質問編集履歴
4
誤字の訂正
title
CHANGED
|
File without changes
|
body
CHANGED
|
@@ -101,7 +101,7 @@
|
|
|
101
101
|
-> 1120 result = forward_call(*input, **kwargs)
|
|
102
102
|
1121 if _global_forward_hooks or self._forward_hooks:
|
|
103
103
|
|
|
104
|
-
~/
|
|
104
|
+
~/06rnn_attentionf6/encoder.py in forward(self, sequence, lengths)
|
|
105
105
|
101 rnn_input \
|
|
106
106
|
--> 102 = nn.utils.rnn.pack_padded_sequence(output,
|
|
107
107
|
103 output_lengths.cpu(), #ここを修正
|
3
コードの修正、追加
title
CHANGED
|
File without changes
|
body
CHANGED
|
@@ -1,5 +1,4 @@
|
|
|
1
|
-
すみません。詳しく書く前に要点だけ書かせてください。落ち着いたら詳細を記述します。
|
|
2
|
-
pytorchにおいてtorchsummaryでモデルの構造をファインチューニングするため、
|
|
1
|
+
pytorchにおいてtorchsummaryでモデルの構造をファインチューニングするため、
|
|
3
2
|
`print(summary(model, input_size=([(10,1684,40),(10)])))`
|
|
4
3
|
を実行しました。forward関数は2入力となっているので、引数を２つ入力しています。モデルのforward関数における引数の次元数はそれぞれ３次元と１次元です。しかし、以下のようなエラーが発生しました。:
|
|
5
4
|
|
|
@@ -18,7 +17,9 @@
|
|
|
18
17
|
ご助言いただいたことを元に修正を行いました。
|
|
19
18
|
`print(summary(model, input_size=([(10,1684,40),(10,)])))`
|
|
20
19
|
すると以下のエラーメッセージが出ました。
|
|
20
|
+
```python
|
|
21
|
+
---------------------------------------------------------------------------
|
|
21
|
-
|
|
22
|
+
RuntimeError Traceback (most recent call last)
|
|
22
23
|
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
|
|
23
24
|
267 if isinstance(x, (list, tuple)):
|
|
24
25
|
--> 268 _ = model.to(device)(*x, **kwargs)
|
|
@@ -72,7 +73,66 @@
|
|
|
72
73
|
278 "Failed to run torchinfo. See above stack traces for more details. "
|
|
73
74
|
279 f"Executed layers up to: {executed_layers}"
|
|
74
75
|
|
|
76
|
+
RuntimeError: Failed to run torchinfo. See above stack traces for more details. Executed layers up to:
|
|
77
|
+
```
|
|
78
|
+
また、この後、.cpu()により、CPUのマッチングを行い
|
|
79
|
+
`print(summary(model, input_size=([(10,1684,40),(10,)])))`
|
|
80
|
+
を再度実行しましたが、次は以下のエラーが生じました。
|
|
81
|
+
```python
|
|
82
|
+
---------------------------------------------------------------------------
|
|
83
|
+
RuntimeError Traceback (most recent call last)
|
|
84
|
+
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
|
|
85
|
+
267 if isinstance(x, (list, tuple)):
|
|
86
|
+
--> 268 _ = model.to(device)(*x, **kwargs)
|
|
87
|
+
269 elif isinstance(x, dict):
|
|
88
|
+
|
|
89
|
+
~/.local/lib/python3.8/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
|
|
90
|
+
1101 or _global_forward_hooks or _global_forward_pre_hooks):
|
|
91
|
+
-> 1102 return forward_call(*input, **kwargs)
|
|
92
|
+
1103 # Do not call functions when jit is used
|
|
93
|
+
|
|
94
|
+
~/06rnn_attentionf6/my_model.py in forward(self, input_sequence, input_lengths, label_sequence)
|
|
95
|
+
85 # エンコーダに入力する
|
|
96
|
+
---> 86 enc_out, enc_lengths = self.encoder(input_sequence,
|
|
97
|
+
87 input_lengths)
|
|
98
|
+
|
|
99
|
+
~/.local/lib/python3.8/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
|
|
100
|
+
1119
|
|
101
|
+
