Loading
Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
A deliberately trivial MNIST convnet -- the champion selected by an Optuna search and registered in MLflow, exported here as TorchScript and served by revision from this repo.
Model source
Source description
A deliberately trivial MNIST conv_net -- the champion selected
by an Optuna search and registered in MLflow, exported here as TorchScript and
served by revision from this repo.
Sources
1 sourceVerified Sep 28
Model artifacts
1 artifactSource excerpts
2 excerpts--- license: mit tags: - image-classification - mnist - mlflow --- # mnist-classifier (champion v2) A deliberately trivial MNIST `conv_net` -- the **champion** selected by an Optuna search and registered in MLflow, exported here as TorchScript and served *by revision* from this repo. - **Test accuracy:** 0.9866 - **Expected calibration error (ECE):** 0.0023 - **Hyperparameters:** `lr=0.0013292918943162175`, `conv_channels=16`, `dropout=0.07800932022121826`, `hidden_dim=128` - **Source MLflow run:** `4863fc86bb854bd4a8644dd3d49fc216` Preprocessing: grayscale, normalize with mean `0.1307`, std `0.3081` on a centred 28x28 digit. See the write-up for the full MLflow/Optuna story.
lr=0.0013292918943162175, conv_channels=16, dropout=0.07800932022121826, hidden_dim=1284863fc86bb854bd4a8644dd3d49fc216Preprocessing: grayscale, normalize with mean 0.1307, std 0.3081
on a centred 28x28 digit. See the write-up for the full MLflow/Optuna story.
Source context: 4 downloads · 0 likes · Pipeline image-classification · Repo rfflpllcn/mnist-classifier