google / google/flax

ResNet-18 Initialization

Open
#4,388 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
7.3k
Forks
833
Avg merge
5h 11m
Merged PRs (30d)
5

Description

Hi,

I am new to `Flax` and am trying to learn it. Regarding the [ResNet example](https://github.com/google/flax/tree/main/examples/imagenet), it seems that the initial parameters, when not using pretraining, do not match those provided in `torchvision.models`.

```python
from torchvision.models import resnet18
nnet = resnet18(pretrained=False)
```

I was wondering if using `flax.linen.initializers.he_normal` could help address this.

Since there are no official ResNet models in Flax, do you have any suggestions on how to replicate the ResNet-18 model from `torchvision.models` in JAX?

Thank you!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.