-> 1120 result = forward_call(*input, **kwargs)
|
|
102
|
+
1121 if _global_forward_hooks or self._forward_hooks:
|
|
103
|
+
|
|
104
|
+
~/06rnn_attentionf5/encoder.py in forward(self, sequence, lengths)
|
|
105
|
+
101 rnn_input \
|
|
106
|
+
--> 102 = nn.utils.rnn.pack_padded_sequence(output,
|
|
107
|
+
103 output_lengths.cpu(), #ここを修正
|
|
108
|
+
|
|
109
|
+
~/.local/lib/python3.8/site-packages/torch/nn/utils/rnn.py in pack_padded_sequence(input, lengths, batch_first, enforce_sorted)
|
|
110
|
+
248 data, batch_sizes = \
|
|
111
|
+
--> 249 _VF._pack_padded_sequence(input, lengths, batch_first)
|
|
112
|
+
250 return _packed_sequence_init(data, batch_sizes, sorted_indices, None)
|
|
113
|
+
|
|
114
|
+
RuntimeError: Length of all samples has to be greater than 0, but found an element in 'lengths' that is <= 0
|
|
115
|
+
|
|
116
|
+
The above exception was the direct cause of the following exception:
|
|
117
|
+
|
|
118
|
+
RuntimeError Traceback (most recent call last)
|
|
119
|
+
/tmp/ipykernel_715630/614744292.py in <module>
|
|
120
|
+
1 from torchinfo import summary
|
|
121
|
+
----> 2 print(summary(model, input_size=([(10,1684,40),(10,)])))
|
|
122
|
+
|
|
123
|
+
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in summary(model, input_size, input_data, batch_dim, cache_forward_pass, col_names, col_width, depth, device, dtypes, row_settings, verbose, **kwargs)
|
|
124
|
+
199 input_data, input_size, batch_dim, device, dtypes
|
|
125
|
+
200 )
|
|
126
|
+
--> 201 summary_list = forward_pass(
|
|
127
|
+
202 model, x, batch_dim, cache_forward_pass, device, **kwargs
|
|
128
|
+
203 )
|
|
129
|
+
|
|
130
|
+
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
|
|
131
|
+
275 except Exception as e:
|
|
132
|
+
276 executed_layers = [layer for layer in summary_list if layer.executed]
|
|
133
|
+
--> 277 raise RuntimeError(
|
|
134
|
+
278 "Failed to run torchinfo. See above stack traces for more details. "
|
|
135
|
+
279 f"Executed layers up to: {executed_layers}"
|
|
136
|
+
|
|
75
137
|
RuntimeError: Failed to run torchinfo. See above stack traces for more details. Executed layers up to: []
|
|
76
|
-
|
|
77
|
-
コード
|
|
78
138
|
```
|
2
タグの修正
title
CHANGED
|
@@ -1,1 +1,1 @@
|
|
|
1
|
-
|
|
1
|
+
pytorch:モデル可視化,cpuとcudaのデバイス間におけるエラー
|
body
CHANGED
|
File without changes
|
1
入力データを修正
title
CHANGED
|
File without changes
|
body
CHANGED
|
@@ -13,4 +13,66 @@
|
|
|
13
13
|
|
|
14
14
|
二つ目の引数の(10)や(10,20)がlengthsに対応していることはわかりますが、もともとの次元数が１なのに１次元として入力すると上記のようにエラーになってしまうのです。なにか解決策はありますでしょうか。
|
|
15
15
|
|
|
16
|
-
torchsummary は最新版を使用しています。
|
|
16
|
+
torchsummary は最新版を使用しています。
|
|
17
|
+
|
|
18
|
+
ご助言いただいたことを元に修正を行いました。
|
|
19
|
+
`print(summary(model, input_size=([(10,1684,40),(10,)])))`
|
|
20
|
+
すると以下のエラーメッセージが出ました。
|
|
21
|
+
```RuntimeError Traceback (most recent call last)
|
|
22
|
+
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
|
|
23
|
+
