> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/fastgen/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/fastgen/_mcp/server. # Choose a method [View source on GitHub](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/README.md) Training methods for fast single-step or few-step generation from diffusion models. **Multimodal distillation:** DMD2, PDD, and SFT natively support multimodal data (e.g., joint video+audio). When the input is a dict of tensors (e.g., `{"video": ..., "audio": ...}`), losses are computed independently per modality and summed automatically. Per-modality CFG scales can be specified as a dict (e.g., `guidance_cfg.cfg.scale={"video": 3.0, "audio": 7.0}`). ## Methods Overview | Category | README | Method | Class | Description | Reference | | ------------------------- | ---------------------------------------- | ------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Trajectory Matching** | [README](/fastgen/trajectory-matching) | CM | [`CMModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/CM.py) | Consistency model (tuning/distillation) | [Song et al. (2023)](https://arxiv.org/abs/2303.01469), [Geng et al. (2024)](https://arxiv.org/abs/2406.14548) | | | | sCM | [`SCMModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/sCM.py) | Continuous-time CM with TrigFlow | [Lu & Song (2024)](https://arxiv.org/abs/2410.11081) | | | | TCM | [`TCMModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/TCM.py) | Two-stage CM with boundary loss | [Lee et al. (2024)](https://arxiv.org/abs/2410.14895) | | | | MeanFlow | [`MeanFlowModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/mean_flow.py) | Mean velocity prediction | [Geng et al. (2025)](https://arxiv.org/abs/2505.13447), [Sabour et al. (2025)](https://arxiv.org/abs/2506.14603) | | | | AYF | [`AYFModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/ayf.py) | Flow-map distillation with the EMD objective | [Sabour et al. (2025)](https://arxiv.org/abs/2506.14603) | | | | KD | [`KDModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/KD.py) | Learn from pre-computed trajectories | [Luhman & Luhman (2021)](https://arxiv.org/abs/2101.02388) | | | | CausalKD | [`CausalKDModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/KD.py) | Causal version of KD | [Yin et al. (2024)](https://arxiv.org/abs/2412.07772), [Zhu et al. (2026)](https://arxiv.org/abs/2602.02214) | | | | PDD | [`PDDModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/trajectory_matching/pdd.py) | Predict multiple trajectory velocities per network evaluation | [Shaul et al. (2026)](https://arxiv.org/abs/2607.26004) | | **Distribution Matching** | [README](/fastgen/distribution-matching) | DMD2 | [`DMD2Model`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/distribution_matching/dmd2.py) | VSD + GAN distillation | [Yin et al. (2024)](https://arxiv.org/abs/2405.14867) | | | | f-distill | [`FdistillModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/distribution_matching/f_distill.py) | f-divergence weighted DMD2 | [Xu et al. (2025)](https://arxiv.org/abs/2502.15681) | | | | LADD | [`LADDModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/distribution_matching/ladd.py) | Pure adversarial distillation | [Sauer et al. (2024)](https://arxiv.org/abs/2403.12015) | | | | CausVid | [`CausVidModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/distribution_matching/causvid.py) | Causal DMD2 with diffusion forcing | [Yin et al. (2024)](https://arxiv.org/abs/2412.07772) | | | | AnyFlow | [`AnyFlowModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/distribution_matching/anyflow.py) | Any-step DMD2 with a flow-map student | [Gu et al. (2026)](https://arxiv.org/abs/2605.13724) | | | | Self-Forcing | [`SelfForcingModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/distribution_matching/self_forcing.py) | Causal DMD2 with self-forcing | [Huang et al. (2025)](https://arxiv.org/abs/2506.08009) | | **Fine-Tuning** | [README](/fastgen/fine-tuning) | SFT | [`SFTModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/fine_tuning/sft.py) | Finetuning with denoising score matching | [Ho et al. (2020)](https://arxiv.org/abs/2006.11239), [Song et al. (2020)](https://arxiv.org/abs/2011.13456), [Lipman et al. (2022)](https://arxiv.org/abs/2210.02747), [Albergo et al. (2023)](https://arxiv.org/abs/2303.08797) | | | | CausalSFT | [`CausalSFTModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/fine_tuning/sft.py), [`CausalTeacherForcingSFTModel`](https://github.com/NVlabs/FastGen/blob/7889625364c56fa02d2a54c1683c99a98bc891c3/fastgen/methods/fine_tuning/sft.py) | Causal finetuning with diffusion/teacher forcing | [Yin et al. (2024)](https://arxiv.org/abs/2412.07772), [Zhu et al. (2026)](https://arxiv.org/abs/2602.02214) |