我想實作一個接收單詞串列并回傳張量的函式
這個想法是為每個單詞創建張量,其中每一行都填充零和一個代表該位置的字母。例如,單詞"abc"將由以下張量表示:
tensor([[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]])
uj5u.com熱心網友回復:
您可以使用字典映射的組合來實作這一點,F.one_hot它是執行密集到一個熱轉換的內置函式。
構造字典映射str到索引。我們將插入一個額外的字符
"0"來方便地為短于最大長度的單詞添加一些填充:>>> alphabet = dict(zip('0abcdefghijklmnopqrstuvwxyz', range(27)))找到最大長度并填充較短的單詞:
>>> max_length = max(len(w) for w in words)用適當數量的
0字符填充所有單詞:>>> padded = [w '0'*(max_length - len(w)) for w in words] ['cd0', 'abc']現在通過遍歷所有填充的單詞和字符來構建張量:
>>> dense = torch.tensor([[alphabet[c] for c in w] for w in padded]) tensor([[3, 4, 0], [1, 2, 3]])
最后申請torch.functional.one_hot:
>>> x = F.one_hot(dense, num_classes=27)
tensor([[[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
[[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]])
一直以來,我們都在使用包含26 1值的字典,其中額外的字符"0"用于對填充值進行單熱編碼。我們現在可以洗掉對應于的第一列"0":
>>> x[..., 1:]
tensor([[[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
[[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]])
轉載請註明出處,本文鏈接:https://www.uj5u.com/ruanti/342909.html
