk_old_storage, nodes_mem)
new_order = fgraph.profile.node_executed_order new_order = fgraph.profile.node_executed_order
# A list of new executed node order # A list of new executed node order
new_storage = fgraph.profile.node_cleared_order new_storage = fgraph.profile.node_cleared_order
# A list of variables that get freed # A list of variables that get freed
new_node_memory_size, new_running_memory_size, new_running_max_memory_size, new_node_memory_saved_by_view, new_node_memory_saved_by_inplace = count_running_memory(new_order, new_storage, nodes_mem) new_running_memory = count_running_memory(new_order, new_storage, nodes_mem)
# Store the max of some stats by any function in this profile. # Store the max of some stats by any function in this profile.
max_sum_size = max(max_sum_size, sum_size) max_sum_size = max(max_sum_size, sum_size)
max_node_memory_size = max(max_node_memory_size, node_memory_size) max_node_memory_size = max(max_node_memory_size, old_running_memory[0])
max_running_max_memory_size = max(max_running_max_memory_size, max_running_max_memory_size = max(max_running_max_memory_size,
running_max_memory_size) old_running_memory[2])
max_node_memory_saved_by_view = max(max_node_memory_saved_by_view, max_node_memory_saved_by_view = max(max_node_memory_saved_by_view,
node_memory_saved_by_view) old_running_memory[4])
max_node_memory_saved_by_inplace = max( max_node_memory_saved_by_inplace = max(
max_node_memory_saved_by_inplace, node_memory_saved_by_inplace) max_node_memory_saved_by_inplace, old_running_memory[3])
# Store max of some stats with new order # Store max of some stats with new order
new_max_node_memory_size = max(new_max_node_memory_size, new_node_memory_size) new_max_node_memory_size = max(new_max_node_memory_size, new_running_memory[0])
new_max_running_max_memory_size = max(new_max_running_max_memory_size, new_max_running_max_memory_size = max(new_max_running_max_memory_size,
new_running_max_memory_size) new_running_memory[2])
new_max_node_memory_saved_by_view = max(new_max_node_memory_saved_by_view, new_max_node_memory_saved_by_view = max(new_max_node_memory_saved_by_view,
new_node_memory_saved_by_view) new_running_memory[4])
new_max_node_memory_saved_by_inplace = max( new_max_node_memory_saved_by_inplace = max(
new_max_node_memory_saved_by_inplace, new_node_memory_saved_by_inplace) new_max_node_memory_saved_by_inplace, new_running_memory[3])
del fgraph, nodes_mem, post_thunk_old_storage, node del fgraph, nodes_mem, post_thunk_old_storage, node
...@@ -745,38 +736,25 @@ class ProfileStats(object): ...@@ -745,38 +736,25 @@ class ProfileStats(object):
print >> file, "Memory Profile" print >> file, "Memory Profile"
print >> file, "(Sparse variables are ignored)" print >> file, "(Sparse variables are ignored)"
print >> file, "(For values in brackets, it's for linker = c|py"
print >> file, "---" print >> file, "---"
# print >> file, " Max if no gc, inplace and view: %dKB" % int( # print >> file, " Max if no gc, inplace and view: %dKB" % int(
# round(max_sum_size / 1024)) # round(max_sum_size / 1024))
print >> file, " Max if linker=cvm (default): unknown" print >> file, " Max if linker=cvm (default): %dKB" %int(round(
new_max_node_memory_size / 1024.))
print >> file, " Max if no gc (allow_gc=False): %dKB" % int(round( print >> file, " Max if no gc (allow_gc=False): %dKB" % int(round(
max_node_memory_size / 1024.)) max_node_memory_size / 1024.))
print >> file, " Max if linker=c|py: %dKB" % int(round( print >> file, " Max if linker=c|py: %dKB" % int(round(
max_running_max_memory_size / 1024.)) max_running_max_memory_size / 1024.))
print >> file, " Memory saved if views are used: %dKB" % int( print >> file, " Memory saved if views are used: %dKB (%dKB)" % (int(
round(max_node_memory_saved_by_view / 1024.)) round(new_max_node_memory_saved_by_view / 1024.)), int(
print >> file, " Memory saved if inplace ops are used: %dKB" % \ round(max_node_memory_saved_by_view / 1024.)))
int(round(max_node_memory_saved_by_inplace / 1024.)) print >> file, " Memory saved if inplace ops are used: %dKB (%dKB)" % \
print >> file, " Memory saved if gc is enabled (linker=c|py): %dKB" % int( (int(round(new_max_node_memory_saved_by_inplace / 1024.)), int(round(max_node_memory_saved_by_inplace / 1024.)))
round(max_node_memory_size - max_running_max_memory_size) / 1024.) print >> file, " Memory saved if gc is enabled: %dKB (%dKB)" % (int(
round(new_max_node_memory_size - new_max_running_max_memory_size) / 1024.), int(
print >> file, " Memory Profile with new executed node order" round(max_node_memory_size - max_running_max_memory_size) / 1024.))
print >> file, "---"
# print >> file, " Max if no gc, inplace and view: %dKB" % int(
# round(max_sum_size / 1024))
print >> file, " Max if linker=cvm (default): unknown"
print >> file, " Max if no gc (allow_gc=False): %dKB" % int(round(
new_max_node_memory_size / 1024.))
print >> file, " Max if linker=c|py: %dKB" % int(round(
new_max_running_max_memory_size / 1024.))
print >> file, " Memory saved if views are used: %dKB" % int(
round(new_max_node_memory_saved_by_view / 1024.))
print >> file, " Memory saved if inplace ops are used: %dKB" % \
int(round(new_max_node_memory_saved_by_inplace / 1024.))
print >> file, " Memory saved if gc is enabled (linker=c|py): %dKB" % int(
round(new_max_node_memory_size - new_max_running_max_memory_size) / 1024.)
if (hasattr(theano, 'sandbox') and if (hasattr(theano, 'sandbox') and
hasattr(theano.sandbox, 'cuda') and hasattr(theano.sandbox, 'cuda') and
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论