Boho Festival LoRA

Overview

This training job focused on creating a LoRA extension to capture the visual and stylistic characteristics of a boho festival concept using the Z-Image De-Turbo diffusion architecture.

Model & Architecture

• Base Model: ostris/Z-Image-De-Turbo

• Architecture: zimage:deturbo

• Quantization: qfloat8 (model and text encoder)

• Precision: bf16

Training Configuration

• Training Steps: 3000

• Optimizer: AdamW 8-bit

The training emphasized a balanced content-to-style objective, suitable for flexible prompt usage.

Trigger Words

• Activation Tokens: boho; festival

Summary

This training successfully produced a lightweight LoRA extension capable of injecting bohemian festival aesthetics into Z-Image Turbo generations. The configuration balances quality, flexibility, and VRAM efficiency, making the resulting model suitable for creative image generation workflows and further fine-tuning if required.