Skip to main content

← All model recipe tasks

sana_wm model family

Tasks

Supported architectures and task heads

The Hugging Face values below are copied from checkpoint metadata at the recorded revision. The TRTMC task contract comes from the exact E2E recipe. Architecture identity selects or describes a source graph; it does not imply that TRTMC reproduces every Hugging Face head with the same model_type. For example, an encoder recipe may consume the base model and intentionally return hidden states instead of the checkpoint's pretraining or classification logits.

HF model_typeHF architecture / pipeline classTRTMC task contractExact recipe profiles
SanaMSVideoCamCtrl_1600M_P1_D20
Source: config.model.model
Metadata: config.yaml at e96271d77398
diffusion_media_generation
sana-wm-bidirectional

Declared recipes

Each row comes from a model-owned E2E manifest. It declares a test recipe; it is not a live pass receipt for every hardware target.

RecipeExact Hugging Face checkpointTaskBuild configurationDeclared E2E casesManifest
sana-wm-bidirectionalEfficient-Large-Model/SANA-WM_bidirectionalbf16; single devicesana-wm-bidirectionaltests/e2e/models/sana_wm/manifests/sana-wm-bidirectional.json

Family-owned configuration

These keys are extracted from the family's registered config schema. Pass one key per repeatable --set namespace.field=value argument. Build-time keys belong on trtmc build; session keys belong on the family's inference command.

sana_wm

Schema source: src/runtime/models/sana_wm/config_schema.cpp

--set keyTypeDefaultCLI surfaceAllowed configuration layers
sana_wm.image_pathstring""runtimeSession Request, Platform Profile
sana_wm.actionstring""runtimeSession Request, Platform Profile
sana_wm.translation_speedfloat-1.0runtimeSession Request, Platform Profile
sana_wm.rotation_speed_degfloat-1.0runtimeSession Request, Platform Profile
sana_wm.num_framesint64-1runtimeSession Request, Platform Profile
sana_wm.fpsint64-1runtimeSession Request, Platform Profile
sana_wm.flow_shiftfloat-1.0runtimeSession Request, Platform Profile
sana_wm.intrinsicsstring""runtimeSession Request, Platform Profile
sana_wm.no_refinerboolfalseruntimeSession Request, Platform Profile
trtmc generate-video <bundle.bundle> <task-inputs> --set sana_wm.image_path=<value>

The registry rejects unknown namespaces and fields. A --config file can set the same keys in JSON or YAML form.

Family-specific CLI contracts

Inputs and options below are filtered by the declared E2E task and the methods and configuration fields used by that family's native runtime implementation. The global CLI parser accepts a wider union of flags; flags absent here are not declared for this family.

trtmc generate-video

Generate video frames from a text prompt.

Declared recipes: sana-wm-bidirectional

trtmc generate-video <bundle.bundle> --prompt "<text>" --output <output-dir> [generation options]
Supported input or optionRequirementRuntime behavior
--prompt <TEXT>RequiredText conditioning input.
--output <DIR>OptionalDirectory for generated PNG frames.
--seed <N>OptionalSet the sampling or diffusion seed.
--num-steps <N>OptionalSet the number of diffusion or flow-matching steps.
--guidance-scale <F>OptionalSet classifier-free guidance strength.
--cfg-scale <F>OptionalSet the conditional classifier-free guidance scale.
--initial-latents-raw <PATH>OptionalLoad packed float32 initial latents.
--negative-prompt <TEXT>OptionalOverride the bundle default negative prompt.
Code basis

Runtime provider: sana_wm

  • src/cli/main.cpp
  • include/trtmc/pipeline.h
  • tests/e2e/models/sana_wm/manifests/sana-wm-bidirectional.json
  • src/runtime/models/sana_wm/pipeline.cpp
  • src/runtime/models/sana_wm/pipeline.h

Bundle lifecycle commands

These commands apply to every declared recipe in this family; they do not add task inputs or model capabilities.

trtmc build

Build one exact checkpoint into a TensorRT-Model-Connect bundle.

trtmc build <hf-id> --output <bundle.bundle>

trtmc inspect

Inspect bundle metadata, runtime identity, and packaged sections.

trtmc inspect <bundle.bundle>

See the CLI Reference for all options and limitations.