datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
pixart-sigma-toy-datasetpixart-alpha-critic5-2-data-0-5m-20260325
Critic5.2 Training Data Bundles
This dataset repo stores the training data path used by train-1.sh.
Included data groups:
Critic5.2-root-metadata.tar.zst
Critic5.2-images.tar.part-*
Critic5.2-caption_feature_wmask.tar.part-*
Critic5.2-img_vae_features-512resolution-noflip.tar.part-*
Restore archives
Images:
cat Critic5.2-images.tar.part-* > Critic5.2-images.tar
tar -xf Critic5.2-images.tar
Caption features:
cat Critic5.2-caption_feature_wmask.tar.part-* >… See the full description on the dataset page: https://huggingface.co/datasets/yilingwang/pixart-alpha-critic5-2-data-0-5m-20260325.PixArt-Eval-30KPixArt-Eval30KSAM-LLaVA-Captions10Mdata_toyImages are from JourneyDB
pixart_sigma_subsetpixart_sigma_baseline_coco_1024housey-home-pixart-alpha
Housey Home v2: PixArt Alpha Split
I was in the process of producing a fully synthetic dataset for ungrounded image generation using an unconventional combination of layers. As such, I needed a dataset of highly similar objects with 'themes'. In order to produce log(x, y) combinations of options in the final model. This is that dataset.
The initial ( 07/15/2024 ) release includes ~1.3k unique houses, each processed using a VQA, PixArt-alpha/PixArt-Sigma-XL-2-1024-MS, to be precise.… See the full description on the dataset page: https://huggingface.co/datasets/MrOvkill/housey-home-pixart-alpha.pixart-alpha-512-ms-critic-0.5m-bundles-20260325
PixArt Alpha Train Bundles
This dataset repo contains:
train-1.sh
archives/PixArt-alpha-code.tar.gz
archives/512-MS-Critic-0.5M-metadata.tar.zst
archives/512-MS-Critic-0.5M-checkpoint-parts/epoch_0_step_0.pth.zst.part-*
Restore checkpoint
cat epoch_0_step_0.pth.zst.part-* > epoch_0_step_0.pth.zst
zstd -d epoch_0_step_0.pth.zst -o epoch_0_step_0.pth
PixArt-Spatial-allseeds-rectifiedPixArt-Count-seedmining-v1.01PixArt-Spatial-allseedsPixArt-Spatial-allseeds-replacedPixArt-Count-allseeds-rectifiedPixArt-Spatial-seedmining-v1.01-rectifiedPixArt-Spatial-seedmining-v1.01PixArt-Spatial-seedmining-v1.01-replacedPixArt-Count-seedmining-v1.01-rectifiedpixart-0.5m-step50pixart-0.5mPixArt-Count-allseedshousey-home-pixart-alpha
Housey Home v2: PixArt Alpha Split
I was in the process of producing a fully synthetic dataset for ungrounded image generation using an unconventional combination of layers. As such, I needed a dataset of highly similar objects with 'themes'. In order to produce log(x, y) combinations of options in the final model. This is that dataset.
The initial ( 07/15/2024 ) release includes ~1.3k unique houses, each processed using a VQA, PixArt-alpha/PixArt-Sigma-XL-2-1024-MS, to be precise.… See the full description on the dataset page: https://huggingface.co/datasets/aimbea/housey-home-pixart-alpha.Data-collectionpixar_training_sl_01