The ONNX model was exported from the best validation-loss checkpoint from the selected recipe. The test metrics were not used to select the checkpoint and were logged in W&B run 8h2ywarm.

Model Details

Dataset: Fashion-MNIST
Architecture: shallow CNN
Channels: [64, 128, 256]
Dropout: 0.2
Optimizer: Adam
Learning rate: 0.001
Weight decay: 3e-5
Scheduler: cosine
Label smoothing: 0.02
Weight averaging: EMA
Training augmentation: random affine rotation/translation/scale
Source W&B run: qwfq2ihy
Source checkpoint: 044.ckpt
Final test W&B run: 8h2ywarm

Input / Output

Use model.onnx for code-independent inference.

Input name: images
Input shape: [batch, 1, 28, 28]
Input dtype: float32
Output name: logits
Output shape: [batch, 10]

Preprocessing:

Convert image to grayscale.
Resize to 28 x 28.
Scale pixel values to [0, 1].
Normalize with mean 0.2860 and standard deviation 0.3530.
Arrange the tensor as channels-first [batch, 1, 28, 28].

Usage

Install the runtime dependencies:

pip install huggingface_hub onnxruntime pillow numpy

Run inference with the ONNX model:

python

import numpy as np
import onnxruntime as ort
from huggingface_hub import hf_hub_download
from PIL import Image

LABELS = {
    0: "T-shirt/top",
    1: "Trouser",
    2: "Pullover",
    3: "Dress",
    4: "Coat",
    5: "Sandal",
    6: "Shirt",
    7: "Sneaker",
    8: "Bag",
    9: "Ankle boot",
}

model_path = hf_hub_download(
    repo_id="tsilva/fashion-mnist-classifier-cnn",
    filename="model.onnx",
)

image = Image.open("example.png").convert("L").resize((28, 28))
x = np.asarray(image, dtype=np.float32) / 255.0
x = (x - 0.2860) / 0.3530
x = x[None, None, :, :].astype(np.float32)

session = ort.InferenceSession(model_path, providers=["CPUExecutionProvider"])
logits = session.run(["logits"], {"images": x})[0]
prediction = int(logits.argmax(axis=1)[0])

print(prediction, LABELS[prediction])

Labels

Fashion-MNIST labels:

id	label
0	T-shirt/top
1	Trouser
2	Pullover
3	Dress
4	Coat
5	Sandal
6	Shirt
7	Sneaker
8	Bag
9	Ankle boot

Files

model.onnx: ONNX export of the validation-selected checkpoint, using opset 17. Prefer this file for portable inference.
model.ckpt: PyTorch Lightning checkpoint for the same model. This is code-dependent and mainly useful for PyTorch-based inspection or continued experimentation.
config.yaml: resolved training config.
metrics.csv: training/validation metric history.
modeling.py: minimal PyTorch model definition and checkpoint loader for users who specifically need the checkpoint path.
label_mapping.json: label id to class name mapping.

Limitations

Most remaining validation errors are visually plausible confusions between related Fashion-MNIST classes, especially shirt/top/pullover/coat and sneaker/ankle boot cases.

tsilva/fashionmnist-classifier-cnn

Fashion-MNIST CNN Classifier

Architecture

Results