You can deploy this model and start earning money today!

kerasformers/
mixer_l16_224_goog_in21k_ft_in1k

Name: mixer_l16_224_goog_in21k_ft_in1k
Availability: InStock

timm/mixer_l16_224.goog_in21k_ft_in1k

finetune

apache-2.0

Views

No views yet

See our collection for all versions of MLP-Mixer.

Run MLP-Mixer with Keras 3: JAX, PyTorch, or TensorFlow

GitHub

zeromodels/mixer_l16_224_goog_in21k_ft_in1k

Paper: MLP-Mixer: An all-MLP Architecture for Vision (arXiv:2105.01601) · HF Papers

MLP-Mixer mixes tokens and channels with MLPs only (no attention). Classifier or per-block feature backbone.

For more details on the model, please go to the upstream model card.

Pure-Keras 3 conversion of timm/mixer_l16_224.goog_in21k_ft_in1k for zeromodels. One implementation runs unmodified on TensorFlow / Torch / JAX.

This is an image-classification / backbone checkpoint (MLPMixerImageClassify / MLPMixerModel).

✨ Quick start

python

import os

os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"

from PIL import Image
from zeromodels.models.mlp_mixer import MLPMixerImageClassify, MLPMixerModel, MLPMixerImageProcessor

model = MLPMixerImageClassify.from_weights("zeromodels/mixer_l16_224_goog_in21k_ft_in1k")
processor = MLPMixerImageProcessor.from_weights("zeromodels/mixer_l16_224_goog_in21k_ft_in1k")

image = Image.open("your_image.jpg").convert("RGB")
pixels = processor(image)  # resize + normalize (normalization lives in the processor)
logits = model(pixels, training=False)
print(logits.shape)  # (1, num_classes)

# Feature extraction: the backbone without the classifier head
backbone = MLPMixerModel.from_weights("zeromodels/mixer_l16_224_goog_in21k_ft_in1k", as_backbone=True)
features = backbone(pixels, training=False)

Load any MLP-Mixer variant the same way with from_weights("zeromodels/<variant>"):

Variant	Hub
`mixer_b16_224_goog_in21k`	`zeromodels/mixer_b16_224_goog_in21k`
`mixer_b16_224_goog_in21k_ft_in1k`	`zeromodels/mixer_b16_224_goog_in21k_ft_in1k`
`mixer_b16_224_miil_in21k_ft_in1k`	`zeromodels/mixer_b16_224_miil_in21k_ft_in1k`
`mixer_l16_224_goog_in21k`	`zeromodels/mixer_l16_224_goog_in21k`
`mixer_l16_224_goog_in21k_ft_in1k`	`zeromodels/mixer_l16_224_goog_in21k_ft_in1k`

Tips

Set KERAS_BACKEND before importing Keras / zeromodels.
MLPMixerImageClassify returns class logits; MLPMixerModel returns features (as_backbone=True for multi-scale stages).
See docs and Loading Weights.
Upstream / timm checkpoints: MLPMixerImageClassify.from_weights("hf:timm/mixer_l16_224.goog_in21k_ft_in1k").

Special Thanks

A huge thank you to the MLP-Mixer authors and the timm / Hub communities for creating and releasing these models.

License: see YAML license (usually matches the upstream checkpoint).

kerasformers/mixer_l16_224_goog_in21k_ft_in1k