metadata

license: other
license_name: fair-ai-public-license-1.0-sd
license_link: https://freedevproject.org/faipl-1.0-sd/
language:
  - en
base_model:
  - Laxhar/noobai-XL-1.0
pipeline_tag: text-to-image
library_name: diffusers
tags:
  - safetensors
  - diffusers
  - stable-diffusion
  - stable-diffusion-xl
  - art

V-Prediction Loss Weighting Test

Notice

This repository contains personal experimental records. No guarantees are made regarding accuracy or reproducibility.

Overview

This repository is a test project comparing different loss weighting schemes for Stable Diffusion v-prediction training, examining the effects of static weighting curves versus adaptive neural network-based weighting.

Environment

sd-scripts dev branch
- Commit hash: [6adb69b] + Modified

Test Cases

This repository includes test models using four different weighting schemes:

test_normal_weight
- Baseline model using standard weighting
test_edm2_weighting
- New loss weighting scheme
test_min_snr_1(incomplete)
- Baseline model with --min_snr_gamma = 1
test_debias_scale-like(incomplete)
- Baseline model with additional parameters:
  - --debiased_estimation_loss
  - --scale_v_pred_loss_like_noise_pred

Training Parameters

For detailed parameters, please refer to the .toml files in each model directory. Each model directory uses sdxl_train.py (and sdxl_train.py and t.py for edm2).

Common parameters:

Samples: 57,373
Epochs: 3
U-Net only
Learning rate: 3.5e-6
Batch size: 8
Gradient accumulation steps: 4
Optimizer: Adafactor (stochastic rounding)
Training time: 13.5 GPU hours (RTX4090) per trial