Z-Image LoRA Checkpoints & Evaluation Results

Repository: LongDuy/zimage-checkpoint-2

This repository contains trained checkpoints and evaluation metrics for Z-Image-Turbo across 3 experimental runs.

📊 Summary of Evaluation Results

Run	Resolution	Text-Guided Loss	No-Text Loss	Weighted Loss	Trained Folder	Evaluation Folder
Run 1	`[512, 768, 1024]`	`0.049565`	`0.060313`	`0.057626`	`trained_z-image_20260811T035640Z-14477`	`evaluation_z-image_20260811T044607Z-18647`
Run 2	`[1024]`	`0.052288`	`0.062500`	`0.059947`	`trained_z-image_20260811T052622Z-23440`	`evaluation_z-image_20260811T062211Z-27854`
Run 3	`[768, 1024, 1280]`	`0.053941`	`0.060982`	`0.059222`	`trained_z-image_20260811T062739Z-28154`	`evaluation_z-image_20260811T072111Z-31822`

Loss Formula

$\text{Weighted Loss} = 0.25 \times \text{text\_guided\_average} + 0.75 \times \text{no\_text\_average}$

📁 Repository Structure

.
├── run_1_res_512_768_1024_20260811T035640Z/
│   ├── config.yaml
│   ├── last_000000500.safetensors
│   ├── last_000000750.safetensors
│   ├── last_000001000.safetensors
│   ├── last_000001250.safetensors
│   ├── last.safetensors
│   ├── zimage-best.safetensors
│   ├── zimage_checkpoint_evaluations.json
│   ├── optimizer.pt
│   └── evaluation_results.json
│
├── run_2_res_1024_20260811T052622Z/
│   ├── config.yaml
│   ├── last_000000500.safetensors
│   ├── last_000000750.safetensors
│   ├── last_000001000.safetensors
│   ├── last_000001250.safetensors
│   ├── last.safetensors
│   ├── zimage-best.safetensors
│   ├── zimage_checkpoint_evaluations.json
│   ├── optimizer.pt
│   └── evaluation_results.json
│
└── run_3_res_768_1024_1280_20260811T062739Z/
    ├── config.yaml
    ├── last_000000500.safetensors
    ├── last_000000750.safetensors
    ├── last_000001000.safetensors
    ├── last_000001250.safetensors
    ├── last.safetensors
    ├── zimage-best.safetensors
    ├── zimage_checkpoint_evaluations.json
    ├── optimizer.pt
    └── evaluation_results.json

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support