---
license: bsd-3-clause
library_name: libreyolo
pipeline_tag: image-classification
datasets:
  - imagenet-1k
tags:
  - image-classification
  - alexnet
  - torchvision
  - imagenet
  - libreyolo
---

# LibreAlexNetb-cls

AlexNet image classification weights repackaged for LibreYOLO. This is
torchvision's single-tower, 64-channel-stem variant: no local response
normalization and no grouped convolutions. It is not the two-GPU 2012 graph.

```python
from libreyolo import LibreYOLO

model = LibreYOLO("LibreAlexNetb-cls.pt")
result = model.predict("image.jpg")
print(result.probs.top1, result.probs.top5)
```

## Source

Derived from [pytorch/vision](https://github.com/pytorch/vision) at commit
[`336d36e8db990a905498c73933e35231876e28bc`](https://github.com/pytorch/vision/commit/336d36e8db990a905498c73933e35231876e28bc).
Copyright (c) Soumith Chintala 2016 and torchvision contributors. The source
implementation is BSD-3-Clause.

Official checkpoint: [alexnet-owt-7be5be79.pth](https://download.pytorch.org/models/alexnet-owt-7be5be79.pth)  
Official checkpoint SHA-256: `7be5be791159472b1fbf3c69796f7cb30dca7ad8466c2df70058c37116cdee02`  
Official checkpoint bytes: `244408911`  
Published ImageNet-1K accuracy: 56.522% top-1, 79.066% top-5.

## Modifications

LibreYOLO metadata and the canonical filename were added. Learned tensors and
state-dict keys are unchanged. The native graph strict-loads the official state
dict and produces bit-identical logits (`max_abs_diff == 0.0`).

Converted checkpoint SHA-256: `95f6996b7b4c5526e7e47ad99cf78b2a3643baa3ba1d4107ab840a05e73d1f5e`  
Converted checkpoint bytes: `244431825`.

The converter and parity tests are in the
[LibreYOLO source repository](https://github.com/LibreYOLO/libreyolo).

## License

The checkpoint publisher did not attach a separate per-object license file.
This mirror applies the releasing project's BSD-3-Clause license on an
**implied**, not publisher-confirmed, basis. Torchvision warns that pretrained
models may have their own licenses or terms derived from training data and
that users must determine whether they have permission for their use case.
The weights were trained on ImageNet-1K; ImageNet's dataset and image-source
terms remain the downstream user's responsibility. See [`LICENSE`](./LICENSE)
and [`NOTICE`](./NOTICE).