-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest.py
143 lines (124 loc) · 3.68 KB
/
test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
"""
-*- coding: utf-8 -*-
time: 2023/10/26 21:24
author: suyunsen
email: suyunsen2023@gmail.com
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
import math, copy, time
from torch.autograd import Variable
import matplotlib.pyplot as plt
from typing import List,Optional,Tuple
from torch import Tensor
import numpy as np
def clone(model,N):
return nn.ModuleList([copy.deepcopy(model) for _ in range(N)])
class MyModule(nn.Module):
def __init__(self):
super(MyModule,self).__init__()
# self.linears = nn.ModuleList([nn.Linear(10, 20) for i in range(10)])
self.linears = nn.Linear(10,10)
def forward(self, x):
# ModuleList can act as an iterable, or be indexed using ints
return self.linears(x)
class all(nn.Module):
def __init__(self,modle):
super(all,self).__init__()
self.item = 6
self.modle = clone(modle,4)
def forward(self,x):
x = self.modle(x)
def _scaled_dot_product_attention(
q: Tensor,
k: Tensor,
v: Tensor,
attn_mask: Optional[Tensor] = None,
dropout_p: float = 0.0,
) -> Tuple[Tensor, Tensor]:
B, Nt, E = q.shape
q = q / math.sqrt(E)
attn = torch.bmm(q, k.transpose(-2, -1))
attn = F.softmax(attn, dim=-1)
output = torch.bmm(attn, v)
return output,attn
def subsequent_mask(bs,tar_len,scr_len):
"Mask out subsequent positions."
attn_shape = (bs, tar_len, tar_len)
subsequent_mask = np.triu(np.ones(attn_shape), k=1).astype('uint8')
return torch.from_numpy(subsequent_mask) == 0
if __name__ == '__main__':
# module = MyModule()
# model = all(module)
# a = torch.rand(10)
# res = module(a)
# print(res)
# print(subsequent_mask(3))
# torch.nn.BCELoss
#线下变换
q = torch.rand((5,2,100))
print(q.size(-1))
b = torch.rand((300,100))
c = torch.rand((300))
res = F.linear(q,b,c)
print(res.shape)
# tem = res.chunk(3,dim=-1)[0]
#
# t = tem.contiguous().view(-1,2*4,25)
# print(t)
# print(t.shape)
# print(t.transpose(0, 1).shape)
# tes = True
# if not tes:
# print("hha")
# else:
# print("nihap")
# print(type(module.parameters()))
# idx = 1
# print(sum(p.numel() for p in module.parameters()))
# cnt = sum(p.numel() for p in model.parameters())
# print(cnt)
# q = torch.rand((2,3,100))
# v = torch.rand((2,3,100))
# k = torch.rand((2,3,100))
# q = q.contiguous().view(3, 2 * 4, 25).transpose(0, 1)
# k = k.contiguous().view(-1, 2 * 4, 25).transpose(0, 1)
# v = v.contiguous().view(-1, 2 * 4, 25).transpose(0, 1)
# print(q.shape)
# print(k.shape)
# print(v.shape)
# print(_scaled_dot_product_attention(q,k,v)[0].shape)
# r = torch.tensor([[0, 0, 0, 0], [0, 0, 0, 0]])
# mask = torch.tensor([[1, 1, 1, 1], [0, 0, 0, 0]])
# print(r.masked_fill(mask == 0, -1e9))
# attn_shape = (5, 5)
# print(np.ones(attn_shape))
# res = np.triu(np.ones(attn_shape),k=1)
# print(res)
# v = torch.rand((2,12,100))
# res = subsequent_mask(2,3,3)
# print(res.shape)
# a = torch.randn((2,3,3))
# print(a)
# t = a.masked_fill(res == 0 , -1e9)
# print(t)
# print(F.softmax(t,dim=-1))
#
# # 将输入张量与掩码相乘，实现注意力掩码
# masked_input = input_tensor * attention_mask.unsqueeze(-1)
#
# print(masked_input)
# P = torch.zeros((1, 512, 768))
# print(P.shape[1])
# t = torch.randn((3,5))
# print(t)
# t = t.unsqueeze(dim=1)
# print(t.shape)
# print(t)
# t = t.repeat((1,5,1))
# print(t.shape)
# print(t)
# print(t.contiguous().view(-1,5,3))
# print(t)
# print(t.shape)