pyiqa v.0.1.12

❗Important Change: Remove set_random_seed in test process.

In this version, set_random_seed has been removed due to its potential negative impact on the training process. For metrics that were previously tested on multiple crops, we have adopted a uniform cropping approach to ensure consistent results. This method aims to cover the entire image as much as possible, providing a fairer evaluation.
Please note that this change may lead to slight differences in the following metrics. Below is a comparison of results on sample images:

Metrics	I03.bmp	I04.bmp	I06.bmp	I08.bmp	I19.bmp	average
ahiq	-0.1059	0.5855	0.7066	0.5176	0.2566	0.3921
v0.1.12	-0.0996	0.5952	0.7331	0.5710	0.2196	0.4039
hyperiqa	0.1572	0.7397	0.7056	0.7204	0.2898	0.5226
v0.1.12	0.1692	0.7270	0.6972	0.7130	0.3105	0.5234
maniqa	0.2518	0.6176	0.4449	0.5197	0.2309	0.430
v0.1.12	0.2320	0.6033	0.4573	0.5201	0.2670	0.415
tres	27.86	85.95	87.19	95.53	80.02	75.31
v0.1.12	25.67	85.64	85.31	96.49	77.49	74.12

📢 Fix Bugs

1. Fix bug in inference of liqe. 0508684

✨ New features

1. Improve codes to allow training of lpips, dists, pieapp 036f2cc
2. Add distributed training. 18f123e

🛠️ Improvements

1. Update demo codes. c8ae6a7
2. Add alert to make README more readable. bee2c14

🍻 New Contributors

• Add ARNIQA by @LorenzoAgnolucci at #158
• Update requirements @shakedel by at #164

Many thanks to the valuable contributions 🤗 !

Full Changelog: v0.1.11...v0.1.12