Whyx-PROmpTea / requirements.txt
ArtShumov's picture
feat(tagger): pose extraction + Qwen3-VL optional NL captioner + WD14x3 ensemble weights recalibration
2c574fa
Raw
History Blame
197 Bytes
audioop-lts; python_version>='3.13'
huggingface_hub>=0.25.0,<1.0
pandas>=2.0.0
pillow>=10.0.0
numpy>=1.26.4
torch>=2.3,<3
torchvision>=0.18,<1
timm>=1.0.12,<2
transformers>=4.45
onnxruntime>=1.16