Training loss reduce to zero in first epoch

I am a beginner in deep learning and in coding also, kindly help in solving the issue that I am facing. I am training the SketchANet model using the shoe2V dataset, including the sketch and photo images. I am preparing the triplet dataset that is anchor, positive image and negative image. while training the model I am using the nn.TripletMarginLoss(margin=0.3) loss model. while I am training the model the training loss starts from 0.3 and reduces to 0 in the first epoch itself. Kindly someone helps in solve this issue… Thanks in advance