SECourses Wan 2.1 Windows App Now Supports Sage Attention - 37% Faster than Flash Attention - 1-Click to Install
Added 2025-03-25 12:10:41 +0000 UTCHopefully many more features coming and we already added so many features.
App link : https://www.patreon.com/posts/123105403
I will hopefully make a full new how to use tutorial once all features are completed.
1-Click to install on Windows into a Python 3.10 VENV - no WSL
First tutorial : https://youtu.be/hnAhveNy-8s
Second tutorial : https://youtu.be/ueMrzmbdWBg
Audio generation tutorial : https://youtu.be/504f8S4MLTw
Interface below and at attachments
The above video generated as below
Prompt: Close-up shot of a smiling young boy with a joyful expression, sitting comfortably in a cozy room. The boy has tousled brown hair and wears a colorful t-shirt. Bright, soft lighting highlights his happy face. Medium close-up, slightly tilted camera angle.
Negative Prompt: Overexposure, static, blurred details, subtitles, paintings, pictures, still, overall gray, worst quality, low quality, JPEG compression residue, ugly, mutilated, redundant fingers, poorly painted hands, poorly painted faces, deformed, disfigured, deformed limbs, fused fingers, cluttered background, three legs, a lot of people in the background, upside down
Used Model: WAN 2.1 14B Text-to-Video
Number of Inference Steps: 20
CFG Scale: 6
Sigma Shift: 10
Seed: 200
Number of Frames: 81
Extend Factor: 1x
Number of Segments: 1
Extension Segment: 0 of 0
Used Last Frame From: original
Input Image:
Denoising Strength: N/A
LoRA Model: None
TeaCache Enabled: True
TeaCache L1 Threshold: 0.15
TeaCache Model ID: Wan2.1-T2V-14B
Precision: BF16
Auto Crop: Enabled
Final Resolution: 1280x720
Comments
v56 has sage attention. in that case after installing run Use_Flash_Attention_Instead_Of_Sage_Attention.bat and it should work but on my system 3090 works very fine with sage attention also did you follow requirements video? https://youtu.be/DrhUHnYfwC0
Furkan Gözükara
2025-03-27 09:32:12 +0000 UTCHi, I just did that but still the same Problem. v52 works v56 not. The same v56 zip-file works on my other system without this problem. Python, cuda etc. at both systems is the same.
MathiasF
2025-03-27 08:55:04 +0000 UTChi easy. run Clear_Triton_Cache.bat and then delete venv and run installer again and it should fix
Furkan Gözükara
2025-03-26 22:56:19 +0000 UTChi could you be using older installer file? did you get a fresh zip file and installed freshly?
Furkan Gözükara
2025-03-26 22:55:40 +0000 UTCNobody else is having troubles with this exception? I always have to fix `DiffSynth-Studio\diffsynth\pipelines\wan_video.py`in order to make it work. Even after doing fresh install. Maybe it's a problem with my models folder? I have it located in `d:\Tools\wan21\Wan2.1\models`. Traceback (most recent call last): File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\gradio\queueing.py", line 625, in process_events response = await route_utils.call_process_api( File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api output = await app.get_blocks().process_api( File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\gradio\blocks.py", line 2137, in process_api result = await self.call_function( File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\gradio\blocks.py", line 1663, in call_function prediction = await anyio.to_thread.run_sync( # type: ignore File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\anyio\to_thread.py", line 56, in run_sync return await get_async_backend().run_sync_in_worker_thread( File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 2470, in run_sync_in_worker_thread return await future File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 967, in run result = context.run(func, *args) File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\gradio\utils.py", line 890, in wrapper response = f(*args, **kwargs) File "d:\Tools\wan21\Wan2.1\App.py", line 1220, in generate_videos loaded_pipeline = load_wan_pipeline(model_choice, torch_dtype, vram_value, lora_path=effective_loras, lora_alpha=None) File "d:\Tools\wan21\Wan2.1\App.py", line 1935, in load_wan_pipeline pipe = WanVideoPipeline.from_model_manager(model_manager, torch_dtype=torch.bfloat16, device=device) File "D:\Tools\wan21\Wan2.1\DiffSynth-Studio\diffsynth\pipelines\wan_video.py", line 142, in from_model_manager pipe.fetch_models(model_manager) File "D:\Tools\wan21\Wan2.1\DiffSynth-Studio\diffsynth\pipelines\wan_video.py", line 131, in fetch_models self.prompter.fetch_tokenizer(os.path.join(os.path.dirname(tokenizer_path), "google/umt5-xxl")) File "D:\Tools\wan21\Wan2.1\DiffSynth-Studio\diffsynth\prompters\wan_prompter.py", line 94, in fetch_tokenizer self.tokenizer = HuggingfaceTokenizer(name=tokenizer_path, seq_len=self.text_len, clean='whitespace') File "D:\Tools\wan21\Wan2.1\DiffSynth-Studio\diffsynth\prompters\wan_prompter.py", line 45, in __init__ self.tokenizer = AutoTokenizer.from_pretrained(name, **kwargs) File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\transformers\models\auto\tokenization_auto.py", line 857, in from_pretrained tokenizer_config = get_tokenizer_config(pretrained_model_name_or_path, **kwargs) File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\transformers\models\auto\tokenization_auto.py", line 689, in get_tokenizer_config resolved_config_file = cached_file( File "d:\Tools\wan21\Wan2.1\venv\lib\site-packages\transformers\utils\hub.py", line 469, in cached_file raise EnvironmentError( OSError: Incorrect path_or_model_id: 'models\google/umt5-xxl'. Please provide either the path to a local folder or the repo_id of a model on the Hub.
