SolitaryThinker
7ac2ff0d1c
Per review: dispatch should be a directly-mapped HF-string -> card registry (like fastvideo), shared by every entrypoint (VideoGenerator + a future CLI / server), not buried in the generator. - New v2/registry.py mirrors fastvideo's fastvideo/registry.py hybrid resolution: (1) exact HF repo id in an explicit ModelEntry registry [PRIMARY — correct per-model card/capabilities, and the only way to split same-architecture capability variants like Wan2.1 T2V vs the i2v 'InP' 1.3B], (2) short repo-name match, (3) architecture inference [FALLBACK — local paths / unregistered repos]. resolve(model_path[, root]) is the single shared entry point. - video_generator.py: moved _read_arch_signature/_select_builders into the registry; from_config now calls resolve() — registered ids resolve with no config read, else arch inference on a cheap *.json snapshot. Reconciles the earlier 'no brittle table' refactor with the 'map hf string -> card' ask: one clean registry + a fallback, not three coupled structures. - Verified: registry resolves exact-id / short-name / arch-fallback / unregistered correctly; CPU suite green (202+2); wan21 GPU smoke generates via the new resolve path.
75 lines
7.2 KiB
Markdown
75 lines
7.2 KiB
Markdown
# v2 porting status — fastvideo models → the v2 (recipe, runtime) substrate
|
||
|
||
Goal: drive each `examples/inference/basic/` model through the **v2 `VideoGenerator`** (typed
|
||
`fastvideo.api` configs + the real torch backend) and generate video, with a matching `v2_*` example.
|
||
|
||
**Active scope (per request): the Wan family + LTX-2 only.** The Cosmos/Hunyuan/SD3.5/LongCat/Flux2/
|
||
audio/interactive families below are kept as an upstream reference map but are out of the current scope.
|
||
|
||
Dispatch is **architecture-driven** (`v2/video_generator.py`): `from_config` reads the checkpoint's
|
||
pipeline / transformer / VAE class names (+ `z_dim`, `transformer_2`) and picks the v2 card — mirroring
|
||
fastvideo's `get_pipeline_config_cls_from_name`, with no hardcoded repo-id table.
|
||
|
||
The v2 mini implements a small set of loops/samplers (**flow-match Euler**, causal **chunk_rollout**,
|
||
**FlowGRPO SDE**) and adapters (Wan DiT/VAE/UMT5, causal Wan, LTX-2 DiT/VAE/Gemma/**spatial upsampler**).
|
||
A model ports **cleanly** only when its architecture *and* sampler match what v2 implements; otherwise it
|
||
needs new per-model work (a card, an adapter for the DiT/VAE/encoder forward, and/or a new sampler), and
|
||
some are blocked by the environment (gated weights, local-only weights, non-video modality, no VSA).
|
||
|
||
The v2 bring-up runs **single-GPU, resident, on the `TORCH_SDPA` backend** (no fastvideo-kernel / VSA /
|
||
FP4), so distilled models that depend on VSA/SLA/FP4 kernels or non-flow-match samplers won't match.
|
||
(Wan2.2-A14B MoE is the exception to "resident": its two 14B experts are CPU-offloaded and swapped onto
|
||
the GPU one at a time at the boundary-timestep transition, so it fits a single 80GB GPU.)
|
||
|
||
**Environment status:** the box was rescheduled from an aarch64 host to an x86_64 host mid-session
|
||
(`ipp1a1…` → `ipp2-0493`). The aarch64 `.venv` couldn't execute on x86, so the venv was **rebuilt for
|
||
x86** (torch 2.11.0+cu128 x86 + numpy 2.5.0rc1 x86 + fastvideo[dev]) and re-validated: the v2 CPU suite
|
||
is green and wan21 + LTX-2 base both generate real video on the x86 stack. (If the box flips arch again,
|
||
rebuild the venv for the current arch — system python `/usr/bin/python3.12`, reinstall the matching
|
||
`uv`, `uv pip install -e ".[dev]"` with `fastvideo-kernel` commented out.)
|
||
|
||
## Working today (verified, real video) — committed on `v2`
|
||
| Official example(s) | Model | v2 card | v2 example |
|
||
|---|---|---|---|
|
||
| `basic.py`, `basic_mps.py`*, `basic_ray.py`* | `Wan-AI/Wan2.1-T2V-1.3B-Diffusers` | wan21 | `v2_basic.py`, `v2_basic_new_api.py` |
|
||
| `basic_self_forcing_causal.py` | `wlsaidhi/SFWan2.1-T2V-1.3B-Diffusers` | wan_causal | `v2_basic_self_forcing_causal.py`, `v2_basic_new_api.py` |
|
||
| `basic_ltx2_distilled.py`, `basic_ltx2_distilled_fast_profile.py` | `FastVideo/LTX2-Distilled-Diffusers` | ltx2 — real `LTX2LatentUpsampler` between base/refine (un_normalize→learned 2× upsample→normalize) | `v2_basic_new_api.py` |
|
||
| `basic_wan2_2_ti2v.py` (T2V branch) | `Wan-AI/Wan2.2-TI2V-5B-Diffusers` | wan2.2-ti2v | `v2_basic_wan2_2_ti2v.py` |
|
||
| `basic_wan2_2.py` | `Wan-AI/Wan2.2-T2V-A14B-Diffusers` (MoE) | wan2.2-a14b + **CPU expert offload** (one 14B expert resident at a time, swapped at the boundary → 60GB peak) | `v2_basic_wan2_2.py` |
|
||
| `basic_ltx2.py` | `Davids048/LTX2-Base-Diffusers` | ltx2 base (single-stage, request-driven many-step) | `v2_basic_ltx2.py` |
|
||
| `basic_ltx2_3_distilled.py` | `FastVideo/LTX-2.3-Distilled-Diffusers` (18.99B) | ltx2 base card (single-stage; few-step) | `v2_basic_ltx2_3_distilled.py` |
|
||
|
||
All seven rows GPU-verified to generate real video (LTX-2 base / 2.3 show inter-frame motion 4.5 / 6.6
|
||
on the rebuilt x86 stack). \* `basic_mps.py` (Apple MPS) and `basic_ray.py` (ray executor) are
|
||
device/executor variants of the same Wan2.1 model — covered by wan21; those runtime modes are out of scope.
