python - 在 Numba @jitclass 中使用列表

Question

我正在模拟一个非常蹩脚的游戏，它基本上会计算玩家在游戏过程中收集的硬币和敌人的数量。该代码包含两个 jitclass:一个 player jitclass 和一个 game jitclass。

对于player类，我们有一些属性和一些方法来描述玩家在游戏中的进展。

from numba import jitclass, int64, float64, deferred_type
from numba.typed import List
import random


specs_player = OrderedDict()
specs_player['level'] = int64
specs_player['coins'] = float64


@jitclass(specs_player)
class Player:

    def __init__(self):
        self.level = 0
        self.coins = 0
        self.enemies = List()

    def pass_level(self):
        self.level += 1

    def collect_coins(self, c):
        self.coins += c

    def collect_enemies(self, e):
        self.enemies.append(e)

    def reset_player(self):
        self.level = 0
        self.coins = 0
        self.enemies = List()

如您所见，属性enemies 是一个列表，它会随着玩家在游戏中的进展而附加值。

game jitclass 使用前两行调用 player jitclass 作为属性:

Player_type = deferred_type()
Player_type.define(Player.class_type.instance_type)

specs_Game = OrderedDict()
specs_Game['last_level'] = int64
specs_Game['diff_threshold'] = float64
specs_Game['player'] = Player_type



class Game:

    def __init__(self, l, t):
        self.player = Player()
        self.last_level = l
        self.diff_threshold = t

    def play_gameround(self):

        random_draw = random.uniform(0, 1)

        if random_draw > self.diff_threshold:
            # Pass Level
            self.player.pass_level()

            # Collect coins
            coins_earned = 100*(random_draw - self.diff_threshold)
            self.player.collect_coins(coins_earned)

            #Collect enemies
            if coins_earned > 10:
                self.player.collect_enemies()


    def reset_gameplay(self):
        self.player.reset_player()

    def continue_playing(self):
        condition = self.player.level < self.last_level

        return(condition)

最后，一个名为 run_one_player 的函数模拟一名玩家的进度并返回一个包含所有必要数据的数组:

def run_one_player(gameplay):
    while gameplay.continue_playing():
        gameplay.play_gameround()

        player_data = np.array([gameplay.player.level,
                                gameplay.player.coins,
                                gameplay.player.enemies])

    return (player_data)

要运行我只需键入的代码:

g = Game(l = 10, t = 0.5)

data_list = run_one_player(g)

但是，这不起作用，Numba 返回以下错误，我很确定这是因为我没有正确定义 enemies 字段的 Numba 类型。

---------------------------------------------------------------------------
TypingError                               Traceback (most recent call last)
<ipython-input-51-f6340d9cfcd7> in <module>
      1 players = 10
----> 2 g = Game(l = 10, t = 0.5)
      3 
      4 data_list = run_one_player(g)

<ipython-input-49-6e42f0104cf1> in __init__(self, l, t)
     46 
     47     def __init__(self, l, t):
---> 48         self.player = Player()
     49         self.last_level = l
     50         self.diff_threshold = t

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/numba/jitclass/base.py in __call__(cls, *args, **kwargs)
    124         bind = cls._ctor_sig.bind(None, *args, **kwargs)
    125         bind.apply_defaults()
--> 126         return cls._ctor(*bind.args[1:], **bind.kwargs)
    127 
    128 

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/numba/dispatcher.py in _compile_for_args(self, *args, **kws)
    374                 e.patch_message(msg)
    375 
--> 376             error_rewrite(e, 'typing')
    377         except errors.UnsupportedError as e:
    378             # Something unsupported is present in the user code, add help info

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/numba/dispatcher.py in error_rewrite(e, issue_type)
    341                 raise e
    342             else:
--> 343                 reraise(type(e), e, None)
    344 
    345         argtypes = []

/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/numba/six.py in reraise(tp, value, tb)
    656             value = tp()
    657         if value.__traceback__ is not tb:
--> 658             raise value.with_traceback(tb)
    659         raise value
    660 

TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Failed in nopython mode pipeline (step: nopython frontend)
Cannot resolve setattr: (instance.jitclass.Player#123b442d0<level:int64,coins:float64>).enemies = ListType[undefined]

File "<ipython-input-49-6e42f0104cf1>", line 18:
    def __init__(self):
        <source elided>
        self.coins = 0
        self.enemies = List()
        ^

[1] During: typing of set attribute 'enemies' at <ipython-input-49-6e42f0104cf1> (18)

File "<ipython-input-49-6e42f0104cf1>", line 18:
    def __init__(self):
        <source elided>
        self.coins = 0
        self.enemies = List()
        ^

[1] During: resolving callee type: jitclass.Player#123b442d0<level:int64,coins:float64>
[2] During: typing of call at <string> (3)

[3] During: resolving callee type: jitclass.Player#123b442d0<level:int64,coins:float64>
[4] During: typing of call at <string> (3)


File "<string>", line 3:
<source missing, REPL/exec in use?>

This is not usually a problem with Numba itself but instead often caused by
the use of unsupported features or an issue in resolving types.

To see Python/NumPy features supported by the latest release of Numba visit:
http://numba.pydata.org/numba-doc/latest/reference/pysupported.html
and
http://numba.pydata.org/numba-doc/latest/reference/numpysupported.html

For more information about typing errors and how to debug them visit:
http://numba.pydata.org/numba-doc/latest/user/troubleshoot.html#my-code-doesn-t-compile

If you think your code should work with Numba, please report the error message
and traceback, along with a minimal reproducer at:
https://github.com/numba/numba/issues/new

Answer 1

首先:我认为你不应该使用 numba 来做这样的事情。 Numba 是一种专门的工具，非常擅长解决特定类型的问题，但这不是其中之一:

1.1.2. Will Numba work for my code?

This depends on what your code looks like, if your code is numerically orientated (does a lot of math), uses NumPy a lot and/or has a lot of loops, then Numba is often a good choice

[...]

---

但是，在您的特定情况下，您需要完整输入 jitclass 的所有属性。这意味着您必须使用 numba 能够理解的类型(支持的类型之一或另一个 jitclass)来输入“enemies”，否则它将无法工作。

由于您没有提供类型，所以我们假设它是一个整数:

import numba as nb

specs_player = {}
specs_player['level'] = nb.int64
specs_player['coins'] = nb.float64
specs_player['enemies'] = nb.types.List(nb.int64)

@nb.jitclass(specs_player)
class Player:
    def __init__(self):
        self.level = 0
        self.coins = 0
        self.enemies = []

当您创建新实例时，此操作会失败，因为 numba 无法推断空列表的类型(至少目前如此)。所以你必须用某种类型来初始化。我还没有找到比创建一个包含项目的列表然后清除它更好的方法:

import numba as nb

specs_player = {}
specs_player['level'] = nb.int64
specs_player['coins'] = nb.float64
specs_player['enemies'] = nb.types.List(nb.int64)

@nb.njit
def empty_int64_list():
    l = [nb.int64(10)]
    l.clear()
    return l

@nb.jitclass(specs_player)
class Player:
    def __init__(self):
        self.level = 0
        self.coins = 0
        self.enemies = empty_int64_list()

如果您的敌人不是整数，情况可能会复杂得多。然而，我认为在你的情况下这是不值得的，因为这不是 numba 比纯 Python 更有效地解决的问题。