python - Numba 给出 : ' TypingError: %r not allowed in a homogenous sequence'

Question

我真的不知道那个错误是关于什么的，因为我什至不直接使用 %r。我怀疑某处有多种类型的混合，但我无法理解会发生在哪里。除了其他加速代码的建议之外，我们将不胜感激。

import numpy as np
from numba import jit, float64

c = 3*10**8
epsilon = 8.854187817 * 10**(-12)
mu = 4*np.pi *10**(-7)

@jit( nopython=True)
def cross(vec1, vec2):

    result = np.array([0.,0.,0.])

    a1, a2, a3 = vec1[0],vec1[1], vec1[2]
    b1, b2, b3 = vec2[0], vec2[1], vec2[2]

    result[0] = a2 * b3 - a3 * b2
    result[1] = a3 * b1 - a1 * b3
    result[2] = a1 * b2 - a2 * b1  

    return result       

@jit( float64[:,:](float64[:],float64,float64,float64[:],float64[:],float64[:]), nopython = True)

def jit_EM_field(position,length,ladung,velocity,acceleration,R):
    #using solutions to lienard wiechert potential
    radius = np.linalg.norm(R - position)
    if radius != 0:
        unitradius = (R - position)/radius
    else:
        unitradius = np.array([0.,0.,0.])

    if radius != 0 and np.dot(unitradius, velocity)!=1:
        charge      = ladung / ( (1 - np.dot(unitradius, velocity)/c)** 3)


        if radius < length:
            radius = length

        radius2     = radius ** 2

        velocity_in_c = velocity/c

        oneMinusV2  = 1 - np.dot(velocity_in_c, velocity_in_c)
        uMinusV     = unitradius - velocity_in_c            
        aCrossUmV   = cross(uMinusV, acceleration)
        Eleft       = (oneMinusV2 * (unitradius - velocity_in_c)) / radius2
        Eright      = cross(unitradius, aCrossUmV) / (radius*c**2)
        E           = (charge/(4*np.pi*epsilon)) * (Eleft - Eright)

        B           = cross(unitradius/c, ((mu*epsilon*charge*c**2) * (Eleft - Eright)))

        EM_field = np.array([E,B], dtype = float)
    else:
        EM_field = np.zeros((2,3), dtype = float)

    return EM_field

jit_EM_field( np.array([0.,1.,0.]),1.,0.1,np.array([0.,1.,0.]),np.array([0.,1.,0.])
,np.array([7.2,5.6,0.1]))

这是完整的错误消息。

runfile('C:/Users/Elios/testingjit.py', wdir='C:/Users/Elios')
Traceback (most recent call last):

  File "<ipython-input-26-221208a798d4>", line 1, in <module>
    runfile('C:/Users/Elios/testingjit.py', wdir='C:/Users/Elios')

  File "C:\Users\Elios\Anaconda4\lib\site-packages\spyder\utils\site\sitecustomize.py", line 866, in runfile
    execfile(filename, namespace)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\spyder\utils\site\sitecustomize.py", line 102, in execfile
    exec(compile(f.read(), filename, 'exec'), namespace)

  File "C:/Users/Elios/testingjit.py", line 32, in <module>
    @jit( float64[:,:](float64[:],float64,float64,float64[:],float64[:],float64[:]), nopython = True)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\decorators.py", line 176, in wrapper
    disp.compile(sig)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\dispatcher.py", line 531, in compile
    cres = self._compiler.compile(args, return_type)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\dispatcher.py", line 80, in compile
    flags=flags, locals=self.locals)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 725, in compile_extra
    return pipeline.compile_extra(func)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 369, in compile_extra
    return self.compile_bytecode(bc, func_attr=self.func_attr)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 378, in compile_bytecode
    return self._compile_bytecode()

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 690, in _compile_bytecode
    return self._compile_core()

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 677, in _compile_core
    res = pm.run(self.status)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 257, in run
    raise patched_exception

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 249, in run
    stage()

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 476, in stage_nopython_frontend
    self.locals)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\compiler.py", line 828, in type_inference_stage
    infer.propagate()

