WorldFoundry
English中文

World models · three_dimension

HunyuanWorld-Mirror

Tencent Hunyuan / HY-World

WorldFoundry exposes framework inference through the in-tree HunyuanMirrorPipeline and vendored inference-only WorldMirror model code.

Runner verifiedUnified environmenthunyuanworld-mirror

Build your run

Choose a recorded variant, then copy the exact setup or inspection command.

worldfoundry.pipeline
Taskthree_dimension
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
worldfoundry-eval evaluate \
  --mode model \
  --model-id hunyuanworld-mirror \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/hunyuanworld-mirror \
  --metric artifact_count \
  --json
01

Compatibility & versions

Manifest-backed recipe

Integration
integrated
Runner evidence
verified
Environment
worldfoundry-unified-cu128
Python
3.10
CUDA
CUDA 12.8
PyTorch
torch
Source revision
—
Checkpoint revision
—
Runtime profile
hunyuanworld-mirror
Pipeline binding
hunyuanworld-mirror
Runner
worldfoundry.pipeline
Pipeline target
worldfoundry.pipelines.hunyuan_world.pipeline_hunyuan_mirror:HunyuanMirrorPipeline
Backend stage
framework_pipeline
Runtime status
framework_inference_gpu_verified
Driver status
compatible
Environment kind
Unified environment
02

Install environment

The environment resolver reads the recorded profile and chooses the unified or dedicated environment shown here.

bash scripts/setup/model_env_install.sh --model hunyuanworld-mirror
Package constraints 27
  • torch
  • torchvision
  • numpy<2.0.0
  • opencv-python
  • pillow_heif
  • scipy
  • einops
  • torchmetrics
  • trimesh
  • open3d==0.18.0
  • roma
  • matplotlib
  • tqdm
  • huggingface-hub[torch]>=0.22
  • jaxtyping
  • lpips
  • colorspacious
  • moviepy==1.0.3
  • plyfile
  • pycolmap==3.10.0
  • tyro
  • onnxruntime
  • safetensors
  • gradio==5.49.1
  • omegaconf
  • pydantic
  • requests
Conda packages 10
  • python
  • pip
  • ffmpeg
  • git
  • git-lfs
  • cmake
  • ninja
  • pkg-config
  • libgl
  • libglib
03

Checkpoints & assets

Run the local check before allocating compute. Gated, private, and license fields below come directly from the checkpoint manifest.

worldfoundry-eval zoo model-download --model-id hunyuanworld-mirror --check-local --json
tencent/HunyuanWorld-Mirror
Revision
—
License
—
Gated
—
Private
—
04

Launch & outputs

The generated command uses the shared evaluation boundary and writes durable result manifests and artifacts.

worldfoundry-eval evaluate \
  --mode model \
  --model-id hunyuanworld-mirror \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/hunyuanworld-mirror \
  --metric artifact_count \
  --json

Input contract

FieldRecorded contract
imageRequired
promptOptional
videoOptional

Artifact contract

Artifact kindFilename / path
generated_3d_assethunyuan_mirror_outputs.zip
05

Evidence & sources

Catalog integration, native-demo parity, and runner parity are independent records.

Integration
integrated
Runner evidence
verified
Native demo evidence
Not recorded
Validation imports
torch, numpy, cv2, PIL, huggingface_hub, pycolmap, open3d, onnxruntime
Recipe provenance
worldfoundry/data/models/catalog/world_models/hunyuanworld-mirror.yamlGitHubhttps://github.com/Tencent-Hunyuan/HunyuanWorld-MirrorHugging Facehttps://huggingface.co/tencent/HunyuanWorld-Mirror