The prompts were summarized using QWEN to produce direct English formatted instead of fragmented, which worked better with some models like Kwen and worse with others like Chroma and zimage turbo.
This is the BigLiminal Klein edition.
2733246_training_data.zip
2733247_training_data.zip
K0DWG0TM79YAMK5VZ85TR27530.safetensors