refactor: simplify class to function

revert-12646-feature/jit/xbyak
chenweihang 7 years ago
parent 999d097bbb
commit ea548a794b

@ -23,9 +23,7 @@ This API is still under active development and may change drastically.
from .. import core from .. import core
from ..framework import Program, Variable from ..framework import Program, Variable
__all__ = ['MemoryInfo'] __all__ = ['memory_usage']
DEBUG = False
dtype_to_size = { dtype_to_size = {
core.VarDesc.VarType.FP16: 2, core.VarDesc.VarType.FP16: 2,
@ -38,62 +36,67 @@ dtype_to_size = {
core.VarDesc.VarType.UINT8: 1, core.VarDesc.VarType.UINT8: 1,
} }
DEBUG = False
class MemoryInfo(object):
def __init__(self, program):
if not isinstance(program, Program):
raise TypeError(
"Calculating Memory Usage requires Program as its Parameter."
"But you passed in %s" % (type(prgram)))
self._program = program
def _has_var(self, block, var_name): def memory_usage(program, batch_size):
return block.has_var(str(var_name)) """
Get the estimate memory usage of program with input batch size.
def _find_var(self, block, var_name): Args:
return block.var(str(var_name)) program(Program): The current Program.
batch_size(int): The current input data batch_size.
Returns:
min_total_memory(float): the estimate memory usage lower bound.
max_total_memory(float): the estimate memory usage upper bound.
unit_str(string): the unit of estimate usage result.
Examples:
>>> import paddle.fluid as fluid
>>> lower_usage, upper_usage, unit = fluid.contrib.memory_usage(
fluid.default_main_program(), batch_size=10)
>>> print "memory usage is about %.3f - %.3f %s" % \
(lower_usage, upper_usage, unit)
def get_memory_usage(self, batch_size, with_details=False): """
# get the first block of program # Parameters check
first_block = self._program.global_block() if not isinstance(program, Program):
raise TypeError(
"Calculating Memory Usage requires Program as its Parameter."
"But you passed in %s" % (type(prgram)))
if batch_size <= 0:
raise ValueError("The batch size need to be positive.")
# get the var_name list of first block # Get the var_name list of first block and calculate
# TODO(chenweihang): not find the API get block's var list directly total_memory = 0.0
total_memory = 0.0 for var in program.global_block().vars.itervalues():
for var in self._program.list_vars(): data_count = 1
if DEBUG: for x in var.shape:
print "All Block's Var: %s" % (var.name) if x == -1:
# TODO(chenweihang): why not used program.list_vars() data_count *= batch_size
# calculate all variable's memory directly? else:
if self._has_var(first_block, var.name): data_count *= x
if DEBUG: var_memory = data_count * dtype_to_size[var.dtype]
print "First Block's Var: %s" % (var.name) if DEBUG:
print "Var's shape: ", var.shape print "%s memory usage: %d" % (var.name, var_memory)
print "Var's dtype: ", var.dtype total_memory += var_memory
data_count = 1 if DEBUG:
for x in var.shape: print "total memory usage: %.2f" % (total_memory)
if x == -1:
data_count *= batch_size
else:
data_count *= x
var_memory = data_count * dtype_to_size[var.dtype]
if DEBUG:
print "Var's memory: %d" % (var_memory)
total_memory += var_memory
# Convert unit and make result string # Convert appropriate unit
result_str = "- With current batch size, memory usage is about " unit_str = "B"
unit_str = " B." if total_memory > 1024:
total_memory /= 1024
unit_str = "KB"
if total_memory > 1024: if total_memory > 1024:
total_memory /= 1024 total_memory /= 1024
unit_str = " KB." unit_str = "MB"
if total_memory > 1024:
total_memory /= 1024
unit_str = " MB."
# Append extra memory consumption (5% - 10%) # Append extra memory consumption (5% - 10%)
result_str += str(round(total_memory * 1.05, 3)) + " - " \ min_total_memory = total_memory * 1.05
+ str(round(total_memory * 1.10, 3)) + unit_str max_total_memory = total_memory * 1.1
return result_str return min_total_memory, max_total_memory, unit_str

Loading…
Cancel
Save