From NVIDIA Labs SANA Text-to-Image Model 1-Click Installers for Windows, RunPod, Massed Compute and Free Kaggle Account
Added 2025-07-13 22:01:00 +0000 UTCPatreon exclusive posts index to find our scripts easily, Patreon scripts updates history to see which updates arrived to which scripts and amazing Patreon special generative scripts list that you can use in any of your task.
Join discord to get help, chat, discuss and also tell me your discord username to get your special rank : SECourses Discord
Please also Star, Watch and Fork our Stable Diffusion & Generative AI GitHub repository and join our Reddit subreddit and follow me on LinkedIn (my real profile)
=======
Latest zip file : Sana_v18.zip
Full tutorial video : https://youtu.be/KW-MHmoNcqo
14 July 2025 Update
The app now supports RTX 5000 series on Windows, RunPod and Massed Compute
Newer models and features will be added later hopefully
13 January 2024 Update
Example images : https://www.reddit.com/r/SECourses/comments/1hzx1d6/it_is_now_possible_to_generate_16_megapixel/
New 4K Tutorial Video : https://youtu.be/GjENQfHF4W8
Now the APP will use Diffusers Pipeline and it has huge VRAM optimizations
You need to reinstall
The models will be downloaded into your Hugging Face cache folder when you first time generate something
Please make a fresh install
When you enable all 4 optimizations the VRAM usages are like below
Make sure shared VRAM is enabled because initial loading of the model need more VRAM
Enable VAE Tiling + Enable VAE Slicing + Enable Model CPU Offload +
Enable Sequential CPU Offload
1K (1024x1024) : 4 GB GPUs
2K (2048x2048) : 6 GB GPUs
4K (4096x4096) : 8 GB GPUs
Still in any case may work on your GPU test it
Just Enable VAE Tiling + Enable Model CPU Offload works great in many cases
29 December 2024 Update
Following aspect ratios added
1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21, 4:5, 5:4
When you select the aspect ratio, it will auto set the resolution according to either 1K model or 2K model
Windows_Fix_Model_Download.bat added which will resume download if your model download fails
27 December 2024 Update
With V6 update I have significantly optimized the pipeline
Now it will keep only model or VAE at the same time on the GPU, thus it will work way faster when decoding with VAE
Now works way faster on low VRAM GPUs
CUDA not found error fixed
Removed set CUDA_VISIBLE_DEVICES=0 from bat file
When changing model, wait until model is fully loaded
If you had V5, just using Windows update file should be sufficient but also replace your windows start bat file
Kaggle notebook updated
2K model is not working on Kaggle since GPUs are lacking BF16 capability
Pipelines and other code changes from main repository added to our APP
I have improved the Gradio APP significantly and now the Gradio APP supports the newest following model as well
Sana 1K (1024x1024)
Sana 2K (2048x2048)
The model will be loaded when you first time generate an image
The very first generation will be significantly slower than the consequent ones - totally related to their pipeline
I recommend make a fresh install - it will auto download newest models from the official repo automatically
======
Official page : https://nvlabs.github.io/Sana/
I have developed a better version Gradio APP than they have and made 1-click installers for Windows, RunPod, Massed Compute and a free Kaggle Notebook
It took me literally more than 12 hours of work :(
Speeds for generating per image as below for 1024x1024
Free Kaggle : 15 second
RTX 3060 : 9.5 second
RTX 3090 : 4 second
RTX 4090 : 2 second
Uses 9.5 GB VRAM but someone reported works amazing even on 8 GB GPUs
Windows Requirements
Python 3.10, FFmpeg, CUDA 12.8, cuDNN 9.7, C++ Tools, MSVC and Git
If it doesn't work make sure to below NEW tutorial and install everything exactly as shown in this below tutorial
How To Install And Use
Windows Requirements
Python 3.10, FFmpeg, CUDA 12.8, cuDNN 9.7, C++ Tools and Git
If it doesn't work make sure to below NEW tutorial and install everything exactly as shown in this below tutorial
RunPod Instructions
Please register for RunPod at : https://get.runpod.io/955rkuppqv4h
Select a GPU with a minimum of 24 GB VRAM, such as the RTX 3090.