267 if isinstance(x, (list, tuple)):
|
|
24
|
+
--> 268 _ = model.to(device)(*x, **kwargs)
|
|
25
|
+
269 elif isinstance(x, dict):
|
|
26
|
+
|
|
27
|
+
~/.local/lib/python3.8/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
|
|
28
|
+
1101 or _global_forward_hooks or _global_forward_pre_hooks):
|
|
29
|
+
-> 1102 return forward_call(*input, **kwargs)
|
|
30
|
+
1103 # Do not call functions when jit is used
|
|
31
|
+
|
|
32
|
+
~/06rnn_attentionf6/my_model.py in forward(self, input_sequence, input_lengths, label_sequence)
|
|
33
|
+
85 # エンコーダに入力する
|
|
34
|
+
---> 86 enc_out, enc_lengths = self.encoder(input_sequence,
|
|
35
|
+
87 input_lengths)
|
|
36
|
+
|
|
37
|
+
~/.local/lib/python3.8/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
|
|
38
|
+
1119
|
|
39
|
+
-> 1120 result = forward_call(*input, **kwargs)
|
|
40
|
+
1121 if _global_forward_hooks or self._forward_hooks:
|
|
41
|
+
|
|
42
|
+
~/06rnn_attentionf6/encoder.py in forward(self, sequence, lengths)
|
|
43
|
+
101 rnn_input \
|
|
44
|
+
--> 102 = nn.utils.rnn.pack_padded_sequence(output,
|
|
45
|
+
103 output_lengths,
|
|
46
|
+
|
|
47
|
+
~/.local/lib/python3.8/site-packages/torch/nn/utils/rnn.py in pack_padded_sequence(input, lengths, batch_first, enforce_sorted)
|
|
48
|
+
248 data, batch_sizes = \
|
|
49
|
+
--> 249 _VF._pack_padded_sequence(input, lengths, batch_first)
|
|
50
|
+
250 return _packed_sequence_init(data, batch_sizes, sorted_indices, None)
|
|
51
|
+
|
|
52
|
+
RuntimeError: 'lengths' argument should be a 1D CPU int64 tensor, but got 1D cuda:0 Long tensor
|
|
53
|
+
|
|
54
|
+
The above exception was the direct cause of the following exception:
|
|
55
|
+
|
|
56
|
+
RuntimeError Traceback (most recent call last)
|
|
57
|
+
/tmp/ipykernel_704668/614744292.py in <module>
|
|
58
|
+
1 from torchinfo import summary
|
|
59
|
+
----> 2 print(summary(model, input_size=([(10,1684,40),(10,)])))
|
|
60
|
+
|
|
61
|
+
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in summary(model, input_size, input_data, batch_dim, cache_forward_pass, col_names, col_width, depth, device, dtypes, row_settings, verbose, **kwargs)
|
|
62
|
+
199 input_data, input_size, batch_dim, device, dtypes
|
|
63
|
+
200 )
|
|
64
|
+
--> 201 summary_list = forward_pass(
|
|
65
|
+
202 model, x, batch_dim, cache_forward_pass, device, **kwargs
|
|
66
|
+
203 )
|
|
67
|
+
|
|
68
|
+
~/.local/lib/python3.8/site-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
|
|
69
|
+
275 except Exception as e:
|
|
70
|
+
276 executed_layers = [layer for layer in summary_list if layer.executed]
|
|
71
|
+
--> 277 raise RuntimeError(
|
|
72
|
+
278 "Failed to run torchinfo. See above stack traces for more details. "
|
|
73
|
+
279 f"Executed layers up to: {executed_layers}"
|
|
74
|
+
|
|
75
|
+
RuntimeError: Failed to run torchinfo. See above stack traces for more details. Executed layers up to: []
|
|
76
|
+
|
|
77
|
+
コード
|
|
78
|
+
```
|