- LTX-2: the refine settings are read from pipeline.ltx2.refine.* and
pipeline.components.upsampler_weights. The checkpoint's model_index.json
defaults fill the fields that are still None, and the remaining switches
get their defaults, in one recorded override that rebinds the pipeline's
config. An explicit input value now always wins over the checkpoint
default; the flat sentinels (the generic refine_* names and the class
default of the step count) are gone.
- MiniMax-H3: the sequential load, decode backend, TAEH3, parallel VAE,
offload, and checkpoint schedule settings are read from typed paths.
- Wan, MMAudio, MagiHuman, Stable Audio: flow_shift, dmd_denoising_steps,
output_type, workload_type, offload, and autocast settings are read from
typed paths. MagiHuman no longer reads the never-set log_level_progress
name.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
- ComponentState (fastvideo/pipelines/component_state.py) holds each
pipeline's component paths and residency, which runtime code kept on
FastVideoArgs as model_paths and model_loaded.
- torch_compile_kwargs(resolved_config) derives the torch.compile keyword
arguments from engine.compile.
- thaw(value) returns a mutable copy of a value read from a resolved
config.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
ResolvedGeneratorConfig becomes the runtime config of every mode. Resolution
builds the model's PipelineConfig once, freezes it, and attaches it as
resolved_config.pipeline_config; it also exposes mode, training_mode, and
inference_mode. VideoGenerator, fastvideo serve, the OpenAI server state,
the executors, the worker, and ComposedPipelineBase receive the resolved
config instead of a FastVideoArgs, and generator_config_to_fastvideo_args
is no longer on the inference path.
- Schema: ExecutionMode and WorkloadType move to fastvideo/api/schema.py;
typed homes for mode, output_type, boundary_ratio, master_port,
moba_config, and the LTX-2 refine settings; Enum parsing.
- Resolution: steps for the refine preset overrides, flat experimental
keys, the V-MoBA config, the validation that check_fastvideo_args did,
the num_gpus bump, the model's vae_tiling default, and a final
fill_runtime_defaults step for typed fields no earlier step decides.
- fastvideo/api/training_schema.py: TrainingRunConfig and
PreprocessRunConfig roots, their resolvers, and a --config plus dotted
override loader.
- fastvideo/api/device_policy.py: the device offload policy as functions
that return an overridden config; the worker keeps the result. The LTX-2
and MiniMax-H3 checkpoint defaults rebind the pipeline's config.
- fastvideo/api/flat_name_fallback.py answers the flat FastVideoArgs names
on the resolved config until the readers move to typed paths.
Golden changes: the decisions of the steps above are recorded, and
FastVideoArgs.ltx2_vae_tiling holds the model's vae_tiling (it held None;
no runtime code reads it). Against the 3f6893a0 goldens every
FastVideoArgs field, PipelineConfig, and SamplingParam matches, except
disable_autocast and boundary_ratio, which FastVideoArgs held at their
defaults because flattening routed them only to PipelineConfig.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
VideoGenerator.from_pretrained accepts the 24 common engine and offload
keywords in FROM_PRETRAINED_KWARGS. Any other keyword raises TypeError
with the config path to use with VideoGenerator.from_config. The
conversion of the removed flat keywords (LTX-2 refine names, component
config prefixes, pipeline_config objects, empty-string paths) is removed
from from_pretrained_kwargs_to_config.
The golden kwargs cases that used removed keywords move to a config
category of typed configs; their golden files are byte-identical. The
DreamVerse contract test and the GPU pool forward-translation tests
checked only the removed keyword mapping; the GPU pool tests keep the
typed-config flattening checks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
VideoGenerator.generate(request) is the one inference entry point. The
generate_video keyword call, its mapping from flat keywords to a request
(legacy_generate_call_to_request and its helpers), and the tests of that
mapping are removed. Comments and docs that named generate_video describe
generate and request extensions instead.
scripts/ltx2_sr_alignment.py keeps its generate_video call, because its
reference run imports the separate FastVideo-internal package.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
The image routes call VideoGenerator.generate with a GenerationRequest
instead of passing flat generate_video keywords. The flat projection
_build_generation_kwargs in video_api.py had no runtime caller, so it is
removed, and its tests exercise build_generation_request instead.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
build_generation_request flattened the serve config's default_request with
explicit_request_updates, merged the request body into the flat dict, and then
rebuilt a nested GenerationRequest through legacy_generate_call_to_request,
the generate_video keyword mapping. It now builds the nested request directly:
- default_request contributes its explicit fields with their sections, through
the new compat.explicit_request_raw;
- each value collected from the OpenAI request goes to a fixed typed path from
_REQUEST_PATHS, and any other collected value goes to extensions.