Get a machine with minimum 48 GB RAM, getting multiple GPU machine better you will get more RAM
Choose a RunPod template:
RunPod Pytorch 2.2.0
runpod/pytorch:2.2.0-py3.10-cuda12.1.1-devel-ubuntu22.04
Follow Runpod_Instructions_READ.txt
Massed Compute Instructions
If you get 2x A6000 GPU it would work best with 96 GB RAM but not mandatory
Watch this instructional video to learn how to use Massed Compute:
Register for Massed Compute using the following link:
Important Notes:
Ensure each copy-paste action is accurate, as Massed Compute may not always copy-paste correctly.
Select an A6000 GPU or A6000 GPU ALT (if other one is not available) and use our code SECourses for huge discount.
Follow Massed_Compute_Instructions_READ.txt
Kaggle :
Upload Kaggle notebook and follow the instructions
Make sure verify phone so you can enable internet and select GPU
Example Kaggle install video : https://youtu.be/LeHfgq_lAXU
Comments
hi please also try this. delete triton cache folder and run again : C:\Users\Furkan\.triton\cache instead of Furkan it will be your username
Furkan Gözükara
2025-01-15 00:54:32 +0000 UTCyes your model downloads failed. please delete files in hugging face cache. i have shown how to fix in this video : https://youtu.be/GjENQfHF4W8 watch first 4 minutes
Furkan Gözükara
2025-01-14 19:42:52 +0000 UTChey i got that error : Error: Error loading Sana 1K (1024x1024) model: Error while deserializing header: InvalidHeaderDeserialization and that's the log F:\Sanav15\Sana>call venv\Scripts\activate.bat * Running on local URL: http://127.0.0.1:7860 To create a public link, set `share=True` in `launch()`. A mixture of bf16 and non-bf16 filenames will be loaded. Loaded bf16 filenames: [text_encoder/model.bf16-00002-of-00002.safetensors, vae/diffusion_pytorch_model.bf16.safetensors, text_encoder/model.bf16-00001-of-00002.safetensors, transformer/diffusion_pytorch_model.bf16.safetensors] Loaded non-bf16 filenames: [transformer/diffusion_pytorch_model-00002-of-00002.safetensors, transformer/diffusion_pytorch_model-00001-of-00002.safetensors If this behavior is not expected, please check your folder structure. Loading checkpoint shards: 0%| | 0/2 [00:00
Sher Lock
2025-01-14 19:10:01 +0000 UTCyep we made huge improvements. glad that it is working better now
Furkan Gözükara
2025-01-14 14:59:24 +0000 UTCOh this time It generated so fast. When I first install, it took 15-20 mins to generate 1 image. Thank you. 😺
Ahmet Inceelli
2025-01-14 14:47:55 +0000 UTCyour requirements wrong. please uninstall python and everything and follow every step of this video into exactly same folders and reinstall and send me reinstall logs : https://youtu.be/DrhUHnYfwC0 if you upgrade gold tier i can connect your pc and install
Furkan Gözükara
2025-01-14 14:28:53 +0000 UTCI got this error: D:\Stable_Diffusion\Sana_v15\Sana>call venv\Scripts\activate.bat Traceback (most recent call last): File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 942, in _get_module return importlib.import_module("." + module_name, self.__name__) File "C:\Users\ADMIN\miniconda3\lib\importlib\__init__.py", line 126, in import_module return _bootstrap._gcd_import(name[level:], package, level) File "", line 1050, in _gcd_import File "", line 1027, in _find_and_load File "", line 1006, in _find_and_load_unlocked File "", line 688, in _load_unlocked File "", line 883, in exec_module File "", line 241, in _call_with_frames_removed File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\loaders\lora_pipeline.py", line 31, in from .lora_base import ( # noqa File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\loaders\lora_base.py", line 27, in from ..models.modeling_utils import ModelMixin, load_state_dict File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\models\modeling_utils.py", line 35, in from ..quantizers import DiffusersAutoQuantizer, DiffusersQuantizer File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\quantizers\__init__.py", line 15, in from .auto import DiffusersAutoQuantizer File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\quantizers\auto.py", line 22, in from .bitsandbytes import BnB4BitDiffusersQuantizer, BnB8BitDiffusersQuantizer File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\quantizers\bitsandbytes\__init__.py", line 2, in from .utils import dequantize_and_replace, dequantize_bnb_weight, replace_with_bnb_linear File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\quantizers\bitsandbytes\utils.py", line 32, in import bitsandbytes as bnb File "C:\Users\ADMIN\AppData\Roaming\Python\Python310\site-packages\bitsandbytes\__init__.py", line 