Skip to content

P
pytensor
提交 6b0a7569 authored 作者: Roy Xue's avatar Roy Xue
浏览文件
操作

Merge pull request #5 from nouiz/GSoC2014_part2

G so c2014 part2
上级 0776ceb1 3c31f11c
隐藏空白字符变更
内嵌 并排
正在显示 包含 118 行增加 和 58 行删除
theano/compile/profiling.py
浏览文件 @ 6b0a7569
......@@ -671,16 +671,23 @@ class ProfileStats(object):
node_memory_saved_by_inplace = 0
dependencies = fgraph.profile.dependencies
# Initial compute_map which is used to check if a node is valid
compute_map = defaultdict(lambda: [0])
for var in fgraph.inputs:
compute_map[var][0] = 1
# two data structure used to mimic Python gc
viewed_by = {}# {var1: [vars that view var1]}
viewed_by = {} # {var1: [vars that view var1]}
# The len of the list is the value of python ref count. But we use a list, not just the ref count value.
# This is more safe to help detect potential bug in the algo
for var in fgraph.variables:
viewed_by[var] = []
view_of = {} # {var1: original var viewed by var1}
view_of = {} # {var1: original var viewed by var1}
# The orignal mean that we don't keep trac of all the intermediate relationship in the view.
for node in order:
for var in node.outputs:
compute_map[var][0] = 1
idx = 0
dmap = getattr(node.op, 'destroy_map', None)
vmap = getattr(node.op, 'view_map', None)
......@@ -695,40 +702,63 @@ class ProfileStats(object):
node_memory_saved_by_view += v
idx += 1
# Update the Python emulating dicts and add the memory allocated by the node
# Update the Python emulating dicts and add the memory
# allocated by the node
idx2 = 0
for out in node.outputs:
if (dmap and idx2 in dmap) or (vmap and idx2 in vmap):
# This is needed for destroy_map in case it return a partial view that is destroyed.
# So the output could be different then the input.
for ins in node.inputs:
assert isinstance(ins, theano.Variable)
view_of[out] = view_of.get(ins, ins)# This get make that we keep trac of view only again the original
viewed_by[ins].append(out)
ins = None
if dmap and idx2 in dmap:
vidx = dmap[idx2]
assert len(vidx) == 1, "Here we only support the possibility to destroy one input"
ins = node.inputs[vidx[0]]
if vmap and idx2 in vmap:
assert ins is None
vidx = vmap[idx2]
assert len(vidx) == 1, "Here we only support the possibility to view one input"
ins = node.inputs[vidx[0]]
if ins is not None:
# This is needed for destroy_map in case it
# return a partial view that is destroyed. So
# the output could be different then the
# input.
assert isinstance(ins, theano.Variable)
# we keep trac of view only again the origin
origin = view_of.get(ins, ins)
view_of[out] = origin
viewed_by[origin].append(out)
else:
running_memory_size += var_mem[out]
node_memory_size += var_mem[out]
idx2 += 1
running_max_memory_size = max(running_max_memory_size, running_memory_size)
running_max_memory_size = max(running_max_memory_size,
running_memory_size)
# Mimic the combination of Theano and Python gc
for ins in node.inputs:
assert not (ins in view_of and viewed_by[ins])
# we keep trac of the original var, so this shouldn't happen
if dependencies[ins] and ins not in fgraph.outputs and ins.owner:
# we trac the original var, so this shouldn't happen
if (dependencies[ins] and
ins not in fgraph.outputs and
ins.owner and
all([compute_map[v][0] for v in dependencies[ins]])):
if ins not in view_of and not viewed_by.get(ins, []):
running_memory_size -= var_mem[ins]
elif ins in view_of:
origin = view_of[ins]
viewed_by[origin].remove(ins)
if not viewed_by[origin] and origin not in fgraph.inputs:
if (not viewed_by[origin] and
origin not in fgraph.inputs):
running_memory_size -= var_mem[origin]
else:
# ins is viewed_by something else, so its memory isn't freed
# ins is viewed_by something else, so its
# memory isn't freed
pass
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]
def count_minimum_peak(node_list, fgraph, nodes_mem):
global mem_count, mem_bound, max_mem_count
......@@ -770,15 +800,6 @@ class ProfileStats(object):
if c != "output" and check_node_state(c):
executable_nodes.add(c)
# two data structure used to mimic Python gc
viewed_by = {}# {var1: [vars that view var1]}
# The len of the list is the value of python ref count. But we use a list, not just the ref count value.
# This is more safe to help detect potential bug in the algo
for var in fgraph.variables:
viewed_by[var] = []
view_of = {}# {var1: original var viewed by var1}
# The orignal mean that we don't keep trac of all the intermediate relationship in the view.
def min_memory_generator(executable_nodes, viewed_by, view_of):
"""
Generate all valid node order from node_list
......@@ -796,8 +817,13 @@ class ProfileStats(object):
if max_mem_count > mem_bound:
continue
viewed_by_temp = viewed_by.copy()
view_of_temp = view_of.copy()
# We don't want a shallow copy, but we don't want
# a deep copy. So this do a "middle" copy, where
# we copy the dict and the list, but not the var
viewed_by_temp = {}
for k, v in viewed_by.iteritems():
viewed_by_temp[k] = list(v)
for var in node.outputs:
compute_map[var][0] = 1
......@@ -810,15 +836,29 @@ class ProfileStats(object):
vmap = getattr(node.op, 'view_map', None)
idx = 0
# Update the Python emulating dicts and add the memory allocated by the node
# Update the Python emulating dicts and add the
# memory allocated by the node
for out in node.outputs:
