expected device cpu but got device cuda:0

Question

import pickle
import docx
import re
import os
from bistmCRF import BiLSTM_CRF
import torch.optim as optim
import torch
import time
EPOCHES =3
EMBEDDING_DIM = 300
HIDDEN_DIM = 500
BATCH_SIZE = 64
learning_rate = 0.001
labels_list = {"SUB": 0, "TEM": 1, "LOC": 2, "RAI": 3, "ACT": 4, "ATT": 5}
delet = ["-1", "-2", "-3", "-4", "-P", "-C", "-W", "-S", "-R", "V", '-L']
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
def extract(path):
    fulllist = []
    contents = []
    f = docx.Document(path)
    parg = f.paragraphs
    for p in parg:
        fulllist.append(p.text)
    fulllist = fulllist[5:len(fulllist)-4]
    for list in fulllist:
        contents += re.findall(r'[\[](.*?)[\]]', list)
    lables = []
    data = {}
    for content in contents:
        content = content.replace(" ", "")
        for lable in labels_list.keys():
            if lable in content:
                lables.append(lable)
                content = content.replace(lable, "")
                if content[:2] in delet:
                    content = content[2:]
                data[content] = lable
                break
    return data

def apart_label_data(content):
    content_label, content_data = [], []
    for con in content:
        content_label += con.values()
        content_data += con.keys()
    return content_data, content_label
def to_ix_processing(pairs):
    word_to_ix = {}
    tag_to_ix= {"<START>": 0, "<STOP>": 1, "<UNK>": 2}
    for pair in pairs:
        for word, tags in pair.items():
            if word not in word_to_ix:
                word_to_ix[word] = len(word_to_ix)
            if tags not in tag_to_ix:
                tag_to_ix[tags] = len(tag_to_ix)
    return word_to_ix, tag_to_ix

def prepare_sequence(datas, word_to_ix):
    idx = []
    for w in datas:
        if w not in word_to_ix.keys():
            w = 'UNK'
        idx.append(word_to_ix[w])
    return torch.tensor(idx, dtype=torch.long)

def trains(model, optimizer, train_data, train_label, to_ix_w, to_ix_l, epoch):
    model.train()
    model = model.cuda()
    total_loss = 0
    start_time = time.time()
    model.zero_grad()
    # prepare_sequence是将数据转换成对应的idx
    sentence_in = prepare_sequence(train_data, to_ix_w).cuda()
    tragets = prepare_sequence(train_label, to_ix_l).cuda()
    loss = model.neg_log_likelihood(sentence_in, tragets).cuda()
    total_loss += loss.item() / len(sentence_in)
    log_interval = 20
    cur_loss = total_loss / log_interval
    elapse = time.time() - start_time

    total_loss = 0
    start_time = time.time()
    loss.backward()
    optimizer.step()

def evaluate(eval_model, test_data, test_label, to_ix_w, to_ix_l):
    eval_model.eval()
    total_loss = 0
    sentence_in = prepare_sequence(test_data, to_ix_w).cuda()
    tragets = prepare_sequence(test_label, to_ix_l).cuda()
    with torch.no_grad():
        loss_test = eval_model.neg_log_likelihood(sentence_in, tragets).cuda()
        total_loss += float(loss_test) / len(sentence_in)
    return total_loss
def main():
    data_all = []
    path = '标注案情750'
    files = os.listdir(path)
    pairs = []
    #将原本结构化数据抽取出来
    for file in files:
        file_dir = os.path.join(path, file)
        data = extract(file_dir)
        data_all.append(data)
    to_ix_w, to_ix_l = to_ix_processing(data_all)

    #定义模型和优化器
    model = BiLSTM_CRF(len(to_ix_w), to_ix_l, EMBEDDING_DIM, HIDDEN_DIM)
    optimizer = optim.Adam(model.parameters(), lr=learning_rate)
    model = model.cuda()
    #将数据划分为8：2的训练集和测试集
    start = int(len(data_all) * 0.8)
    train = data_all[:start]
    test = data_all[start:]
    #将训练集和测试集数据和标签分开
    train_data, train_label = apart_label_data(train)
    test_data, test_label = apart_label_data(test)
    for epoch in range(EPOCHES):
        trains(model, optimizer, train_data, train_label, to_ix_w, to_ix_l, epoch)
        val_loss = evaluate(model, test_data, test_label, to_ix_w, to_ix_l)
        print('|epoch{:3d}|' "loss{:5.5f}".format(epoch, val_loss))


if __name__== '__main__':
    main()

实在找不到问题，求大佬们帮忙！！！！

 

Answer 1

第16行device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")

改为device = torch.device("cpu")

试试