15, in from .nn import modules File "C:\Users\ADMIN\AppData\Roaming\Python\Python310\site-packages\bitsandbytes\nn\__init__.py", line 17, in from .triton_based_modules import ( File "C:\Users\ADMIN\AppData\Roaming\Python\Python310\site-packages\bitsandbytes\nn\triton_based_modules.py", line 6, in from bitsandbytes.triton.dequantize_rowwise import dequantize_rowwise File "C:\Users\ADMIN\AppData\Roaming\Python\Python310\site-packages\bitsandbytes\triton\dequantize_rowwise.py", line 12, in import triton File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\__init__.py", line 20, in from .runtime import ( File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\runtime\__init__.py", line 1, in from .autotuner import (Autotuner, Config, Heuristics, autotune, heuristics) File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\runtime\autotuner.py", line 9, in from ..testing import do_bench, do_bench_cudagraph File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\testing.py", line 7, in from . import language as tl File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\language\__init__.py", line 4, in from . import math File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\language\math.py", line 1, in from . import core File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\language\core.py", line 10, in from ..runtime.jit import jit File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\runtime\jit.py", line 12, in from ..runtime.driver import driver File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\runtime\driver.py", line 1, in from ..backends import backends File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\backends\__init__.py", line 50, in backends = _discover_backends() File "D:\Stable_Diffusion\Sana_v15\Sana\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 "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\triton\backends\__init__.py", line 12, in _load_module spec.loader.exec_module(module) File "D:\Stable_Diffusion\Sana_v15\Sana\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: A dynamic link library (DLL) initialization routine failed. The above exception was the direct cause of the following exception: Traceback (most recent call last): File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 942, in _get_module return importlib.import_module("." + module_name, self.__name__) File "C:\Users\ADMIN\miniconda3\lib\importlib\__init__.py", line 126, in import_module return _bootstrap._gcd_import(name[level:], package, level) File "", line 1050, in _gcd_import File "", line 1027, in _find_and_load File "", line 1006, in _find_and_load_unlocked File "", line 688, in _load_unlocked File "", line 883, in exec_module File "", line 241, in _call_with_frames_removed File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\pipelines\sana\pipeline_sana.py", line 26, in from ...loaders import SanaLoraLoaderMixin File "", line 1075, in _handle_fromlist File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 932, in __getattr__ module = self._get_module(self._class_to_module[name]) File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 944, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.loaders.lora_pipeline because of the following error (look up to see its traceback): DLL load failed while importing libtriton: A dynamic link library (DLL) initialization routine failed. The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\Users\ADMIN\miniconda3\lib\runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, File "C:\Users\ADMIN\miniconda3\lib\runpy.py", line 86, in _run_code exec(code, run_globals) File "D:\Stable_Diffusion\Sana_v15\Sana\app\secourses_diff_app.py", line 19, in from diffusers import SanaPipeline File "", line 1075, in _handle_fromlist File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 933, in __getattr__ value = getattr(module, name) File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 933, in __getattr__ value = getattr(module, name) File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 932, in __getattr__ module = self._get_module(self._class_to_module[name]) File "D:\Stable_Diffusion\Sana_v15\Sana\venv\lib\site-packages\diffusers\utils\import_utils.py", line 944, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.pipelines.sana.pipeline_sana because of the following error (look up to see its traceback): Failed to import diffusers.loaders.lora_pipeline because of the following error (look up to see its traceback): DLL load failed while importing libtriton: A dynamic link library (DLL) initialization routine failed.