The resulting SamplingParam is unchanged. A request-body field still wins over
the same-named field in default_request.stage_overrides, because
request_to_sampling_param applies stage overrides to sampling fields of the same
name. The only difference in the typed request: an operator stage override
that the body does not set now stays under stage_overrides instead of moving
into sampling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
resolve_generation_request in fastvideo/api/resolution.py runs resolution
steps over the settings sections of a GenerationRequest (sampling, runtime,
output, stage_overrides) and freezes them as a ResolvedRequest. The explicit
paths are the ones the request tracked while it was parsed or built. Prompts,
inputs, state, plan, and extensions are request data and are left out, so
large tensors are not copied.
fastvideo/api/request_resolution.py adds resolve_request(request, model_path).
Its step, fill_sampling_defaults[preset <name>], sets each setting that the
request did not write to the model's preset default
(SamplingParam.from_pretrained). This is the same source that
request_to_sampling_param starts from, and a test checks that the two agree
field by field.
VideoGenerator attaches the ResolvedRequest to each GenerationResult as
resolved_request, so the source of every setting can be looked up. The golden
request snapshots record request_resolution_decisions; nothing else changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
A request's embedded_cfg_scale (request extension or legacy generate_video
keyword) was applied to a deep copy of FastVideoArgs.pipeline_config in
VideoGenerator. The worker runs the pipeline with its own FastVideoArgs
(GpuWorker.execute_forward passes self.fastvideo_args), so the copy never
reached the denoising stages and the request value was ignored.
embedded_cfg_scale is now a per-request sampling value:
- the request field request.sampling.embedded_cfg_scale (the extension key
still works);
- SamplingParam.embedded_cfg_scale and ForwardBatch.embedded_cfg_scale;
- embedded_cfg_scale_for_batch(batch, fastvideo_args), which the denoising,
SR denoising, Wan DMD, and Flux2 text-encoding stages read. It returns the
batch value, else pipeline_config.embedded_cfg_scale, so training and
requests without the field are unchanged.
VideoGenerator no longer copies or changes FastVideoArgs per request, and
request_to_pipeline_overrides is removed. The golden snapshots record the new
SamplingParam field, and the request with extensions.embedded_cfg_scale now
carries 3.0 on SamplingParam instead of on a pipeline override.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
generator_config_to_fastvideo_args passes the ResolvedGeneratorConfig to
FastVideoArgs through the resolved_config init-only argument. Such an
object keeps the config as resolved_config, skips the environment folds that
resolution has already decided, and becomes read-only after construction.
Its PipelineConfig becomes read-only too:
- Assigning a configuration field raises AttributeError and names
FastVideoArgs.override.
- Runtime state stays writable: model_paths, model_loaded, and the Ray
placement handles. So do nested component configs, which loaders fill
with checkpoint data.
- A FastVideoArgs built directly, as the training and argparse entry points
do, is unchanged and stays writable.
FastVideoArgs.override(source, values) is the one way to change fields after
construction. It records (source, values) in override_log and records the
change on resolved_config for fields that have a typed path. The writers that
changed configuration in place now go through it:
- device_policy:* for the unified-memory, lazy-load, MPS, and layerwise
offload decisions that each worker makes after binding its device;
- checkpoint:model_index.json for the LTX-2 refine defaults;
- checkpoint:fastvideo_inference.json for the FastH3 DMD schedule;
- request for the per-request embedded_cfg_scale copy.
Behavior change: an explicit false for engine.compile.regional or
pipeline.minimax_h3.vae_parallel_decode/encode now wins over
FASTVIDEO_INFERENCE_TORCH_COMPILE and FASTVIDEO_VAE_PARALLEL_*. The resolution
steps fill these fields only while they are unset, and FastVideoArgs no
longer folds the environment a second time.
A test run that froze every FastVideoArgs logged no other non-test writer in
the stages, pipelines, loader, worker, hooks, entrypoints, and platforms
suites.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
FastVideoArgs._apply_ltx2_vae_overrides turns on PipelineConfig.vae_tiling
when an LTX-2 VAE tile size is set and ltx2_vae_tiling is unset. The
resolution step derive_vae_tiling_from_ltx2_tile_sizes makes the same
decision on pipeline.vae_tiling, so the resolved config holds it with its
source.
