--- title: FastH3 4-step Preview emoji: 🎬 colorFrom: red colorTo: gray sdk: gradio sdk_version: 6.26.0 app_file: app.py short_description: 4-step MiniMax-H3 with sparse attention β€” video + audio python_version: "3.12" startup_duration_timeout: 1h suggested_hardware: zero-a10g --- # FastH3 4-step Preview (VSA, data-free) β€” MiniMax-H3 in four forward passes [`FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree`](https://huggingface.co/FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree) is a **data-free DMD2 distillation** of [`MiniMaxAI/MiniMax-H3`](https://huggingface.co/MiniMaxAI/MiniMax-H3), the 33B dual-modality DiT that emits a video *and* its synchronized soundtrack from one denoising pass. The student keeps the teacher's architecture exactly and only replaces `transformer/`, so it is a drop-in for the same `diffusers` modular pipeline β€” but it needs **four** DiT forwards instead of thirty, and it was distilled **with Video Sparse Attention on**. Everything runs unquantized at **bfloat16**. ## The sampling contract: five grid points, four forwards `MiniMaxH3Scheduler.set_timesteps(n)` builds `linspace(1, 0, n)`, applies the shift `Οƒ' = sΒ·Οƒ / (1 + (sβˆ’1)Β·Οƒ)`, and then drops the trailing zero when it forms the timesteps β€” so **`n` sigma grid points drive `n βˆ’ 1` model evaluations**. The distilled ladder is `t = 999, 749, 500, 250 β†’ 0`: five points, four forwards. This Space therefore fixes `num_inference_steps = 5`, matching the checkpoint's own `fastvideo_inference.json` (`num_inference_steps: 5`, `transformer_forwards: 4`, `dmd_denoising_steps: [999, 749, 500, 250]`, `guidance_scale: 1.0`). It is not a knob. There is no CFG and no negative prompt β€” the teacher is guidance-distilled and the student inherits that. ## Video Sparse Attention is not optional here This checkpoint is the **VSA** variant. Its `fastvideo_inference.json` pins `attention_backend: VIDEO_SPARSE_ATTN_H3`, `vsa_sparsity: 0.9`, `vsa_tile_size: 64`, and the transformer ships 50 trained `attn.to_gate_compress` tensors (~3.6 GiB) that only the sparse path consumes. FastVideo publishes a separate `…-Dense-DataFree` checkpoint for people who want dense β€” running *this* one dense is running it off-distribution. The published kernel is `vsa_kernel: sm100a`, which is GB200-only, and the ZeroGPU pool is **sm120** (RTX PRO 6000 Blackwell). FastVideo's other officially supported route is `--vsa-kernel triton`, which is pure Triton and architecture-agnostic β€” so this Space vendors those two files verbatim from [FastVideo](https://github.com/hao-ai-lab/FastVideo) (Apache-2.0) into `vsa_kernel/` and ports the H3 backend on top of them: | File | What it is | |---|---| | `vsa_kernel/block_sparse_attn_triton.py` | FastVideo's Triton block-sparse attention, verbatim except that the autotune sweep is collapsed to the single config it lands on for Blackwell (re-enable it with `H3_VSA_AUTOTUNE=1`). | | `vsa_kernel/index.py` | FastVideo's `map_to_index` / `topk_index_to_map`, verbatim. | | `vsa_h3.py` | The port: `MiniMaxH3VSAAttnProcessor`, a `diffusers` attention processor reproducing `MiniMaxH3VSABackend`. | `vsa_h3.py` follows FastVideo's `video_sparse_attn_h3.py` step for step: 64-token `(4, 4, 4)` tiles over the post-patchify video grid, segment-pure prefix tiles, per-head fp32 pooled tile scores, `topk = max(1, min(⌈(1 βˆ’ sparsity)Β·n_video_tilesβŒ‰, n_video_tiles))`, prefix keys exempt (always selected) and prefix queries always dense, plus the compression branch `softmax(scores) @ pool(v)` broadcast back over each tile row and scaled by the trained gate with **no** activation. The tile geometry is derived per-forward from the pipeline's own `token_tags` / `position_ids`, so the `[text | cond | audio | video]` packing stays authoritative. `diffusers` 0.40.0 has no `to_gate_compress`, so `vsa_h3.add_gate_compress_modules()` patches `MiniMaxH3TransformerBlock.