Luong Huynh
2025-01-14 12:03:36 +0000 UTCnice. triton is usually not mandatory :D
Furkan Gözükara
2025-01-13 20:49:26 +0000 UTCYes I do have python installed. the strange thing it is working perfectly and it is really fast 😊
Tobe2d
2025-01-13 20:48:49 +0000 UTCdo you have python 3.10? it looks like you don't have because my installer installs pre compiled triton 3 based on python 3.10. make a fresh install and try again
Furkan Gözükara
2025-01-13 19:05:07 +0000 UTCHi, as always thank you so much for the great work. When I installed it with no errors and start Windows_Start.bat I get the below log. It works and its really fast, just want to know what is this error? G:\Nvidia_Sana>cd Sana G:\Nvidia_Sana\Sana>call venv\Scripts\activate.bat A matching Triton is not available, some optimizations will not be enabled Traceback (most recent call last): File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\xformers\__init__.py", line 57, in _is_triton_available import triton # noqa File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\__init__.py", line 20, in from .runtime import ( File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\runtime\__init__.py", line 1, in from .autotuner import (Autotuner, Config, Heuristics, autotune, heuristics) File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\runtime\autotuner.py", line 9, in from ..testing import do_bench, do_bench_cudagraph File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\testing.py", line 7, in from . import language as tl File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\language\__init__.py", line 4, in from . import math File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\language\math.py", line 1, in from . import core File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\language\core.py", line 10, in from ..runtime.jit import jit File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\runtime\jit.py", line 12, in from ..runtime.driver import driver File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\runtime\driver.py", line 1, in from ..backends import backends File "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\backends\__init__.py", line 50, in backends = _discover_backends() File "G:\Nvidia_Sana\Sana\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 "G:\Nvidia_Sana\Sana\venv\lib\site-packages\triton\backends\__init__.py", line 12, in _load_module spec.loader.exec_module(module) File "G:\Nvidia_Sana\Sana\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: A dynamic link library (DLL) initialization routine failed. * Running on local URL: http://127.0.0.1:7860 To create a public link, set `share=True` in `launch()`.
Tobe2d
2025-01-13 15:48:58 +0000 UTCawesome glad you solved. also int4 model coming soon hopefully i contacted developers for lower VRAM and they told me int4 will be published. i expect 2k to work at 8 gb gpus very well
Furkan Gözükara
2025-01-06 09:41:15 +0000 UTCGood news! I updated the NVidia and looked in task manager and now 16 GB of my system RAM is shared with the GPU, and it works! Thanks for the help!
Nibmeister
2025-01-05 18:26:46 +0000 UTCI also tried out Sana in ComfyUI. It ran thru the sampler just fine, but when it hit the VAE, it OOMed on me.
