MNIST CNN from scratch — SGD lr=0.001
Trained for 60s wall clock (batch 128, SGD momentum 0.9, seed 0), reaching step 10425 with final train loss 0.013465155847370625 and test accuracy 0.9852.
Part of a 10-learning-rate sweep. Live curves: https://huggingface.co/spaces/abidlabs/mnist-lr-sweep-trackio
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support