huggingface · kashif · Dec 16, 2025 · Dec 16, 2025 · Dec 16, 2025 · Dec 16, 2025
diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml
@@ -504,6 +504,8 @@
         title: AuraFlow
       - local: api/pipelines/blip_diffusion
         title: BLIP-Diffusion
+      - local: api/pipelines/block_refinement
+        title: Block Refinement
       - local: api/pipelines/bria_3_2
         title: Bria 3.2
       - local: api/pipelines/bria_fibo
@@ -542,6 +544,8 @@
         title: DDPM
       - local: api/pipelines/deepfloyd_if
         title: DeepFloyd IF
+      - local: api/pipelines/dflash
+        title: DFlash
       - local: api/pipelines/diffedit
         title: DiffEdit
       - local: api/pipelines/dit
@@ -562,6 +566,8 @@
         title: Hunyuan-DiT
       - local: api/pipelines/hunyuanimage21
         title: HunyuanImage2.1
+      - local: api/pipelines/hybrid_token_diffusion
+        title: Hybrid Token Diffusion
       - local: api/pipelines/pix2pix
         title: InstructPix2Pix
       - local: api/pipelines/kandinsky
@@ -614,6 +620,8 @@
         title: Sana Sprint
       - local: api/pipelines/sana_video
         title: Sana Video
+      - local: api/pipelines/sdar
+        title: SDAR
       - local: api/pipelines/self_attention_guidance
         title: Self-Attention Guidance
       - local: api/pipelines/semantic_stable_diffusion
@@ -659,6 +667,8 @@
         title: Stable Diffusion
       - local: api/pipelines/stable_unclip
         title: Stable unCLIP
+      - local: api/pipelines/token_diffusion
+        title: Token Diffusion
       - local: api/pipelines/unclip
         title: unCLIP
       - local: api/pipelines/unidiffuser
@@ -697,6 +707,8 @@
         title: Kandinsky 5.0 Video
       - local: api/pipelines/latte
         title: Latte
+      - local: api/pipelines/llada2
+        title: LLaDA2
       - local: api/pipelines/ltx2
         title: LTX-2
       - local: api/pipelines/ltx_video
@@ -720,6 +732,8 @@
   - sections:
     - local: api/schedulers/overview
       title: Overview
+    - local: api/schedulers/bd3lm_token_diffusion
+      title: BD3LMTokenDiffusionScheduler
     - local: api/schedulers/block_refinement
       title: BlockRefinementScheduler
     - local: api/schedulers/cm_stochastic_iterative
@@ -740,6 +754,8 @@
       title: DDPMScheduler
     - local: api/schedulers/deis
       title: DEISMultistepScheduler
+    - local: api/schedulers/dflash_token_diffusion
+      title: DFlashTokenDiffusionScheduler
     - local: api/schedulers/multistep_dpm_solver_inverse
       title: DPMSolverMultistepInverse
     - local: api/schedulers/multistep_dpm_solver
@@ -766,6 +782,8 @@
       title: HeliosScheduler
     - local: api/schedulers/heun
       title: HeunDiscreteScheduler
+    - local: api/schedulers/hybrid_token_diffusion
+      title: HybridTokenDiffusionScheduler
     - local: api/schedulers/ipndm
       title: IPNDMScheduler
     - local: api/schedulers/stochastic_karras_ve
@@ -786,8 +804,12 @@
       title: ScoreSdeVeScheduler
     - local: api/schedulers/score_sde_vp
       title: ScoreSdeVpScheduler
+    - local: api/schedulers/sdar_token_diffusion
+      title: SDARTokenDiffusionScheduler
     - local: api/schedulers/tcd
       title: TCDScheduler
+    - local: api/schedulers/token_diffusion
+      title: TokenDiffusionScheduler
     - local: api/schedulers/unipc
       title: UniPCMultistepScheduler
     - local: api/schedulers/vq_diffusion

