Hugging Face's logo

Datasets:
QLNI
/
FLUX-Reason-6M-flux2-latents

Tasks:
Text-to-Image
Size:
1M<n<10M
License:
Dataset card Files
xet
 Community
Dataset Viewer
The dataset viewer is not available for this dataset.
Cannot get the config names for the dataset.
Error code:   ConfigNamesError
Exception:    ValueError
Message:      Split name should match '^\w+(\.\w+)*$'' but got '{'fluxdb-aesthetics-part01', 'fluxdb-text', 'fluxdb-imaginative', 'fluxdb-aesthetics-part02'}'.
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/dataset/config_names.py", line 67, in compute_config_names_response
                  config_names = get_dataset_config_names(
                      path=dataset,
                      token=hf_token,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 161, in get_dataset_config_names
                  dataset_module = dataset_module_factory(
                      path,
                  ...<4 lines>...
                      **download_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1182, in dataset_module_factory
                  raise e1 from None
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1157, in dataset_module_factory
                  ).get_module()
                    ~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 640, in get_module
                  patterns = get_data_patterns(base_path, download_config=self.download_config)
                File "/usr/local/lib/python3.14/site-packages/datasets/data_files.py", line 495, in get_data_patterns
                  return _get_data_files_patterns(resolver)
                File "/usr/local/lib/python3.14/site-packages/datasets/data_files.py", line 281, in _get_data_files_patterns
                  raise ValueError(f"Split name should match '{_split_re}'' but got '{splits}'.")
              ValueError: Split name should match '^\w+(\.\w+)*$'' but got '{'fluxdb-aesthetics-part01', 'fluxdb-text', 'fluxdb-imaginative', 'fluxdb-aesthetics-part02'}'.

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.

FLUX-Reason-6M as FLUX.2 latents (256 px)

Ready-to-train latents of the 5.6M images of LucasFang/FLUX-Reason-6M (Apache-2.0), encoded with the FLUX.2 VAE encoder from black-forest-labs/FLUX.2-small-decoder (Apache-2.0). No images are stored; decoding a latent with the FLUX.2 decoder gives the picture back (about 36 dB).

One .npz per source parquet shard (same file name), all four subsets (Aesthetics-Part01, Aesthetics-Part02, Imaginative, Text):

  • lat (N, 128, 16, 16) float16: image center-cropped and resized to 256x256, FLUX.2 encoder mean -> (32, 32, 32), patchified 2x2 to (128, 16, 16), then normalised per channel with the VAE's BatchNorm running stats: (z - running_mean) / sqrt(running_var + 1e-4). To decode: undo the normalisation, un-patchify to (32, 32, 32), vae.decode.
  • caps (N,) object: a list of captions per image: every available one of composition / entity / text / style / imaginative, plus caption_detail cut to 100 words.
  • clip (N,) float16: CLIP ViT-B/16 cosine between the image and its first caption (prompt-match score, for filtering).
  • dino (N, 64, 768) float16, first 60 shards only: DINOv2-base patch features pooled to 8x8 (for REPA-style training).

Made by QLNI for the SHADOW image model. Prep code: prep_flux2.py in the SHADOW repositories.
Downloads last month
5,051
Total file size:
451 GB

Models trained or fine-tuned on QLNI/FLUX-Reason-6M-flux2-latents

Space using QLNI/FLUX-Reason-6M-flux2-latents 1