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typeinfer.py", line 717, in propagate
    raise errors[0]

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typeinfer.py", line 127, in propagate
    constraint(typeinfer)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typeinfer.py", line 372, in __call__
    self.resolve(typeinfer, typevars, fnty)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typeinfer.py", line 385, in resolve
    sig = typeinfer.resolve_call(fnty, pos_args, kw_args)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typeinfer.py", line 972, in resolve_call
    return self.context.resolve_function_type(fnty, pos_args, kw_args)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typing\context.py", line 124, in resolve_function_type
    return func.get_call_type(self, args, kws)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\types\functions.py", line 49, in get_call_type
    sig = temp.apply(args, kws)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typing\templates.py", line 216, in apply
    sig = typer(*args, **kws)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typing\npydecl.py", line 456, in typer
    ndim, seq_dtype = _parse_nested_sequence(self.context, object)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typing\npydecl.py", line 423, in _parse_nested_sequence
    n, dtype = _parse_nested_sequence(context, typ.dtype)

  File "C:\Users\Elios\Anaconda4\lib\site-packages\numba\typing\npydecl.py", line 421, in _parse_nested_sequence
    raise TypingError("%r not allowed in a homogenous sequence")

TypingError: %r not allowed in a homogenous sequence

Answer 1

Numba 似乎缺乏对嵌套数组的支持。已设法绕过以下代码中的问题。 jit_EM_field() 现在返回一个长度为 6 的数组，而不是 2 个长度为 3 的嵌套数组，您可以将其视为 @jit( float64[:]在 jit 装饰器中。还删除了一个多余的 else条款。

import numpy as np
from numba import jit, float64

c = 3*10**8
epsilon = 8.854187817 * 10**(-12)
mu = 4*np.pi *10**(-7)

@jit( nopython=True)
def cross(vec1, vec2):
    result = np.array([0.,0.,0.])
    a1, a2, a3 = vec1[0],vec1[1], vec1[2]
    b1, b2, b3 = vec2[0], vec2[1], vec2[2]
    result[0] = a2 * b3 - a3 * b2
    result[1] = a3 * b1 - a1 * b3
    result[2] = a1 * b2 - a2 * b1  
    return result       

@jit( float64[:](float64[:],float64,float64,float64[:],float64[:],float64[:]), nopython=True)
def jit_EM_field(position,length,ladung,velocity,acceleration,R):
    #using solutions to lienard wiechert potential
    EM_field = np.array([0.,0.,0.,0.,0.,0.])
    radius = np.linalg.norm(R - position)
    if radius != 0:
        unitradius = (R - position)/radius
        if np.dot(unitradius, velocity) != 1:
            charge = ladung / ( (1 - np.dot(unitradius, velocity)/c)** 3)
            if radius < length:
                radius = length
            radius2 = radius ** 2
            velocity_in_c = velocity/c
            oneMinusV2 = 1 - np.dot(velocity_in_c, velocity_in_c)
            uMinusV = unitradius - velocity_in_c            
            aCrossUmV = cross(uMinusV, acceleration)
            Eleft = (oneMinusV2 * (unitradius - velocity_in_c)) / radius2
            Eright = cross(unitradius, aCrossUmV) / (radius*c**2)
            E = (charge/(4*np.pi*epsilon)) * (Eleft - Eright)
            B = cross(unitradius/c, ((mu*epsilon*charge*c**2) * (Eleft - Eright)))
            EM_field = np.array([E[0],E[1],E[2],B[0],B[1],B[2]])
    return EM_field

em_field = jit_EM_field(np.array([0.,1.,0.]),1.,0.1,np.array([0.,1.,0.]),np.array([0.,1.,0.]),np.array([7.2,5.6,0.1]))
em_field_zero = jit_EM_field(np.array([0.,1.,0.]),1.,0.1,np.array([0.,1.,0.]),np.array([0.,1.,0.]),np.array([0.,1.,0.]))
import pprint as pp
pp.pprint(em_field)
pp.pprint(em_field_zero)