text_amon_API / requirements.txt
Abdullahcoder54's picture
DreamX-Creator 1.0 on ZeroGPU: vendored videox_fun + dreamx_inference from AMAP-ML upstream; generate(image, prompt)->(mp4, last-frame PNG, seed), neutral keyframe when image empty, DREAMX_CKPT_DIR for persistent checkpoints, diffusers 0.37.1 stack
982899c
Raw History Blame Contribute Delete
619 Bytes
# For ZeroGPU, gradio/spaces/huggingface_hub/torch are platform-managed; the
# Dockerfile installs torch explicitly from the cu124 index and then this file.
# DreamX-Creator (Wan2.2-TI2V-5B + gated A2V/V2A) engine pins mirror the upstream
# Space (hugging-apps/gd-ml-dreamx-creator) adapted to our known-good base.
diffusers==0.37.1
transformers==5.16.1
accelerate>=1.0
safetensors
sentencepiece
# .pth checkpoints are saved pre-torch-2.6; app.py wraps torch.load(weights_only=False).
# torch 2.6.0+cu124 predates numpy 2.5 ABI; 2.1.x is the known-good pair.
numpy==2.1.3
omegaconf
einops
pillow
soundfile
tqdm
protobuf