__init__` **before** the pipeline loads; otherwise the 50 gate tensors load as "unexpected keys" and are silently dropped. Blocks whose gate is all-zero have it removed again after load, exactly as FastVideo does. A hidden `/selftest` API endpoint runs the ported kernel at sparsity 0 against `F.scaled_dot_product_attention` on a real packed layout, so a wrong tile order or transpose is caught without spending a generation. ## Split across two Spaces MiniMax-H3 is ~196 GiB in bfloat16 and a ZeroGPU Space is evicted at **150 GB of storage**, so no single unquantized Space can hold it. Cutting `MiniMaxH3Blocks` at its `text_encoder` step splits it in two, and both halves fit: | Space | Subfolders | Download | |---|---|---| | [`multimodalart/qwen3vl-conditioner`](https://huggingface.co/spaces/multimodalart/qwen3vl-conditioner) | `text_encoder/` + `tokenizer/` + `processor/` | 66.7 GB β†’ 62.15 GiB bf16 | | this one | `transformer/` + `vae/` + `audio_vae/` | ~85 GB β†’ 75.7 GiB resident | The conditioner is a **public** Space and is unchanged by the distillation β€” the student's `modular_model_index.json` points its `text_encoder` at the same Qwen3-VL weights β€” so this Space reuses it as-is and calls it over the gradio API for every request. `h3_split_blocks.py` subclasses the pipeline's blocks with the `text_encoder` step removed. Dropping the step drops the components it declares, so `load_components` resolves only `transformer` / `vae` / `audio_vae` / the two schedulers, and `prompt_embeds` + `text_token_tags` become ordinary required pipeline inputs. The wire format is those two tensors β€” `(1, num_text_tokens, 5120)` bfloat16 and `(num_text_tokens,)` int64 β€” in one safetensors file with the resolved `height` / `width` / `num_frames` in its metadata header. ## Text-to-video+audio only The preview distills **the T2VA path only**, so this Space exposes no keyframe / reference inputs β€” the student's `transformer_ref` tower is not packaged with the checkpoint. Image conditioning is what [`multimodalart/minimax-h3`](https://huggingface.co/spaces/multimodalart/minimax-h3) (the undistilled 30-step teacher) is for. ## Prompt format MiniMax-H3 is trained on a structured multimodal caption, not a bare sentence: ``` integrated_multimodal_description: overall_soundscape: non_diegetic_music: ``` **Expand prompt** (on by default) sends a short prompt through the conditioner's Qwen3-VL prompt rewriter, which writes that structure for you and returns it. Turn it off when you have already written a full-format prompt β€” the examples that carry MiniMax's own official prompts do exactly that. ## Examples The two long examples are MiniMax's own published prompts, taken verbatim from the base model's repo (Apache-2.0 code / docs): - the starship-bridge two-shot from `scripts/readme/reproducible-768p-t2va-request.sh` - the bakery two-shot, Case 1 of `docs/VIDEO_PROMPT_WRITING_GUIDE_base_en.md` ## Generation constraints Fixed by the checkpoint: 24 fps, a 768 pixel short edge at the training canvas, `num_frames` snapped up to the next `17Β·n + 5`. The distillation's operating point is **1344Γ—768 Γ— 124 frames (β‰ˆ5 s)** β€” that is the default, and it is exactly the layout VSA was tuned on (`grid = 37Γ—24Γ—42`, 37 296 video rows, 672 tiles, 66 selected). The duration slider reaches 8 s and the canvas dropdown offers smaller/faster grids, both outside the distilled operating point, so quality degrades gracefully rather than being guaranteed. The 