import numpy as np # from tensorflow import keras X = np.random.randint(100, size=10*1000).reshape(1000, 10) Y = np.random.randint(100, size=15*1000).reshape(1000, 15) X_decoder = np.c_[np.zeros((1000, 1)), Y[:, :-1]] seq_lengths = np.full([1000], 15) print(X) # history = model.fit([X, X_decoder, seq_lengths], Y, epochs=2) '''(1,3) 和 (1,3) 变 (1,6)''' qq = np.c_[np.array([[1,2,3]]), 0, 0, np.array([[4,5,6]])] print(qq)
D:ProgramDataMiniconda3python.exe E:/新脚本主文件夹/训练测试项目/test3/测试乱七八糟的3.py [[65 97 98 ... 78 15 79] [92 18 68 ... 6 41 21] [ 9 20 49 ... 95 99 91] ... [29 17 53 ... 6 68 41] [48 72 2 ... 62 6 22] [60 3 19 ... 55 74 49]] [[1 2 3 0 0 4 5 6]] Process finished with exit code 0
np.c_ 这个玩意儿不是个方法，而是返回一个类然后直接和后面的ndarry相乘。。。相当于concatenate(【】,dim):
c_ = concatenate([np.zeros((1000, 1)), Y[:, :-1]],1) print(c_)
r_和c_里面源码是这样的，虽然我看不太懂。。。。
class AxisConcatenator: """ Translates slice objects to concatenation along an axis. For detailed documentation on usage, see `r_`. """ # allow ma.mr_ to override this concatenate = staticmethod(_nx.concatenate) makemat = staticmethod(matrixlib.matrix) def __init__(self, axis=0, matrix=False, ndmin=1, trans1d=-1): self.axis = axis self.matrix = matrix self.trans1d = trans1d self.ndmin = ndmin def __getitem__(self, key): # handle matrix builder syntax if isinstance(key, str): frame = sys._getframe().f_back mymat = matrixlib.bmat(key, frame.f_globals, frame.f_locals) return mymat if not isinstance(key, tuple): key = (key,) # copy attributes, since they can be overridden in the first argument trans1d = self.trans1d ndmin = self.ndmin matrix = self.matrix axis = self.axis objs = [] scalars = [] arraytypes = [] scalartypes = [] for k, item in enumerate(key): scalar = False if isinstance(item, slice): step = item.step start = item.start stop = item.stop if start is None: start = 0 if step is None: step = 1 if isinstance(step, complex): size = int(abs(step)) newobj = linspace(start, stop, num=size) else: newobj = _nx.arange(start, stop, step) if ndmin > 1: newobj = array(newobj, copy=False, ndmin=ndmin) if trans1d != -1: newobj = newobj.swapaxes(-1, trans1d) elif isinstance(item, str): if k != 0: raise ValueError("special directives must be the " "first entry.") if item in ('r', 'c'): matrix = True col = (item == 'c') continue if ',' in item: vec = item.split(',') try: axis, ndmin = [int(x) for x in vec[:2]] if len(vec) == 3: trans1d = int(vec[2]) continue except Exception as e: raise ValueError( "unknown special directive {!r}".format(item) ) from e try: axis = int(item) continue except (ValueError, TypeError): raise ValueError("unknown special directive") elif type(item) in ScalarType: newobj = array(item, ndmin=ndmin) scalars.append(len(objs)) scalar = True scalartypes.append(newobj.dtype) else: item_ndim = ndim(item) newobj = array(item, copy=False, subok=True, ndmin=ndmin) if trans1d != -1 and item_ndim < ndmin: k2 = ndmin - item_ndim k1 = trans1d if k1 < 0: k1 += k2 + 1 defaxes = list(range(ndmin)) axes = defaxes[:k1] + defaxes[k2:] + defaxes[k1:k2] newobj = newobj.transpose(axes) objs.append(newobj) if not scalar and isinstance(newobj, _nx.ndarray): arraytypes.append(newobj.dtype) # Ensure that scalars won't up-cast unless warranted final_dtype = find_common_type(arraytypes, scalartypes) if final_dtype is not None: for k in scalars: objs[k] = objs[k].astype(final_dtype) res = self.concatenate(tuple(objs), axis=axis) if matrix: oldndim = res.ndim res = self.makemat(res) if oldndim == 1 and col: res = res.T return res def __len__(self): return 0 # separate classes are used here instead of just making r_ = concatentor(0), # etc. because otherwise we couldn't get the doc string to come out right # in help(r_)
注意下倒数第三行源码有错误，就是那个注释，里面写的concatentor应该是concatenate
然后调用的类很简单，如下，在同一个文件里，indextricks.py
class CClass(AxisConcatenator): """ Translates slice objects to concatenation along the second axis. This is short-hand for ``np.r_['-1,2,0', index expression]``, which is useful because of its common occurrence. In particular, arrays will be stacked along their last axis after being upgraded to at least 2-D with 1's post-pended to the shape (column vectors made out of 1-D arrays). See Also -------- column_stack : Stack 1-D arrays as columns into a 2-D array. r_ : For more detailed documentation. Examples -------- >>> np.c_[np.array([1,2,3]), np.array([4,5,6])] array([[1, 4], [2, 5], [3, 6]]) >>> np.c_[np.array([[1,2,3]]), 0, 0, np.array([[4,5,6]])] array([[1, 2, 3, ..., 4, 5, 6]]) """ def __init__(self): AxisConcatenator.__init__(self, -1, ndmin=2, trans1d=0) c_ = CClass()
就是返回了一个上面的类，然而很高级的样子，也看不太懂。。。。