christian-tarantino/cc12m-500k-encoded
CC12M 500k Pre-encoded Latents Curated subset of the CC12M dataset (~500,000 text-image pairs) pre-encoded into high-performance HDF5 (.h5) format for training Multimodal Diffusion Transformers (MMDiT). Dataset Specifications Total Samples: ~500k pairs (Split across multi-bucket resolutions). Format: HDF5 (.h5) containing pre-extracted latent representations and tokenized text conditioning. Resolution Buckets: Optimized for 512x512, 768x512 and 512x768 aspect… See the full description on the dataset page: https://huggingface.co/datasets/christian-tarantino/cc12m-500k-encoded.
| Author / organization | christian-tarantino |
|---|---|
| License (as declared) | mit |
| Downloads | 0 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 22 October 2026 |
| Last updated at source | 22 October 2026 |
Recent changes
- 22 Oct 2026 Description: none → CC12M 500k Pre-encoded Latents Curated subset of the CC12M dataset (~500,000 text-image pairs) pre-encoded into high-performance HDF5 (.h5) format for training Multimodal Diffusion Transformers (MMDiT). Dataset Specifications Total Samples:
- 22 Oct 2026 Topics added dataset, hdf5, cc12m, latents, miniflux, rectified-flow
Changes detected by Silytics between refreshes of the source.
More from christian-tarantino
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 22 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.