taohan10200/CRA5-model

CRA5: Extreme Compression of ERA5

This repository contains the pretrained weights for the CRA5 project.

Usage

You can use these weights with the CRA5 library.

from cra5.models.compressai.zoo import vaeformer_pretrained
# Note: You need to update the model URLs in your library to point here
# or manually load the state dict.

# Example if using updated library:
model = vaeformer_pretrained(quality=268, pretrained=True)

Models

• cra5_268v_300k.pth: Model with 268 variables.
• cra5_159v_150k.pth: Model with 159 variables.

See GitHub Repository for more details.