PipelineConfig is unchanged. In the golden case kwargs/ltx2_vae_tile_sizes,
the carrier field FastVideoArgs.ltx2_vae_tiling is now True instead of
None, because the decision now reaches FastVideoArgs as input.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
The resolution step fill_pipeline_config_defaults[<PipelineConfig class>]
fills each typed field that is still None with the value that the model's
PipelineConfig holds for the field's flat name. That PipelineConfig is the
one flattening starts from: the registry class for model_path, plus a
pipeline config file or object when one is given. The covered fields are the
engine.precision fields, pipeline.vae_sp, flow_shift, embedded_cfg_scale,
dmd_denoising_steps, and the pipeline.longcat BSA settings.
The resolved config therefore holds the effective value of these fields,
and provenance tells a model default apart from user input. The step runs
after the environment steps and before the derived values, so the precedence
is user input, then environment variables, then model defaults.
inference_resolution_steps(config) builds the ordered step list.
The values that the step fills are the values that PipelineConfig already
had, so FastVideoArgs and PipelineConfig are unchanged; the golden snapshots
add the decisions. Tests that flatten unregistered model paths skip this
step.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
An unsupported FASTVIDEO_ATTENTION_BACKEND name, such as a typo, was ignored
and the run fell back to automatic backend selection without telling the
user. An explicit attention_backend argument with the same name raised.
get_env_variable_attn_backend now raises ValueError for an unsupported name,
using the same parse as an explicit request (coerce_attn_backend). Every
reader goes through it: the fill_attention_backend_from_env resolution
step, FastVideoArgs.__post_init__, and the attention selector's
environment fallbacks.
The golden case environment/attention_backend_unknown_name now records the
error, and the registry description and env_vars.md table say that an
unsupported name raises.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
The first resolution steps take over startup rewrites that
FastVideoArgs.__post_init__ and check_fastvideo_args make in place:
- fill_attention_backend_from_env: FASTVIDEO_ATTENTION_BACKEND fills
engine.attention.backend while it is unset.
- fill_regional_compile_from_env: FASTVIDEO_INFERENCE_TORCH_COMPILE turns
engine.compile.regional on.
- fill_vae_parallel_from_env: the FASTVIDEO_VAE_PARALLEL_* variables set
pipeline.minimax_h3, and the decode strategy defaults to gather.
- derive_parallel_sizes: the -1 placeholders of tp_size, sp_size, and
hsdp_shard_dim become 1, num_gpus, and num_gpus.
Each step repeats the current FastVideoArgs rule exactly, so FastVideoArgs is
unchanged: the golden snapshots differ only in their resolution decisions.
The FastVideoArgs code stays for FastVideoArgs built directly by the
training and argparse entry points; for a resolved config it finds the
values already decided.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
generator_config_to_fastvideo_args now runs resolve_inference_config, in
the new fastvideo/api/inference_resolution.py, before it flattens the
config. The resolution goes through the ordered, provenance-recording driver
in fastvideo/api/resolution.py. VideoGenerator.from_config keeps the frozen
result as resolved_config, so the source of every startup value can be
inspected.
A mapping input is the raw user input, and every leaf in it counts as
explicit. A GeneratorConfig object does not record which fields the caller
wrote. written_fields therefore treats the fields that differ from the
schema defaults as explicit, and parsing its output gives back an equal
config.
INFERENCE_RESOLUTION_STEPS is empty in this commit, so FastVideoArgs is
unchanged. The golden snapshots of the YAML and from_pretrained cases add
the list of resolution decisions, which is empty here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
The from_pretrained keyword ltx2_refine_lora_path set
pipeline.components.lora_path, which reaches FastVideoArgs.lora_path and loads
as the main LoRA on the stage-1 transformer. The refine stage reads
FastVideoArgs.ltx2_refine_lora_path, which stayed unset. The conversion also
turned an empty string into None. An empty string is how callers such as the
Dreamverse GPU pool disable the refine LoRA, and None makes LTX2Pipeline load
the checkpoint default instead: fastvideo_refine_lora_path in model_index.json
is FastVideo/LTX2-Distilled-LoRA for the LTX-2 Diffusers checkpoints.