if (dmap and idx in dmap) or (vmap and idx in vmap):
# This is needed for destroy_map in case it return a partial view that is destroyed.
# So the output could be different then the input.
for ins in node.inputs:
assert isinstance(ins, theano.Variable)
view_of_temp[out] = view_of_temp.get(ins, ins)# This get make that we keep trac of view only again the original
viewed_by_temp[ins].append(out)
ins = None
if dmap and idx in dmap:
vidx = dmap[idx]
assert len(vidx) == 1, "Here we only support the possibility to destroy one input"
ins = node.inputs[vidx[0]]
if vmap and idx in vmap:
assert ins is None, "Here we only support the possibility to view one input"
vidx = vmap[idx]
assert len(vidx) == 1
ins = node.inputs[vidx[0]]
if ins is not None:
# This is needed for destroy_map in case it
# return a partial view that is destroyed. So
# the output could be different then the
# input.
assert isinstance(ins, theano.Variable)
# We keep trac of view only again the original
origin = view_of_temp.get(ins, ins)
view_of_temp[out] = origin
viewed_by_temp[origin].append(out)
else:
mem_created += var_mem[out]
idx += 1
......@@ -828,19 +868,23 @@ class ProfileStats(object):
# Mimic the combination of Theano and Python gc.
for ins in node.inputs:
assert not (ins in view_of_temp and viewed_by_temp[ins])
# we keep track of the original var, so this shouldn't happen
if dependencies[ins] and ins not in fgraph.outputs and ins.owner:
if all(compute_map[v] for v in dependencies[ins]):
if ins not in view_of_temp and not viewed_by_temp.get(ins, []):
mem_freed += var_mem[ins]
elif ins in view_of_temp:
origin = view_of_temp[ins]
viewed_by_temp[origin].remove(ins)
if not viewed_by_temp[origin] and origin not in fgraph.inputs:
mem_freed += var_mem[origin]
assert not (ins in view_of_temp and
viewed_by_temp[ins])
# We track of the original var, so this shouldn't happen
if (dependencies[ins] and
ins not in fgraph.outputs and
ins.owner and
all([compute_map[v][0] for v in dependencies[ins]])):
if ins not in view_of_temp and not viewed_by_temp.get(ins, []):
mem_freed += var_mem[ins]
elif ins in view_of_temp:
origin = view_of_temp[ins]
viewed_by_temp[origin].remove(ins)
if not viewed_by_temp[origin] and origin not in fgraph.inputs:
mem_freed += var_mem[origin]
else:
# ins is viewed_by something else, so its memory isn't freed
# ins is viewed_by something else, so its
# memory isn't freed
pass
mem_count -= mem_freed
......@@ -852,13 +896,15 @@ class ProfileStats(object):
if not new_exec_nodes:
yield [node]
#Check and Update mem_bound
# Check and Update mem_bound
if max_mem_count < mem_bound:
mem_bound = max_mem_count
else:
for p in min_memory_generator(new_exec_nodes, viewed_by_temp, view_of_temp):
for p in min_memory_generator(new_exec_nodes,
viewed_by_temp,
view_of_temp):
yield [node]+p
# Reset track variables
mem_count -= mem_created
max_mem_count = max_storage
......@@ -866,8 +912,19 @@ class ProfileStats(object):
for var in node.outputs:
compute_map[var][0] = 0
# two data structure used to mimic Python gc
viewed_by = {} # {var1: [vars that view var1]}
# The len of the list is the value of python ref count. But we use a list, not just the ref count value.
# This is more safe to help detect potential bug in the algo
for var in fgraph.variables:
viewed_by[var] = []
view_of = {} # {var1: original var viewed by var1}
# The orignal mean that we don't keep trac of all the intermediate relationship in the view.
# Loop all valid orders and find min peak(store in mem_bound)
for order in min_memory_generator(executable_nodes, viewed_by, view_of):
for order in min_memory_generator(executable_nodes,
viewed_by,
view_of):
continue
return mem_bound
......@@ -875,7 +932,7 @@ class ProfileStats(object):
for fgraph, nodes_mem in fct_memory.iteritems():
# Sum of the size of all variables in bytes
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()])
order = fgraph.toposort()
# A list of intermediate variable that are not need
......@@ -888,24 +945,27 @@ class ProfileStats(object):
new_order = fgraph.profile.node_executed_order
# A list of new executed node order
new_running_memory = count_running_memory(new_order, fgraph, nodes_mem)
new_running_memory = count_running_memory(new_order,
fgraph, nodes_mem)
# Store the max of some stats by any function in this profile.
max_sum_size = max(max_sum_size, sum_size)
max_node_memory_size = max(max_node_memory_size, old_running_memory[0])
max_node_memory_size = max(max_node_memory_size,
old_running_memory[0])
max_running_max_memory_size = max(max_running_max_memory_size,
old_running_memory[2])
old_running_memory[2])
max_node_memory_saved_by_view = max(max_node_memory_saved_by_view,
old_running_memory[4])
max_node_memory_saved_by_inplace = max(
max_node_memory_saved_by_inplace, old_running_memory[3])
# Store max of some stats with new order
new_max_node_memory_size = max(new_max_node_memory_size, new_running_memory[0])
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_running_memory[2])
new_running_memory[2])
new_max_node_memory_saved_by_view = max(new_max_node_memory_saved_by_view,
new_running_memory[4])
new_running_memory[4])
new_max_node_memory_saved_by_inplace = max(
new_max_node_memory_saved_by_inplace, new_running_memory[3])
......
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