World models · interactive-world-model · game-world-model · action-conditioned-video · image-to-video

SCOPE

zizhaotong

Official SCOPE inference source, DiffSynth fork, README, license, and examples are vendored under worldfoundry.synthesis.visual_generation.scope.scope_runtime.

Runner verified统一环境scope

构建运行命令

选择仓库中已记录的 variant，然后复制对应的准备、检查或运行命令。

worldfoundry.pipeline

Variant

任务interactive-world-model

环境worldfoundry-unified-cu128

设备CUDA 12.8

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

兼容性与版本

由 Manifest 生成

集成状态: verified
Runner 证据: verified
环境: worldfoundry-unified-cu128
Python: 3.10
CUDA: CUDA 12.8
PyTorch: torch
源码 revision: —
Checkpoint revision: —

Runtime profile: scope
Pipeline binding: scope
Runner: worldfoundry.pipeline
Pipeline target: worldfoundry.pipelines.scope.pipeline_scope:SCOPEPipeline
Backend stage: in_tree_official_cli
Runtime 状态: in_tree_scope_runtime_checkpoint_pending
Driver 状态: compatible
环境类型: 统一环境

安装环境

环境解析器会读取已记录 profile，并选择此处显示的统一或独立环境。

bash scripts/setup/model_env_install.sh --model scope

依赖版本约束 22

torch
torchvision
numpy
pillow
opencv-python
transformers>=4.57.0,<5
accelerate
huggingface_hub
safetensors
einops
modelscope
sentencepiece
protobuf
ftfy
tqdm
imageio
imageio[ffmpeg]
pandas
pyarrow
pynvml
peft
typing_extensions

Conda 依赖 5

python=3.10
pip
ffmpeg
setuptools
wheel

Checkpoint 与资产

分配算力前先做本地检查；gated、private 与 license 字段直接来自 checkpoint manifest。

worldfoundry-eval zoo model-download --model-id scope --check-local --json

zizhaotong/SCOPE

Revision: —
License: —
Gated: —
Private: —

运行与输出

生成的命令通过共享 evaluation 边界运行，并持久保存结果 manifest 与 artifact。

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

输入契约

字段	已记录契约
`actions`	action_parquet
`image`	Required
`prompt`	Required
`video`	Optional

Artifact 契约

Artifact 类型	文件名 / 路径
`generated_video`	`—`
`generated_video`	`scope.mp4`

证据与来源

Catalog 集成、原生 demo parity 与 runner parity 是三条独立记录。

集成状态: verified
Runner 证据: verified
原生 Demo 证据: 未记录
验证 Imports: torch, torchvision, numpy, PIL, cv2, transformers, accelerate, huggingface_hub, safetensors, einops, modelscope, sentencepiece, pandas, pyarrow, imageio, peft, worldfoundry.base_models.diffusion_model.diffsynth, worldfoundry.base_models.diffusion_model.diffsynth.models.scope_dit, worldfoundry.base_models.diffusion_model.diffsynth.pipelines.scope_pipeline

配方溯源

worldfoundry/data/models/catalog/world_models/scope.yamlProjecthttps://z2tong.github.io/SCOPE/Paperhttps://arxiv.org/abs/2605.23345GitHubhttps://github.com/z2tong/SCOPEHugging Facehttps://huggingface.co/zizhaotong/SCOPE

构建运行命令

输入契约

Artifact 契约

相关配方