Petr K.
2025-03-26 18:46:55 +0000 UTCHi, while v52 works v56 sadly does not :-( System is an AMD 1700x with 80 GB RAM and RTX3090 24 GB vRAM I did a fresh install with v56 and the install script runs without any error messages. But at startup the cmd gives following error: import error: DLL load failed while importing libtriton: W0326 09:28:23.354000 1196 venv\Lib\site-packages\torch\distributed\elastic\multiprocessing\redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs. Traceback (most recent call last): File "E:\Wan21_v56\Wan2.1\App.py", line 25, in from diffsynth import ModelManager, WanVideoPipeline, save_video, VideoData File "E:\Wan21_v56\Wan2.1\DiffSynth-Studio\diffsynth\__init__.py", line 2, in from .models import * File "E:\Wan21_v56\Wan2.1\DiffSynth-Studio\diffsynth\models\__init__.py", line 1, in from .model_manager import * File "E:\Wan21_v56\Wan2.1\DiffSynth-Studio\diffsynth\models\model_manager.py", line 4, in from .downloader import download_models, download_customized_models, Preset_model_id, Preset_model_website File "E:\Wan21_v56\Wan2.1\DiffSynth-Studio\diffsynth\models\downloader.py", line 6, in from ..configs.model_config import preset_models_on_huggingface, preset_models_on_modelscope, Preset_model_id File "E:\Wan21_v56\Wan2.1\DiffSynth-Studio\diffsynth\configs\model_config.py", line 57, in from ..models.wan_video_dit import WanModel File "E:\Wan21_v56\Wan2.1\DiffSynth-Studio\diffsynth\models\wan_video_dit.py", line 21, in from sageattention import sageattn File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\sageattention\__init__.py", line 1, in from .core import sageattn, sageattn_varlen File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\sageattention\core.py", line 20, in from .triton.quant_per_block import per_block_int8 as per_block_int8_triton File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\sageattention\triton\quant_per_block.py", line 18, in import triton File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\__init__.py", line 8, in from .runtime import ( File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\runtime\__init__.py", line 1, in from .autotuner import (Autotuner, Config, Heuristics, autotune, heuristics) File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\runtime\autotuner.py", line 9, in from .jit import KernelInterface File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\runtime\jit.py", line 12, in from ..runtime.driver import driver File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\runtime\driver.py", line 1, in from ..backends import backends File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\backends\__init__.py", line 50, in backends = _discover_backends() File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\backends\__init__.py", line 43, in _discover_backends compiler = _load_module(name, os.path.join(root, name, 'compiler.py')) File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\backends\__init__.py", line 12, in _load_module spec.loader.exec_module(module) File "E:\Wan21_v56\Wan2.1\venv\lib\site-packages\triton\backends\amd\compiler.py", line 2, in from triton._C.libtriton import ir, passes, llvm, amd ImportError: DLL load failed while importing libtriton Any idea?
MathiasF
2025-03-26 08:33:43 +0000 UTCget latest installer zip file extract and use install. it should work
Furkan Gözükara
2025-03-25 13:30:06 +0000 UTCDoes this update require fresh install or simple update will do the trick?
Tomek Deręgowski
2025-03-25 13:28:55 +0000 UTCwell it is based on single test :d but it is significantly faster. thanks all good
Furkan Gözükara
2025-03-25 13:28:36 +0000 UTC37% so very specific lol will give it a whirl thanks Doc. Lot of trouble in Türkiye at the mo hope all is well
Nobodys Hero
2025-03-25 13:21:09 +0000 UTChi sorry for this error. v56 zip file fixed this issue.
Furkan Gözükara
2025-03-25 12:37:25 +0000 UTCHi! I did a fresh install and get this error: Traceback (most recent call last): File "C:\IA\Wan 2.1\Wan2.1\App.py", line 21, in import wan File "C:\IA\Wan 2.1\Wan2.1\wan\__init__.py", line 1, in from . import configs, distributed, modules File "C:\IA\Wan 2.1\Wan2.1\wan\configs\__init__.py", line 7, in from .wan_i2v_14B import i2v_14B File "C:\IA\Wan 2.1\Wan2.1\wan\configs\wan_i2v_14B.py", line 3, in from easydict import EasyDict ModuleNotFoundError: No module named 'easydi
Raúl Tarín Sepúlveda
2025-03-25 12:26:52 +0000 UTC
