Model Isn't Converging

Me: Hey I wonder why my model isn't converg- PyTorch Best I can do is Epoch: 0, Accuracy: 50.27, Loss: nan Epoch: 10, Accuracy: 50.27, Loss: nan Epoch: 20, Accuracy: 50.27, Loss: nan Epoch: 30, Accuracy: 50.27, Loss: nan Epoch: 40, Accuracy: 50.27, Loss: nan

Rick Harrison and Chumlee from Pawn Stars stand behind a counter while a caption jokes about a machine-learning model that refuses to converge. The punchline imitates Rick’s bargaining phrase, “Best I can do is,” applying it to a broken training run stuck at 50.27% accuracy with a NaN loss. It is used as a reaction to failed optimization, debugging frustration, or mediocre results. EN: Rick Harrison and Chumlee from Pawn Stars stand behind a counter while a caption jokes about a machine-learning model that refuses to converge. The punchline imitates Rick’s bargaining phrase, “Best I can do is,” applying it to a broken training run stuck at 50.27% accuracy with a NaN loss. It is used as a reaction to failed optimization, debugging frustration, or mediocre results.

• @profunctor_io / 4668 2020-05-28