다음은 모델을 학습시킨 결과입니다.
Train Epoch: 1 [0/20000 (0%)] Loss: 0.695734 Train Epoch: 1 [5000/20000 (25%)] Loss: 0.697002 Train Epoch: 1 [10000/20000 (50%)] Loss: 0.700943 Train Epoch: 1 [15000/20000 (75%)] Loss: 0.691708 [EPOCH: 1], Validation Loss: 0.69 | Validation Accuracy: 0.51 Train Epoch: 2 [0/20000 (0%)] Loss: 0.692522 Train Epoch: 2 [5000/20000 (25%)] Loss: 0.689648 Train Epoch: 2 [10000/20000 (50%)] Loss: 0.693150 Train Epoch: 2 [15000/20000 (75%)] Loss: 0.691025 [EPOCH: 2], Validation Loss: 0.69 | Validation Accuracy: 0.50 Train Epoch: 3 [0/20000 (0%)] Loss: 0.694444 Train Epoch: 3 [5000/20000 (25%)] Loss: 0.693492 Train Epoch: 3 [10000/20000 (50%)] Loss: 0.693789 Train Epoch: 3 [15000/20000 (75%)] Loss: 0.693217 [EPOCH: 3], Validation Loss: 0.69 | Validation Accuracy: 0.50 Train Epoch: 4 [0/20000 (0%)] Loss: 0.694434 Train Epoch: 4 [5000/20000 (25%)] Loss: 0.697406 Train Epoch: 4 [10000/20000 (50%)] Loss: 0.691963 Train Epoch: 4 [15000/20000 (75%)] Loss: 0.693150 [EPOCH: 4], Validation Loss: 0.69 | Validation Accuracy: 0.50 Train Epoch: 5 [0/20000 (0%)] Loss: 0.696038 Train Epoch: 5 [5000/20000 (25%)] Loss: 0.691857 Train Epoch: 5 [10000/20000 (50%)] Loss: 0.696768 Train Epoch: 5 [15000/20000 (75%)] Loss: 0.695390 [EPOCH: 5], Validation Loss: 0.69 | Validation Accuracy: 0.50
검증 데이터셋을 모델에 적용한 결과 50%의 정확도를 보이고 있습니다. 여전히 높은 예측력이라고 볼 수 없습니다. 또한, RNN 셀과 비교해도 성능의 차이는 없습니다. 더 많은 에포크를 지정한다면 성능은 좋아질 수 있습니다.