The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/folder_based_builder/folder_based_builder.py", line 251, in _split_generators
raise ValueError(
"`file_name`, `*_file_name`, `file_names` or `*_file_names` must be present as dictionary key in metadata files"
)
ValueError: `file_name`, `*_file_name`, `file_names` or `*_file_names` must be present as dictionary key in metadata files
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Texture-rich 1024px benchmark showcase
The original seven-model collection contains 60 hand-curated prompts and 420 generated images: 20 prompts each from Qwen-Image-Bench EN, LongText-EN, and CVTG, rendered by seven models at 1024×1024 with seed 0 protocol alignment.
We first generated 40 candidates per benchmark (720 images), reviewed the actual seven-model comparisons, and downselected prompts for visual detail, material/texture richness, cross-model presentability, and scene diversity.
Models and generation settings
| Model | NFE | CFG | Fixed shift | Prompt input |
|---|---|---|---|---|
| Z-Image Base | 32 | 4 | 6 | original benchmark prompt |
| Z-Image Turbo | 8 | 0 | 3 | original benchmark prompt |
| ZImage AnyFlow2 4-Step-1000 | 4 | 0 | 6 | original benchmark prompt |
| AnyFlow2-Z-Image (Dense Stage2) | 4 | 0 | 6 | original benchmark prompt |
| ZImage-4Step-3000 | 4 | 0 | 6 | original benchmark prompt |
| FLUX.3 Image Klein | 16 | 0 | 16 / alpha | official PUP prompt |
| FLUX.3 4-Step | 4 | 1 | 6.93 | official PUP prompt |
All generation used 1024×1024 output. The two FLUX.3 columns use the official prompt-upsampling captions; both the source benchmark prompt and the exact PUP caption are retained in metadata.jsonl.
Layout
images/<benchmark>/<prompt-key>/<model>.png: full-resolution images.comparisons/<benchmark>/<prompt-key>.jpg: seven-column visual comparisons.metadata.jsonl: prompts, PUP prompts, seeds, and protocols.selection_ranks.json: auditable downselection from each 40-prompt candidate pool.
FLUX.3 Raw2Q-2000 addition
The same 60 prompts, seeds and official PUP captions are rendered by kimi000/Flux3-Klein-4Step-Raw2Q-2000. Native settings: 4 steps / 4 NFE, guidance 1, shift 6.93, final interval scale 0.3; CPU token-major noise, BF16 transformer/text encoder, FP32 integration and VAE. The checkpoint card defaults to raw prompts; this addition intentionally uses the published official PUP captions.
The preceding eight-model collection contains 480 generated model images. Existing full-resolution images, metadata and seven-column comparisons are unchanged. The new comparisons append a Raw2Q column to the already published comparison previews; they are not scoring inputs.
FLUX.1-dev Raw2Q-1000 addition
The same 60 selected prompts and seeds are rendered with kimi000/Flux1-Dev-4Step-Raw2Q-1000. Native settings: 1024px, four steps / four NFE, distilled guidance embedding 3.5, fixed shift exp(1.15), final interval scale 0.3. This FLUX.1 column uses the original prompt, CLIP 77 / T5 512 tokens and no PUP, as specified by its author. The FLUX.3 columns retain their original official PUP captions. This preprocessing difference is intentional and recorded.
Existing images and metadata are unchanged. Comparisons append the new column to the previously published eight-model JPEG previews; full-resolution PNGs remain available.
- Downloads last month
- 134