英文:
RuntimeError: Expected all tensors to be on the same device, but found at least two devices
问题
RuntimeError: 预期所有张量位于相同设备上，但发现至少有两个设备，cuda:0 和 cpu！(在检查方法wrapper_mm中参数mat2的参数时)
我尝试在GPU上运行此代码并打印标签，但出现了此错误。我的训练代码如下：
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # 设备为cuda
with open('intents.json') as f:
intents = json.load(f)
file = 'data.pth'
data = torch.load(file)
input_size = data['input_size']
model_state = data['model_state']
output_size = data['output_size']
hidden_size = data['hidden_size']
all_words = data['all_words']
tags = data['tags']
model = NeuralNetwork(input_size, hidden_size, output_size)
model.load_state_dict(model_state)
model.eval()
@jit(target_backend='cuda')
def get_response(pattern):
sentence = tokenize(pattern)
BoW = bag_of_word(sentence, all_words)
BoW = torch.from_numpy(BoW).to(device)
output = model.forward_propagation(BoW)
# print(output)
_, predicted = torch.max(output, dim=-1)
tag = tags[predicted.item()] # 给出输入语音的预测标签
# print(tag)
probs = torch.softmax(output, dim=-1) # 使输出概率在-1和1之间
# print(props)
prob = probs[predicted.item()] # 选择最大概率
# print(prob)
return prob, tag
pattern = speech_to_text()
prob, tag = get_response(pattern)
print(tag)
英文:
I have gotten the response from the chatbot using GPU, i get the following errors:
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument mat2 in method wrapper_mm)
I tried running this code on GPU and print tag but I get on this error.
My training code is as follows:
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # device = cuda
with open('intents.json') as f:
intents = json.load(f)
file = 'data.pth'
data = torch.load(file)
input_size = data['input_size']
model_state = data['model_state']
output_size = data['output_size']
hidden_size = data['hidden_size']
all_words = data['all_words']
tags = data['tags']
model = NeuralNetwork(input_size,hidden_size,output_size)
model.load_state_dict(model_state)
model.eval()
@jit(target_backend='cuda')
def get_response(pattern):
sentence = tokenize(pattern)
BoW = bag_of_word(sentence,all_words)
BoW = torch.from_numpy(BoW).to(device)
output = model.forward_propagation(BoW)
# print(output)
_,predicted = torch.max(output,dim=-1)
tag = tags[predicted.item()] # give prediction tag for input speech
# print(tag)
probs = torch.softmax(output,dim=-1) # to make output probability between -1 and 1
# print(props)
prob = probs[predicted.item()] # to select the big probability
# print(prob)
return prob,tag
pattern = speech_to_text()
prob,tag = get_response(pattern)
print(tag)
答案1
得分: 1
将模型 to() 移动到设备上：
model = NeuralNetwork(input_size, hidden_size, output_size).to(device)
英文:
Move the model to() the device:
model = NeuralNetwork(input_size, hidden_size, output_size).to(device)
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论