41a3e20440c61f4892d7731f89032b53932cce66,src/solver.py,Trainer,exec,#Trainer#,111

Before Change

                    self.write_log("wer",{"tr_att":cal_er(self.tokenizer,att_output,txt),
                                          "tr_ctc":cal_er(self.tokenizer,ctc_output,txt)})
                // Validation
                if self.step%self.valid_step == 0:
                    self.validate()

                // End of step
                self.step+=1
                if self.step > self.max_step:break

    

After Change

                total_loss = 0
                if ctc_output is not None:
                    if self.paras.ctc_backend =="cudnn":
                        ctc_loss = self.ctc_loss(ctc_output.permute(1,0,2).contiguous(), 
                                                 txt.to_sparse().values().to(device=self.device,dtype=torch.int32),
                                                 encode_len.to(device=self.device,dtype=torch.int32),
                                                 txt_len.to(device=self.device,dtype=torch.int32))
                    else:
                        ctc_loss = self.ctc_loss(ctc_output.permute(1,0,2).contiguous(), txt, encode_len, txt_len)
                    total_loss += ctc_loss*self.asr_model.ctc_weight
                if att_output is not None:

Instances