The refine LoRA gets its own field, pipeline.ltx2.refine.lora_path, whose
flat name is ltx2_refine_lora_path. None keeps the checkpoint default, and an
empty string disables the refine LoRA. The Dreamverse contract test, the
Dynamo contract example, and the parity inventory point at the field.
Golden changes: kwargs/ltx2_refine_lora_path now records the path on
ltx2_refine_lora_path instead of lora_path, and the added case
kwargs/ltx2_refine_lora_disabled pins the empty string.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
The schema class MiniMaxH3Config had the same name as the MiniMax-H3 DiT
config class in fastvideo/configs/models/dits/minimax_h3.py. The family
partitions of PipelineSelection are renamed LTX2Options, LTX2RefineOptions,
MiniMaxH3Options, and LongCatOptions, so their names cannot be confused with
the model and pipeline config classes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
Parameters that belong to one model family, or to one component config,
could be set only through argparse flags or pipeline.experimental. They gain
typed GeneratorConfig fields, each with the flat name that FastVideoArgs or
PipelineConfig uses:
- pipeline.dit and pipeline.vae: dicts of DiT and VAE config field
overrides, passed as dit_config.<key> and vae_config.<key>
- pipeline.ltx2: VAE tile sizes, initial video and audio latent paths, noise
order, distilled sigmas, and refine.{transformer,noise,audio_noise}_path
- pipeline.minimax_h3: sequential load, video decode backend, TAEH3
checkpoint and chunk size, and the sequence-parallel VAE switches
- pipeline.longcat: the block sparse attention (BSA) settings
legacy_from_pretrained_to_config routes dit_config.* and vae_config.*
keywords into the new dicts. Every new scalar field defaults to None and is
not passed to FastVideoArgs while unset. Behavior is unchanged: the golden
config snapshots match.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
These inference parameters could be set only through argparse flags or the
untyped pipeline.experimental dict. Each one gains a typed GeneratorConfig
field whose flat name is its FastVideoArgs or PipelineConfig field, so YAML,
dotted CLI overrides, and from_pretrained keywords all reach it:
- engine.attention: backend, vsa_sparsity, vsa_tile_size, moba_config_path
- engine.precision: dit, vae, vae_decode, image_encoder, text_encoders
- engine.compile.regional (inference_torch_compile)
- pipeline: vae_sp, flow_shift, embedded_cfg_scale, dmd_denoising_steps
- pipeline.components.lora_target_modules
Every field defaults to None, which keeps the runtime or model default; a
None field is not passed to FastVideoArgs. pipeline.experimental still
accepts the flat names and still wins over the typed fields. The parity
inventory lists the typed paths, and the docs show the typed forms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
Each GeneratorConfig field that has a name in the flat keyword API
(VideoGenerator.from_pretrained kwargs and FastVideoArgs fields) declares it
through flat_field(...) metadata. legacy_from_pretrained_to_config and
generator_config_to_fastvideo_args read those names instead of keeping two
hand-written mappings, so adding a typed field with a flat name needs no
compat change.
The keywords that need a conversion keep explicit branches: the
torch_compile_kwargs split, the transformer_quant instance, the LTX-2 refine
preset keys, empty-string refine paths, and a non-string pipeline_config.
_SPECIALLY_MAPPED_PATHS lists every field without a flat name, and
test_generator_config_flat_names.py fails when a field has neither.
Behavior is unchanged: the 150 golden config snapshots match. The flat
keyword dict no longer passes None for revision, dist_timeout, and
lazy_module_load, and always passes the default LoRA nickname and strength and
empty per-component compile dicts; each equals the FastVideoArgs default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
Add fastvideo/api/resolution.py. resolve_generator_config parses a raw
GeneratorConfig mapping and runs an ordered list of steps over it. Each
step receives a read-only ResolutionView and returns {dotted_path: value};
it never assigns fields. The driver records every decision as
(source, values), where source is the step's qualified name, and the last
decision for a path wins. Explicit paths are taken from the keys of the
raw mapping, so "the user set this" is never inferred from defaults.
The result is a frozen ResolvedGeneratorConfig with the same nesting as
GeneratorConfig, per-path provenance, and with_override(source, values)
as the only way to change it after resolution; every override is logged.
The module is not wired into FastVideoArgs, compat.py, or any entry
point yet, so runtime behavior is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012M91wnVFmPEvJ9h39r5BH7