|
||
|
||
## Needs per-model work (architecture/sampler matches partially)
|
||
| Official example(s) | Model | What v2 needs |
|
||
|---|---|---|
|
||
| `basic_dmd.py`, `basic_dmd_new_api.py` | FastWan2.1-T2V-1.3B (WanDMDPipeline) | **Blocked (env):** loading needs a *non-strict* load to skip the VSA-only `to_gate_compress` (root cause: the checkpoint key mis-maps under the generic Wan path, and `TransformerLoader` is strict except Cosmos2.5 — `component_loader.py:1005`). Even loaded, the DMD model is trained for VSA, which is **not built here (no nvcc)**, so SDPA output would be degraded. Needs the fastvideo-kernel VSA build + a non-strict WanDMD load. |
|
||
| `basic_turbodiffusion*.py` | TurboWan2.1/2.2 (RCM) | **New sampler:** Reparameterized Consistency Model (`fastvideo/models/schedulers/scheduling_rcm.py:RCMScheduler`) — v2 has only flow-match/chunk/SDE loops; needs a new RCM consistency loop. |
|
||
| `basic_wan2_2_i2v.py`, `basic_wan2_2_Fun.py` | Wan2.2-I2V / Fun | **I2V mechanism (identified, `pipelines/stages/image_encoding.py`):** SigLIP `image_encoder`+`image_processor` → `image_embeds` (the DiT's `encoder_hidden_states_image`, already accepted) **and** VAE-encode the first frame, concat `[noisy_latent, image_latent, mask]` (first-frame mask=1) → an i2v DiT (larger in_channels). v2 needs: image_encoder/image_processor components+adapters, a VAE-encode-image node, the concat-mask conditioning in the wan loop, the i2v DiT in_channels, `image_path` handling (reuse A14B offload). Fun adds a control input. |
|
||
| `basic_lucy_edit.py` | `decart-ai/Lucy-Edit-Dev` (Wan v2v) | **V2V mechanism (identified):** `WanVideoToVideoPipeline` + `VideoVAEEncodingStage` — VAE-encode the input video to a conditioning latent. v2 needs input-video loading + a VAE-encode-video node + the v2v conditioning (no image_encoder). |
|
||
| `basic_ltx2_3_distilled_i2v*.py` | LTX-2.3-Distilled I2V | Image conditioning on the ltx2 single-stage program (encode the input image + condition the denoise). |
|
||
|
||
## Out of current scope — new families (kept as an upstream reference map only)
|
||
Each would need a new card + adapters (DiT/VAE/encoder forwards) + maybe a sampler:
|
||
`basic_cosmos2_5_*` (Cosmos-Predict2.5), `basic_gen3c.py` (GEN3C — **local `converted_weights/` only**),
|
||
`basic_hy15*.py` / `basic_gamecraft.py` / `basic_hyworld.py` (Hunyuan family),
|
||
`basic_longcat_*.py` (LongCat + LoRA), `basic_lingbotworld_base_cam.py`,
|
||
`basic_sd35_t2i.py` (SD3.5 image), `basic_flux2*.py` (**gated** black-forest-labs).
|
||
|
||
## Different modality / interactive (out of the t2v VideoGenerator surface)
|
||
`basic_stable_audio*.py` (audio), `basic_matrixgame2*.py` / `basic_matrixgame3.py`
|
||
(interactive mouse/keyboard-conditioned world models).
|
||
|
||
## How to add a model to the v2 VideoGenerator
|
||
1. Ensure a v2 card exists (`v2/models/<family>/`) whose `load_id`s resolve to real `fastvideo.models.*`.
|
||
2. Register it in the shared `v2/registry.py` (used by every entrypoint — VideoGenerator, CLI, server):
|
||
add a `ModelEntry` mapping its HF id(s) → builders (the explicit PRIMARY path — the only way to split
|
||
same-architecture capability variants, e.g. T2V vs the i2v "InP"), and/or a branch in
|
||
`select_by_architecture` for the arch-inference FALLBACK (local paths / unregistered repos).
|
||
3. If a DiT/VAE/encoder/upsampler forward differs, add an adapter + class-detection in `torch_adapters.py`
|
||
and register the component kind in `torch_cuda.py`.
|
||
4. Add `examples/inference/basic/v2_<name>.py` mirroring the official example; generate to confirm.
|