vit-base-timm-int8-litert

ViT-Base/16 224 (timm augreg) optimized for Arm-based mobile CPUs with SME2

ViT-Base/16 224 (timm augreg_in21k_ft_in1k) optimized with INT8 dynamic post-training quantization for image classification, exported to LiteRT .tflite format for efficient inference on Arm-based mobile CPUs with SME2.

Summary

This repository contains an Arm-optimized version of timm/vit_base_patch16_224.augreg_in21k_ft_in1k for image classification, quantized to INT8 via dynamic post-training quantization — per-channel symmetric INT8 weights, with activations quantized to INT8 dynamically at inference. The model is provided in LiteRT .tflite format, targeting Mobile CPU systems.

ViT-Base/16 is a Vision Transformer that splits a 224x224 RGB image into 16x16 patches. This checkpoint was pre-trained on ImageNet-21k with the AugReg recipe and fine-tuned on ImageNet-1k, and produces logits for 1,000 ImageNet-1k classes.

This version is intended to demonstrate efficient inference on Arm-based platforms while preserving the original model's intended behavior. Arm has evaluated this model on ImageNet-1k and measured performance on a representative evaluation target.

Key results

Original model

Model files

Performance

Performance was measured on the vivo X300 reference configuration below. The Graviton environment documented for example.py verifies portability of the inference flow but does not provide or imply equivalent Graviton performance.

Reference configuration

Performance results

Accuracy

Accuracy was evaluated using the same preprocessing, input resolution, and evaluation protocol described below. Where possible, the optimized model is compared against the original model under the same evaluation conditions.

Evaluation setup

Accuracy results

Accuracy was measured using the evaluation setup described above. Users should re-evaluate the model on their own data before production use.

Arm optimization approach

Arm optimized this model for efficient inference on Arm-based platforms using a hardware-aware conversion and validation flow.

For this release, Arm used:

The goal of this process is to improve deployment characteristics such as latency, memory use, model size, and runtime compatibility while preserving the model's intended behavior. Detailed conversion scripts, calibration configuration, or backend-specific implementation details may be provided separately where appropriate.

Using this model

Python environment

Note: The Python/uv example runs on AWS Graviton (Ubuntu arm64) to confirm runtime compatibility only, and is intended as a guideline for building an equivalent run script on Mobile CPU systems.

Install dependencies

Dependencies are declared in pyproject.toml, which ships with this repository. Resolve and install them into a local virtual environment with uv:

uv python install
uv sync --frozen

Run the example

uv run example.py

The example loads the optimized .tflite weight and sample_input.jpg from its own directory, prints the top-5 predictions, and writes predictions.json beside itself. Edit the MODEL_PATH and IMAGE_PATH constants at the top of the script to point it at other files.

Expected input

Expected output

Intended use

This model is intended for developers evaluating image classification workloads on Arm-based platforms. It is suitable as a reference implementation for benchmarking, prototyping, and integration exploration.

Limitations

• Performance depends on the target device, runtime version, backend/delegate support, memory configuration, and system load.
• Accuracy was evaluated on ImageNet-1k val and may not generalize to all domains.
• This release preserves the original model's intended task and behavior, but users should validate it for their own application, data, and deployment environment.
• This repository is not a replacement for the original model documentation.

Additional notes

The export is fixed at 224x224 input resolution, and it targets the CPU backend.

Dynamic quantization keeps INT8 weights on disk and in memory and dequantizes them per inference. No calibration dataset is required, and no layers were skipped from quantization: only operators without an INT8 dynamic kernel on the LiteRT XNNPACK path remain in FP32.

The export captures the model graph with torch.export, applies PT2E dynamic INT8 quantization with per-channel weights, and converts the result to .tflite. The attention blocks are exported with an eager, batched-matmul attention implementation so that the graph is capturable and quantizable.

• Sample input: sample_input.jpg is derived from Samoyed on Beach by Appleinfl, via Wikimedia Commons (public domain).

About this version

Original Model: timm/vit_base_patch16_224.augreg_in21k_ft_in1k by Google Research (AugReg ViT weights), ported to PyTorch by Ross Wightman in timm - Repository

Optimization/conversion: Arm-Optimized version for execution on Arm-based platforms.

Converted/optimized by: Arm

License: The Original Model and the Optimized Model are subject to Apache-2.0.

This repository contains a converted or optimized version of the Original Model (the “Optimized Model”). The Original Model has been converted or optimized as described above for execution on Arm-based platforms.

No retraining or fine-tuning of the Original Model was performed as part of the conversion or optimization. The conversion or optimization was not intended to change the Original Model’s behavior or intended use.

Original Model and Documentation

For information about the Original Model, including its development, training data, intended uses, limitations and other relevant information, please refer to the Original Model repository. Information in that repository was provided by the original developer or other third parties and, unless expressly stated otherwise, has not been independently verified by Arm.

Licenses and Third-Party Terms

Use of the Original Model and the Optimized Model is subject to the applicable licenses, usage restrictions and other terms identified above and in the relevant repositories. Publication of the Optimized Model does not grant any rights beyond those provided under the applicable license terms.

You are responsible for reviewing those terms and ensuring that your use of the Original Model and the Optimized Model is permitted.

Purpose of this Release

The Optimized Model is provided as a reference implementation to demonstrate and evaluate execution and performance on Arm-based systems. It is not a production-ready or supported solution.

Arm’s publication of the Optimized Model does not constitute an endorsement or certification of the Original Model or a representation that the Optimized Model is suitable for production use or any particular purpose.

To the fullest extent permitted by applicable law (i) the Optimized Model is provided “as is.” Arm makes no representations or warranties that the Original Model, the Optimized Model or their outputs are accurate, safe, secure, non-infringing, legally compliant, suitable for production use or fit for any particular purpose; and (ii) Arm will not be liable for any loss or damage arising from or in connection with the Optimized Model, its use or its outputs.

You are responsible for independently evaluating the Optimized Model, its outputs and its suitability for your intended use, including compliance with applicable legal, regulatory, safety and security requirements.

Arm does not commit to provide ongoing support, maintenance or updates for the Optimized Model. Any use of or reliance on the Optimized Model or its outputs is at your own risk.