search

649453932 / Chinese-Text-Classification-Pytorch

中文文本分类，TextCNN，TextRNN，FastText，TextRCNN，BiLSTM_Attention，DPCNN，Transformer，基于pytorch，开箱即用。
MIT License
5.25k stars 1.22k forks source link

请问是否可以写一个预测类 #41

Open blue9735 opened 4 years ago

myechona commented 4 years ago

自己写了个函数：

    if args.status == 'predict':
        line = "union select or 1=1"
        vocab = build_vocab(train_path, args.word, args.vocab_size, args.min_freq)
        n_vocab = len(vocab)
        model = Model(n_vocab, args.num_classes, args.embed, args.num_filters, args.filter_sizes, args.dropout).to(device)
        print(model.parameters)
        # print(model.predict(vocab, args.word, line))
        #predict(model, vocab, args.word, args.max_sequence_length, line, '1', device)

        if args.word:
            tokenizer = lambda x: x.split(' ')
        else:
            tokenizer = lambda x: [y for y in x]
        content = tokenizer(line)
        data = [vocab[x] for x in content if x in vocab]
        data = kr.preprocessing.sequence.pad_sequences([data], 32)

        tests = [np.asarray(data)]
        test_batch = Variable(torch.LongTensor(tests))

        # 其一
        predict = model(test_batch).data.max(1, keepdim=True)[1]
        label = predict[0][0]
        print(label)

        #其二
        outputs = model(test_batch)
        predic = torch.max(outputs.data, 1)[1].cpu().numpy()
        print(predic)
ArtificialZeng commented 4 years ago

自己写了个函数：

    if args.status == 'predict':
        line = "union select or 1=1"
        vocab = build_vocab(train_path, args.word, args.vocab_size, args.min_freq)
        n_vocab = len(vocab)
        model = Model(n_vocab, args.num_classes, args.embed, args.num_filters, args.filter_sizes, args.dropout).to(device)
        print(model.parameters)
        # print(model.predict(vocab, args.word, line))
        #predict(model, vocab, args.word, args.max_sequence_length, line, '1', device)

        if args.word:
            tokenizer = lambda x: x.split(' ')
        else:
            tokenizer = lambda x: [y for y in x]
        content = tokenizer(line)
        data = [vocab[x] for x in content if x in vocab]
        data = kr.preprocessing.sequence.pad_sequences([data], 32)

        tests = [np.asarray(data)]
        test_batch = Variable(torch.LongTensor(tests))

        # 其一
        predict = model(test_batch).data.max(1, keepdim=True)[1]
        label = predict[0][0]
        print(label)

        #其二
        outputs = model(test_batch)
        predic = torch.max(outputs.data, 1)[1].cpu().numpy()
        print(predic)

在本项目中用不了