8 s ceiling is a VRAM limit, not a model limit: 75.7 GiB of weights sit resident on a 95.0 GiB card and the sparse working set grows with sequence length. The checkpoint itself is a **preview**. ## Measured On this Space, over `gradio_client`. Startup β€” the ~85 GB download plus the load, with no placement and therefore no pack β€” is **78 s**. | Canvas Γ— frames | Packed rows | Video tiles kept | Denoise + decode | Per forward | Peak allocated | |---|---|---|---|---|---| | 544Γ—544 Γ— 56 (2.3 s) | 4 913 | 10/100 | 6 s | 1.4 s | 81.05 GiB | | 1344Γ—768 Γ— 124 (5.2 s) | 37 296 | 66/660 | 59–72 s | 14.7–18.0 s | 83.70 GiB | | 1344Γ—768 Γ— 192 (8.0 s) | 57 456 | 102/1020 | 87 s | 21.9 s | 88.00 GiB | Sparse attention makes the cost **linear** in the packed rows rather than quadratic, so `get_duration` drops the quadratic term a dense path needs. The spread on the 5 s point is how fast a slice of the pool the request lands on; the fit is taken at the slow end (`1.95e-3 Β· rows`) so a slow slice is not aborted mid-video. That books 98 s at the 5 s default and 143 s at the 8 s maximum. A cold worker pays a one-time **68 s**: 75.7 GiB across PCIe (~11 s) plus the Triton JIT of the vendored kernels (~57 s). The same 544Γ—544 request measures 74 s cold against 6 s warm. That is booked only on the first request a process serves rather than padded onto every request. At 88.00 GiB on a 95.0 GiB card the 8 s ceiling leaves ~7 GiB β€” which is why the slider stops at 8 s and not 10 s. The vendored kernels are checked against `F.scaled_dot_product_attention` on the live GPU by the hidden `/selftest` endpoint: at `sparsity = 0` the ported path reproduces dense attention to **4.1e-03 relative / 0.999996 cosine**, which is bf16 rounding. ## Placement `H3_PLACEMENT=lazy`: the weights move onto the card on the first GPU call and stay there. `spaces`' startup `torch.pack()` would write a second on-disk copy of every resident CUDA tensor, and 85 + 75 GB exceeds the 150 GB quota, so packing is not an option here. The one-time `.to("cuda")` plus the Triton JIT of the block-sparse kernels lands inside the first request of a cold worker; after that there is **no offloading in the request path at all**. ## Space variables | Variable | Default | Meaning | |---|---|---| | `H3_MODEL_REPO` | `FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree` | The distilled diffusers-layout checkpoint. Public. | | `H3_CONDITIONER` | `multimodalart/qwen3vl-conditioner` | The public Space this one asks for embeddings. | | `H3_ATTENTION` | `vsa` | The ported VSA-H3 backend. Any other value is passed to `set_attention_backend` as a dense escape hatch (e.g. `_native_cudnn`). | | `H3_VSA_SPARSITY` | `0.9` | The checkpoint's trained sparsity. | | `H3_VSA_AUTOTUNE` | unset | Set to `1` to restore FastVideo's full Triton autotune sweep instead of the pinned Blackwell config. | | `H3_PLACEMENT` | `lazy` | `lazy` moves all weights onto the card on the first GPU call; `offload` hands placement to `ComponentsManager.enable_auto_cpu_offload`. | | `H3_GPU_SIZE` | `xlarge` | ZeroGPU allocation size. `large` (48 GB) does not fit 75 GiB of weights. | ## Secrets None. Every weight this Space downloads is public, and the conditioner is a public Space called on the requesting user's own ZeroGPU token. ## License The weights are under the **MiniMax H3 Community License**, inherited from the base model β€” it carries territory and acceptable-use restrictions. Read [the base model's license](https://huggingface.co/MiniMaxAI/MiniMax-H3/blob/main/LICENSE) before using outputs. The vendored kernels in `vsa_kernel/` are Apache-2.0, from [hao-ai-lab/FastVideo](https://github.com/hao-ai-lab/FastVideo).