diff --git a/docs/source/en/api/pipelines/dflash.md b/docs/source/en/api/pipelines/dflash.md
@@ -0,0 +1,24 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# DFlash
+
+`DFlashPipeline` performs block-diffusion speculative decoding using a diffusion draft model and a target causal LM.
+The draft model is conditioned on target hidden features extracted during prefill and verification steps.
+
+## DFlashPipeline
+[[autodoc]] DFlashPipeline
+    - all
+    - __call__
+
+## DFlashPipelineOutput
+[[autodoc]] pipelines.DFlashPipelineOutput
diff --git a/docs/source/en/api/pipelines/hybrid_token_diffusion.md b/docs/source/en/api/pipelines/hybrid_token_diffusion.md
@@ -0,0 +1,23 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Hybrid Token Diffusion
+
+`HybridTokenDiffusionPipeline` is an alias of `TokenDiffusionPipeline` for hybrid-transition schedulers.
+
+## HybridTokenDiffusionPipeline
+[[autodoc]] HybridTokenDiffusionPipeline
+    - all
+    - __call__
+
+## TokenDiffusionPipelineOutput
+[[autodoc]] pipelines.TokenDiffusionPipelineOutput
diff --git a/docs/source/en/api/pipelines/overview.md b/docs/source/en/api/pipelines/overview.md
@@ -48,11 +48,14 @@ The table below lists all the pipelines currently available in 🤗 Diffusers an
 | [Dance Diffusion](dance_diffusion) | unconditional audio generation |
 | [DDIM](ddim) | unconditional image generation |
 | [DDPM](ddpm) | unconditional image generation |
+| [DFlash](dflash) | text2text |
+| [SDAR](sdar) | text2text |
 | [DeepFloyd IF](deepfloyd_if) | text2image, image2image, inpainting, super-resolution |
 | [DiffEdit](diffedit) | inpainting |
 | [DiT](dit) | text2image |
 | [Flux](flux) | text2image |
 | [Hunyuan-DiT](hunyuandit) | text2image |
+| [Hybrid Token Diffusion](hybrid_token_diffusion) | text2text |
 | [I2VGen-XL](i2vgenxl) | image2video |
 | [InstructPix2Pix](pix2pix) | image editing |
 | [Kandinsky 2.1](kandinsky) | text2image, image2image, inpainting, interpolation |
@@ -85,6 +88,7 @@ The table below lists all the pipelines currently available in 🤗 Diffusers an
 | [T2I-Adapter](stable_diffusion/adapter) | text2image |
 | [Text2Video](text_to_video) | text2video, video2video |
 | [Text2Video-Zero](text_to_video_zero) | text2video |
+| [Token Diffusion](token_diffusion) | text2text |
 | [unCLIP](unclip) | text2image, image variation |
 | [UniDiffuser](unidiffuser) | text2image, image2text, image variation, text variation, unconditional image generation, unconditional audio generation |
 | [Value-guided planning](value_guided_sampling) | value guided sampling |

