Why does self.current_epoch in trainer.validate(model) is 0

[talhaanwarch]

This is my code

class OurModel(LightningModule):
    def __init__(self,train_split,val_split):
        super(OurModel,self).__init__()
        #architecute
        #lambda resnet
        
        self.train_split=train_split
        self.val_split=val_split
        #########TIMM#################
        model_name='resnest50d'
        self.model =  timm.create_model(model_name,pretrained=True)

        self.fc1=nn.Linear(1000,500)
        self.relu=nn.ReLU()
        self.fc2= nn.Linear(500,250)
        self.fc3= nn.Linear(250,1)
        self.drp=nn.Dropout(0.25)
        #parameters
        self.lr=1e-3
        self.batch_size=32
        self.numworker=2
        self.criterion=nn.BCEWithLogitsLoss()
        self.metrics=torchmetrics.Accuracy()

        self.trainloss,self.valloss=[],[]
        self.trainacc,self.valacc=[],[]
        
        self.sub_pred=0
    def forward(self,x):
        x= self.model(x)
        x=self.fc1(x)
        x=self.relu(x)
        x=self.drp(x)
        x=self.fc2(x)
        x=self.relu(x)
        x=self.drp(x)
        x=self.fc3(x)
        return x

    def configure_optimizers(self):
        opt=torch.optim.AdamW(params=self.parameters(),lr=self.lr )
        return opt
        
    def train_dataloader(self):
        return DataLoader(DataReader(self.train_split,True), batch_size = self.batch_size, 
                          num_workers=self.numworker,pin_memory=True,shuffle=True)

    def training_step(self,batch,batch_idx):
        image,label=batch
        pred = self(image)
        loss=self.criterion(pred.flatten(),label.float()) #calculate loss
        acc=self.metrics(pred.flatten(),label)#calculate accuracy
        return {'loss':loss,'acc':acc}

    def training_epoch_end(self, outputs):
        loss=torch.stack([x["loss"] for x in outputs]).mean().detach().cpu().numpy().round(2)
        acc=torch.stack([x["acc"] for x in outputs]).mean().detach().cpu().numpy().round(2)
        self.trainloss.append(loss)
        self.trainacc.append(acc)
        print('training acc',acc)
        self.log('train_loss', loss)
        
    def val_dataloader(self):
        ds=DataLoader(DataReader(self.val_split), batch_size = self.batch_size,
                      num_workers=self.numworker,pin_memory=True, shuffle=False)
        return ds

    def validation_step(self,batch,batch_idx):
        image,label=batch
        pred = self(image)
        loss=self.criterion(pred.flatten(),label.float()) #calculate loss
        acc=self.metrics(pred.flatten(),label)#calculate accuracy
        return {'loss':loss,'acc':acc}

    def validation_epoch_end(self, outputs):
        loss=torch.stack([x["loss"] for x in outputs]).mean().detach().cpu().numpy().round(2)
        acc=torch.stack([x["acc"] for x in outputs]).mean().detach().cpu().numpy().round(2)
        self.valloss.append(loss)
        self.valacc.append(acc)
        print('validation acc',self.current_epoch,acc)
        self.log('val_loss', loss)
        self.log('val_acc', acc)

this is how i train and validate it

trainer = Trainer(max_epochs=5, auto_lr_find=False, auto_scale_batch_size=False,
                        deterministic=True,
                        gpus=gpu,precision=gpup,
                        accumulate_grad_batches=4,
                        enable_progress_bar = True,
                        num_sanity_val_steps=0,
                        callbacks=[lr_monitor],
   
                        )
    trainer.fit(model)
    trainer.validate(model)

when i run trainer.validate(model) the value of self.current_epoch is 5. why it is not 0.

[rohitgr7]

hey @talhaanwarch !
because you have already run model.fit which has updated this property.