Video · text-to-image · text-to-video · video-to-video · image-conditioned-generation

Sana

Efficient-Large-Model

Official Sana source, configs, ControlNet runner, Sprint runner, and video runner are vendored under worldfoundry.base_models.diffusion_model.image.sana.

Runtime Ported统一环境sana

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-image

环境worldfoundry-unified-cu128

设备CUDA 12.8

worldfoundry-eval evaluate \
  --mode model \
  --model-id sana-600m-512px \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/sana-600m-512px \
  --metric artifact_count \
  --json

兼容性与版本

由 Manifest 生成

集成状态: runtime ported
Runner 证据: pending
环境: worldfoundry-unified-cu128
Python: 3.11
CUDA: CUDA 12.8
PyTorch: torch
源码 revision: —
Checkpoint revision: bfbcf738fec4e4bb059a68d9d09c9c4bfc4a1cde

Runtime profile: sana
Pipeline binding: sana
Runner: worldfoundry.pipeline
Pipeline target: worldfoundry.pipelines.sana.pipeline_sana:SanaPipeline
Backend stage: in_tree_official_cli
Runtime 状态: in_tree_official_cli_streaming_variants_a100_workspace_validated
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 sana-600m-512px

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 sana --check-local --json

Efficient-Large-Model/Sana_600M_512px

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

Efficient-Large-Model/Sana_600M_1024px

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

Efficient-Large-Model/Sana_1600M_512px

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

Efficient-Large-Model/Sana_1600M_512px_MultiLing

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

Efficient-Large-Model/Sana_1600M_1024px

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

Efficient-Large-Model/Sana_1600M_1024px_MultiLing

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

Efficient-Large-Model/Sana_1600M_1024px_BF16

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

Efficient-Large-Model/Sana_1600M_2Kpx_BF16

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

Efficient-Large-Model/Sana_1600M_4Kpx_BF16

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

Efficient-Large-Model/SANA1.5_1.6B_1024px

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

Efficient-Large-Model/SANA1.5_4.8B_1024px

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

Efficient-Large-Model/Sana_Sprint_0.6B_1024px

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

Efficient-Large-Model/Sana_Sprint_1.6B_1024px

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

Efficient-Large-Model/Sana_600M_1024px_ControlNet_HED

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

Efficient-Large-Model/Sana_1600M_1024px_BF16_ControlNet_HED

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

Efficient-Large-Model/SANA-Video_2B_480p

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

Efficient-Large-Model/SANA-Video_2B_720p

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

Efficient-Large-Model/SANA-Video_2B_480p_LongLive

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

Efficient-Large-Model/SANA-Streaming

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

Efficient-Large-Model/SANA-Streaming_bidirectional

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

Efficient-Large-Model/Sana_1600M_1024px

Revision: bfbcf738fec4e4bb059a68d9d09c9c4bfc4a1cde
License: —
Gated: —
Private: —

Efficient-Large-Model/SANA-Streaming

Revision: 534c946811822f70117b8549b7e072120a4414d2
License: —
Gated: —
Private: —

Efficient-Large-Model/SANA-Streaming_bidirectional

Revision: 9c68733952a5fa6f96f597a18211fcc493ac66e2
License: —
Gated: —
Private: —

Lightricks/LTX-2

Revision: 47da56e2ad66ce4125a9922b4a8826bf407f9d0a
License: —
Gated: —
Private: —

Efficient-Large-Model/gemma-2-2b-it

Revision: 569d9809d0c8b6722d4d31b5a77a2ec7a400650a
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 sana-600m-512px \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/sana-600m-512px \
  --metric artifact_count \
  --json

输入契约

字段	已记录契约
`image`	Required
`prompt`	Required
`video`	Required

Artifact 契约

Artifact 类型	文件名 / 路径
`generated_image`	`sana.png`

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: runtime ported
Runner 证据: pending
原生 Demo 证据: 未记录
验证 Imports: torch, torchvision, numpy, transformers, diffusers, xfuser, worldfoundry

配方溯源

worldfoundry/data/models/catalog/video/sana.yamlProjecthttps://nvlabs.github.io/Sana/Paperhttps://arxiv.org/abs/2410.10629GitHubhttps://github.com/NVlabs/SanaHugging Facehttps://huggingface.co/Efficient-Large-Model/Sana_1600M_1024pxHugging Facehttps://huggingface.co/Efficient-Large-Model/SANA-StreamingHugging Facehttps://huggingface.co/Efficient-Large-Model/SANA-Streaming_bidirectional

构建运行命令

输入契约

Artifact 契约

相关配方