Video · text-to-video · image-to-video

Wan2.1

Wan-AI

Only official Wan-Video GitHub and Wan-AI Hugging Face repos are listed.

Runner verified统一环境源码已固定wan2.1

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text-to-video

环境worldfoundry-unified-cu128

设备CUDA 12.8

worldfoundry-eval evaluate \
  --mode model \
  --model-id wan2.1-t2v-1.3b \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/wan2.1-t2v-1.3b \
  --metric artifact_count \
  --json

兼容性与版本

由 Manifest 生成

集成状态: integrated
Runner 证据: verified
环境: worldfoundry-unified-cu128
Python: 3.11
CUDA: CUDA 12.8
PyTorch: torch
源码 revision: 9737cba9c1c3c4d04b33fcad41c111989865d315
Checkpoint revision: 37ec512624d61f7aa208f7ea8140a131f93afc9a

Runtime profile: wan2.1-t2v-1.3b-832x480-81f
Pipeline binding: wan2.1
Runner: worldfoundry.pipeline
Pipeline target: worldfoundry.pipelines.wan.pipeline_wan_2p1_t2v:Wan2p1T2VPipeline
Backend stage: in_tree_runtime_plan
Runtime 状态: catalog_demo_and_runner_parity_recorded_checkpoint_runtime_required
Driver 状态: compatible_unified_default
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 wan2.1-t2v-1.3b

依赖版本约束 4

torch
torchvision
torchaudio
xfuser

Conda 依赖 5

python=3.11
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 wan2.1 --check-local --json

Wan-AI/Wan2.1-T2V-1.3B

Revision: 37ec512624d61f7aa208f7ea8140a131f93afc9a
License: apache-2.0
Gated: false
Private: false

Wan-AI/Wan2.1-T2V-14B

Revision: a064a6c71f5be440641209c07bf2a5ce7a2ff5e4
License: apache-2.0
Gated: false
Private: false

Wan-AI/Wan2.1-I2V-14B-480P

Revision: 6b73f84e66371cdfe870c72acd6826e1d61cf279
License: apache-2.0
Gated: false
Private: false

Wan-AI/Wan2.1-I2V-14B-720P

Revision: 8823af45fcc58a8aa999a54b04be9abc7d2aac98
License: apache-2.0
Gated: false
Private: false

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 wan2.1-t2v-1.3b \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/wan2.1-t2v-1.3b \
  --metric artifact_count \
  --json

输入契约

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

Artifact 契约

Artifact 类型	文件名 / 路径
`generated_video`	`wan2.1-t2v-1.3b-832x480-81f.mp4`
`expected artifact`	`wan21_t2v_1p3b_seed42_832x480_81f.mp4`
`expected artifact`	`wan21_t2v_1p3b_worldfoundry_seed42_832x480_81f.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: integrated
Runner 证据: verified
原生 Demo 证据: verified
验证 Imports: torch, torchvision, numpy, transformers, diffusers, xfuser, worldfoundry

配方溯源

worldfoundry/data/models/catalog/video/wan2.1.yamlGitHub9737cba9c1c3c4d04b33fcad41c111989865d315

构建运行命令

输入契约

Artifact 契约

相关配方