diff --git a/docs/source/en/api/pipelines/sdar.md b/docs/source/en/api/pipelines/sdar.md
@@ -0,0 +1,23 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# SDAR
+
+`SDARPipeline` performs block diffusion decoding with iterative remasking strategies.
+
+## SDARPipeline
+[[autodoc]] SDARPipeline
+    - all
+    - __call__
+
+## SDARPipelineOutput
+[[autodoc]] pipelines.SDARPipelineOutput
diff --git a/docs/source/en/api/pipelines/token_diffusion.md b/docs/source/en/api/pipelines/token_diffusion.md
@@ -0,0 +1,24 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Token Diffusion
+
+`TokenDiffusionPipeline` provides a generic token-space diffusion sampler for discrete denoising over token IDs. It
+pairs a token denoiser model with a token diffusion scheduler.
+
+## TokenDiffusionPipeline
+[[autodoc]] TokenDiffusionPipeline
+    - all
+    - __call__
+
+## TokenDiffusionPipelineOutput
+[[autodoc]] pipelines.TokenDiffusionPipelineOutput
diff --git a/docs/source/en/api/schedulers/dflash_token_diffusion.md b/docs/source/en/api/schedulers/dflash_token_diffusion.md
@@ -0,0 +1,22 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# DFlashTokenDiffusionScheduler
+
+`DFlashTokenDiffusionScheduler` implements the acceptance and posterior sampling logic used in DFlash-style block
+diffusion speculative decoding.
+
+## DFlashTokenDiffusionScheduler
+[[autodoc]] DFlashTokenDiffusionScheduler
+
+## DFlashTokenDiffusionSchedulerOutput
+[[autodoc]] schedulers.scheduling_dflash_token_diffusion.DFlashTokenDiffusionSchedulerOutput
diff --git a/docs/source/en/api/schedulers/hybrid_token_diffusion.md b/docs/source/en/api/schedulers/hybrid_token_diffusion.md
@@ -0,0 +1,22 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# HybridTokenDiffusionScheduler
+
+`HybridTokenDiffusionScheduler` defines hybrid discrete token diffusion updates with separate transitions for
+masked and unmasked tokens.
+
+## HybridTokenDiffusionScheduler
+[[autodoc]] HybridTokenDiffusionScheduler
+
+## HybridTokenDiffusionSchedulerOutput
+[[autodoc]] schedulers.scheduling_hybrid_token_diffusion.HybridTokenDiffusionSchedulerOutput
diff --git a/docs/source/en/api/schedulers/overview.md b/docs/source/en/api/schedulers/overview.md
@@ -54,6 +54,25 @@ Many schedulers are implemented from the [k-diffusion](https://github.com/crowso
 | exponential              | init with `timestep_spacing="linspace"`, `use_exponential_sigmas=True`     |
 | beta                     | init with `timestep_spacing="linspace"`, `use_beta_sigmas=True`            |
 
+## Token diffusion schedulers
+
+These schedulers operate over categorical token IDs instead of continuous latents. They are designed for discrete
+token diffusion models and expose the same `set_timesteps`/`step` interface as other schedulers.
+
+Differences between the discrete token schedulers:
+- `TokenDiffusionScheduler`: token-level diffusion with per-token corruption (e.g. mask/uniform) and a single-step `step` to denoise logits.
+- `HybridTokenDiffusionScheduler`: hybrid transitions that combine token- and block-wise updates in the same schedule.
+- `DFlashTokenDiffusionScheduler`: block diffusion scheduler specialized for speculative decoding with a draft model and target acceptance.
+- `SDARTokenDiffusionScheduler`: block diffusion scheduler with remasking strategies (sequential/low-confidence/entropy-bounded) per step.
+
+[[autodoc]] TokenDiffusionScheduler
+
+[[autodoc]] HybridTokenDiffusionScheduler
+
+[[autodoc]] DFlashTokenDiffusionScheduler
+
+[[autodoc]] SDARTokenDiffusionScheduler
+
 All schedulers are built from the base [`SchedulerMixin`] class which implements low level utilities shared by all schedulers.
 
 ## SchedulerMixin

diff --git a/docs/source/en/api/schedulers/sdar_token_diffusion.md b/docs/source/en/api/schedulers/sdar_token_diffusion.md
@@ -0,0 +1,21 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# SDARTokenDiffusionScheduler
+
+`SDARTokenDiffusionScheduler` implements block diffusion remasking and sampling for SDAR-style decoding.
+
+## SDARTokenDiffusionScheduler
+[[autodoc]] SDARTokenDiffusionScheduler
+
+## SDARTokenDiffusionSchedulerOutput
+[[autodoc]] schedulers.scheduling_sdar_token_diffusion.SDARTokenDiffusionSchedulerOutput
diff --git a/docs/source/en/api/schedulers/token_diffusion.md b/docs/source/en/api/schedulers/token_diffusion.md
@@ -0,0 +1,22 @@
+<!--Copyright 2025 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# TokenDiffusionScheduler
+
+`TokenDiffusionScheduler` defines discrete token diffusion updates over categorical token IDs and supports multiple
+forward processes and alpha schedules.
+
+## TokenDiffusionScheduler
+[[autodoc]] TokenDiffusionScheduler
+
+## TokenDiffusionSchedulerOutput
+[[autodoc]] schedulers.scheduling_token_diffusion.TokenDiffusionSchedulerOutput