This repository adapts the official ShandaAI/FloodDiffusion model at revision b70b9d7e7e7002eec136bf763245d76361fcf4a1. The official source is Apache-2.0 and the upstream implementation is also available at ShandaAI/FloodDiffusion.
The official diffusion and VAE safetensors were loaded strictly in the QuietPhoenix Ascend container. The adapted runtime completed a real npu:0 forward with torch_npu, using the official minimum legal denoise schedule of five steps. The result is finite and is a smoke-only execution record, not a full precision or generation-quality claim.
| Item | Result |
|---|---|
| Device | npu:0 |
| Model revision | b70b9d7e7e7002eec136bf763245d76361fcf4a1 |
| Latent output | [1,4] |
| Decoded output | [1,263] |
| Dtype | torch.float32 |
| Finite | true |
| Parameters | 131,221,455 including VAE |
| NPU forward | 0.420409s |
complex64; Ascend rejects the official complex128 cat path.Weights are intentionally excluded from Git by .gitignore; model-lock.json records the official revision, file sizes, and SHA-256 values. Place model.safetensors and vae.safetensors beside ldf.yaml before running.
source /usr/local/Ascend/ascend-toolkit/set_env.sh
python inference.py --model-dir . --device npu --output validation/npu-runtime.json

