Upgrade Hunyuanvideo to use the new diffusers format

README.md

@@ -232,15 +232,9 @@ Below is a list of validated diffusers version requirements. If the model is not
 | Model Name | Diffusers version |
 | --- | --- |
 | [Flux](https://huggingface.co/black-forest-labs/FLUX.1-dev) | >= 0.35.2 |
+| [HunyuanVideo](https://github.com/Tencent/HunyuanVideo) | >= 0.35.2 |
 
 
-#### HunyuanVideo
-
-- Supports `diffusers<=0.32.2` (breaking commit diffusers @ [8907a70](https://github.com/huggingface/diffusers/commit/8907a70a366c96b2322656f57b24e442ea392c7b))
-  - For this constraint, supports `xfuser<=0.4.3post3` (breaking commit xDiT @ [0c20492](https://github.com/xdit-project/xDiT/commit/0c2049281d5f698622d006658fb99d1aa92a9d50))
-
-- Supports bs=1 only
-
 <h2 id="dev-guide">📚  Develop Guide</h2>
 
 We provide a step-by-step guide for adding new models, please refer to the following tutorial.

examples/hunyuan_video_usp_example.py

@@ -3,6 +3,10 @@
 from typing import Any, Dict, Union, Optional
 import logging
 import time
+from xfuser.config.diffusers import has_valid_diffusers_version, get_minimum_diffusers_version
+if not has_valid_diffusers_version("hunyuanvideo"):
+    minimum_diffusers_version = get_minimum_diffusers_version("hunyuanvideo")
+    raise ImportError(f"Please install diffusers>={minimum_diffusers_version} to use HunyuanVideo.")
 
 import torch
 
@@ -76,7 +80,7 @@ def new_forward(
         image_rotary_emb = self.rope(hidden_states)
 
         # 2. Conditional embeddings
-        temb = self.time_text_embed(timestep, guidance, pooled_projections)
+        temb, _ = self.time_text_embed(timestep=timestep, pooled_projection=pooled_projections, guidance=guidance)
         hidden_states = self.x_embedder(hidden_states)
         encoder_hidden_states = self.context_embedder(encoder_hidden_states,
                                                       timestep,

xfuser/config/diffusers.py

@@ -4,6 +4,7 @@
 DEFAULT_MINIMUM_DIFFUSERS_VERSION = "0.33.0"
 MINIMUM_DIFFUSERS_VERSIONS = {
     "flux": "0.35.2",
+    "hunyuanvideo": "0.35.2",
 }
 
 def has_valid_diffusers_version(model_name: str|None = None) -> bool: