pytorch训练数据以及测试全部代码(2)

p&＃61;{‘trainBatch’:6, &＃39;nAveGrad&＃39;:1, &＃39;lr&＃39;:1e-07, &＃39;wd&＃39;:0.0005, &＃39;momentum&＃39;:0.9,&＃39;epoch_size&＃39;:10, &＃39;optimizer&＃39;:&＃39;SGD()&＃39;}最后一个optimizer的值是很长的字符串就不全部写出来了。这个字典长度是7。

其中的net 和criterion在稍后来进行讲解

if resume_epoch&＃61;&＃61;0&＃xff0c;那么从头开始训练 training from scratch&＃xff1b;否则权重的初始化时一个已经训练好的模型&＃xff0c;使用net.load_state_dict函数&＃xff0c;这个函数是在torch.nn.Module类里面定义的一个函数。

def load_state_dict(self, state_dict, strict&＃61;True):r"""Copies parameters and buffers from :attr:&＃96;state_dict&＃96; intothis module and its descendants. If :attr:&＃96;strict&＃96; is &＃96;&＃96;True&＃96;&＃96;, thenthe keys of :attr:&＃96;state_dict&＃96; must exactly match the keys returnedby this module&＃39;s :meth:&＃96;~torch.nn.Module.state_dict&＃96; function.Arguments:state_dict (dict): a dict containing parameters andpersistent buffers.strict (bool, optional): whether to strictly enforce that the keysin :attr:&＃96;state_dict&＃96; match the keys returned by this module&＃39;s:meth:&＃96;~torch.nn.Module.state_dict&＃96; function. Default: &＃96;&＃96;True&＃96;&＃96;"""missing_keys &＃61; []unexpected_keys &＃61; []error_msgs &＃61; []# copy state_dict so _load_from_state_dict can modify itmetadata &＃61; getattr(state_dict, &＃39;_metadata&＃39;, None)state_dict &＃61; state_dict.copy()if metadata is not None:state_dict._metadata &＃61; metadatadef load(module, prefix&＃61;&＃39;&＃39;):module._load_from_state_dict(state_dict, prefix, strict, missing_keys, unexpected_keys, error_msgs)for name, child in module._modules.items():if child is not None:load(child, prefix &＃43; name &＃43; &＃39;.&＃39;)load(self)

而里面的torch.load函数定义如下.map_location参数有三种形式:函数,字符串,字典

def load(f, map_location&＃61;None, pickle_module&＃61;pickle):"""Loads an object saved with :func:&＃96;torch.save&＃96; from a file.:meth:&＃96;torch.load&＃96; uses Python&＃39;s unpickling facilities but treats storages,which underlie tensors, specially. They are first deserialized on theCPU and are then moved to the device they were saved from. If this fails(e.g. because the run time system doesn&＃39;t have certain devices), an exceptionis raised. However, storages can be dynamically remapped to an alternativeset of devices using the &＃96;map_location&＃96; argument.If &＃96;map_location&＃96; is a callable, it will be called once for each serializedstorage with two arguments: storage and location. The storage argumentwill be the initial deserialization of the storage, residing on the CPU.Each serialized storage has a location tag associated with it whichidentifies the device it was saved from, and this tag is the secondargument passed to map_location. The builtin location tags are &＃96;&＃39;cpu&＃39;&＃96; forCPU tensors and &＃96;&＃39;cuda:device_id&＃39;&＃96; (e.g. &＃96;&＃39;cuda:2&＃39;&＃96;) for CUDA tensors.&＃96;map_location&＃96; should return either None or a storage. If &＃96;map_location&＃96; returnsa storage, it will be used as the final deserialized object, already moved tothe right device. Otherwise, :math:&＃96;torch.load&＃96; will fall back to the defaultbehavior, as if &＃96;map_location&＃96; wasn&＃39;t specified.If &＃96;map_location&＃96; is a string, it should be a device tag, where all tensorsshould be loaded.Otherwise, if &＃96;map_location&＃96; is a dict, it will be used to remap location tagsappearing in the file (keys), to ones that specify where to put thestorages (values).User extensions can register their own location tags and tagging anddeserialization methods using &＃96;register_package&＃96;.Args:f: a file-like object (has to implement read, readline, tell, and seek),or a string containing a file namemap_location: a function, string or a dict specifying how to remap storagelocationspickle_module: module used for unpickling metadata and objects (has tomatch the pickle_module used to serialize file)Example:>>> torch.load(&＃39;tensors.pt&＃39;)# Load all tensors onto the CPU>>> torch.load(&＃39;tensors.pt&＃39;, map_location&＃61;&＃39;cpu&＃39;)# Load all tensors onto the CPU, using a function>>> torch.load(&＃39;tensors.pt&＃39;, map_location&＃61;lambda storage, loc: storage)# Load all tensors onto GPU 1>>> torch.load(&＃39;tensors.pt&＃39;, map_location&＃61;lambda storage, loc: storage.cuda(1))# Map tensors from GPU 1 to GPU 0>>> torch.load(&＃39;tensors.pt&＃39;, map_location&＃61;{&＃39;cuda:1&＃39;:&＃39;cuda:0&＃39;})# Load tensor from io.BytesIO object>>> with open(&＃39;tensor.pt&＃39;) as f:buffer &＃61; io.BytesIO(f.read())>>> torch.load(buffer)"""


设置使用GPU,这里是

torch.cuda.set_device(device&＃61;0)  告诉编码器cuda使用gpu0号

net.cuda() 将模型放在gpu0号上面

关于writer &＃61; SummaryWriter(log_dir&＃61;log_dir)这个函数在后面会讲解

num_img_tr &＃61; len(trainloader)# 1764
num_img_ts &＃61; len(testloader)# 242 这是batch数目