Datasetpublic

tigerking009/zoya-image-1-experiments

ZOYA IMAGE-1 — Reproducible GGUF Experiments Purpose This dataset stores reproducible ZOYA IMAGE-1 image-generation experiments together with the exact generation parameters, model identities, SHA256 fingerprints, and validation reports. The package is designed for controlled comparisons where the tested variable is changed explicitly and all other relevant variables remain fixed. Current baseline Experiment ID: ZOYA_PHASE0_BASELINE_00001… See the full description on the dataset page: https://huggingface.co/datasets/tigerking009/zoya-image-1-experiments.

sourceHugging Faceotherupdated 2mo agoView on Hugging Face

0likes164downloads

Dataset Card

ZOYA IMAGE-1 — Reproducible GGUF Experiments

Purpose

This dataset stores reproducible ZOYA IMAGE-1 image-generation experiments together with the exact generation parameters, model identities, SHA256 fingerprints, and validation reports.

The package is designed for controlled comparisons where the tested variable is changed explicitly and all other relevant variables remain fixed.

Current baseline

Experiment ID: ZOYA_PHASE0_BASELINE_00001

Exact model components

Component	File
UNet / image generator	`z-image-turbo-iq4_nl.gguf`
Text encoder	`qwen3_4b_f32-iq4_nl.gguf`
VAE	`pig_flux_vae_fp32-f16.gguf`

Exact generation configuration

Parameter	Value
Seed	`169887354867181`
Width	`1024`
Height	`1024`
Orientation	`square`
Steps	`8`
CFG	`1.0`
Sampler	`euler`
Scheduler	`simple`
Denoise	`1.0`
Text encoder type	`lumina2`

Prompt

cute anime style girl with massive fluffy fennec ears and a big fluffy tail blonde messy long hair blue eyes wearing a maid outfit with a long black gold leaf pattern dress and a white apron, it is a postcard held by a hand in front of a beautiful realistic city at sunset and there is cursive writing that says "ZImage, Now in GGUF"

Negative prompt

(empty)

Baseline image

images/ZOYA_PHASE0_BASELINE_00001_.png

Image SHA256:

983e86579609ad66604b2a153d73302e2164881f46eb3673985cebd891e9a5eb

The baseline output was independently validated for:

—PNG signature
—image decoding
—pixel loading
—dimensions
—model identity
—prompt/settings consistency

Fair comparison policy

For a controlled comparison:

1.Prompt must remain identical.
2.Negative prompt must remain identical.
3.Seed must remain identical.
4.Resolution must remain identical.
5.Model identity must remain explicitly recorded.
6.Model SHA256 values must remain explicitly recorded.
7.Generation settings must remain identical unless that setting is the variable under test.

Only the explicitly tested variable should change.

Resolution / orientation policy

The planned collection should be majority landscape-oriented.

Square and portrait images are permitted for controlled diversity, but every record must explicitly store:

—width
—height
—orientation
—aspect ratio

Resolution is therefore never implicit.

Model integrity

Model SHA256 values are stored in:

metadata/experiment_manifest.json

and

metadata/experiment_manifest.csv

Reproducibility artifacts

The repository contains:

—images/ — generated image artifacts
—image_metadata.csv — canonical tabular records
—image_metadata.jsonl — line-oriented machine-readable records
—metadata/experiment_manifest.json — complete experiment manifest
—metadata/experiment_manifest.csv — tabular experiment manifest
—metadata/baseline_visual_fingerprints.json — baseline visual fingerprint
—metadata/dataset_card_notes.txt — experiment notes
—metadata/cell48_first_baseline_generation.json — generation report
—metadata/cell49_baseline_output_validation.json — integrity report
—dataset_schema.json — dataset field contract

Status

The current repository contains the validated Phase-0 baseline record.

Benchmarking has not been claimed from the baseline alone.

Additional experiments should preserve the same provenance and fair-comparison rules.