Nibmeister
2025-01-05 16:36:38 +0000 UTCawesome
Furkan Gözükara
2025-01-05 15:08:41 +0000 UTCit worked after the clean installation
Umut Hasanoglu
2025-01-05 14:22:01 +0000 UTCawesome
Furkan Gözükara
2025-01-05 14:08:13 +0000 UTCinstall error. reinstall and send me logs so we can see error reason : monstermmorpg@gmail.com
Furkan Gözükara
2025-01-05 13:37:14 +0000 UTCTraceback (most recent call last): File "D:\Sanav9\Sana\venv\lib\site-packages\mmcv\utils\registry.py", line 69, in build_from_cfg return obj_cls(**args) File "D:\Sanav9\Sana\diffusion\model\nets\sana_multi_scale.py", line 412, in SanaMS_1600M_P1_D20 return SanaMS(depth=20, hidden_size=2240, patch_size=1, num_heads=20, **kwargs) File "D:\Sanav9\Sana\diffusion\model\nets\sana_multi_scale.py", line 208, in __init__ super().__init__( File "D:\Sanav9\Sana\diffusion\model\nets\sana.py", line 261, in __init__ self.initialize_weights() File "D:\Sanav9\Sana\diffusion\model\nets\sana.py", line 361, in initialize_weights self.pos_embed.data.copy_(torch.from_numpy(pos_embed).float().unsqueeze(0)) RuntimeError: Numpy is not available During handling of the above exception, another exception occurred: Traceback (most recent call last): File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\queueing.py", line 625, in process_events response = await route_utils.call_process_api( File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api output = await app.get_blocks().process_api( File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\blocks.py", line 2047, in process_api result = await self.call_function( File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\blocks.py", line 1606, in call_function prediction = await utils.async_iteration(iterator) File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\utils.py", line 714, in async_iteration return await anext(iterator) File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\utils.py", line 708, in __anext__ return await anyio.to_thread.run_sync( File "D:\Sanav9\Sana\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:\Sanav9\Sana\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 2505, in run_sync_in_worker_thread return await future File "D:\Sanav9\Sana\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 1005, in run result = context.run(func, *args) File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\utils.py", line 691, in run_sync_iterator_async return next(iterator) File "D:\Sanav9\Sana\venv\lib\site-packages\gradio\utils.py", line 852, in gen_wrapper response = next(iterator) File "D:\Sanav9\Sana\app\secourses_app.py", line 321, in generate_multiple images, current_seed, speed_info = generate( File "D:\Sanav9\Sana\venv\lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "D:\Sanav9\Sana\venv\lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "D:\Sanav9\Sana\app\secourses_app.py", line 229, in generate pipe = SanaPipeline(args.config) File "D:\Sanav9\Sana\app\sana_pipeline.py", line 104, in __init__ self.model = self.build_sana_model(config).to(self.device) File "D:\Sanav9\Sana\app\sana_pipeline.py", line 155, in build_sana_model model = build_model( File "D:\Sanav9\Sana\diffusion\model\builder.py", line 35, in build_model model = MODELS.build(cfg, default_args=kwargs) File "D:\Sanav9\Sana\venv\lib\site-packages\mmcv\utils\registry.py", line 237, in build return self.build_func(*args, **kwargs, registry=self) File "D:\Sanav9\Sana\venv\lib\site-packages\mmcv\utils\registry.py", line 72, in build_from_cfg raise type(e)(f'{obj_cls.__name__}: {e}') RuntimeError: SanaMS_1600M_P1_D20: Numpy is not available
Umut Hasanoglu
2025-01-05 13:36:00 +0000 UTCthank you so much, worked
Sher Lock
2025-01-05 11:39:36 +0000 UTCyou can make a fresh nvidia driver install and reset all settings to default. it should enable it.
Furkan Gözükara
2025-01-05 00:23:14 +0000 UTCIt's not obvious to me how to set it to shared VRAM. Can you help?
Nibmeister
2025-01-05 00:05:26 +0000 UTC1K model uses 8.7 GB VRAM after first generation. it should work with shared VRAM usage. do you have it enabled? i will check if we can use FP8
Furkan Gözükara
2025-01-04 23:58:35 +0000 UTCI installed it on my laptop with 3070TI 8 GB VRAM and though it installed without a problem, ran out of memory trying to make a 1k image. Any way to tweak it to work on this GPU?
Nibmeister
2025-01-04 23:54:53 +0000 UTCi agree. it is nvidia's fault :D
Furkan Gözükara
2025-01-04 23:43:33 +0000 UTCThanks. I don't know why they do that since .safetensors has been made a safe standard quite a long time ago, probably since SDXL came out. Maybe there's a way online or somewhere to check the files for any pickle stuff. Ill try searching just in case there's a risk. There might not be a risk if they're from the official place but it's hard to tell.
cool1
2025-01-04 23:42:18 +0000 UTCsorry for late reply. edit the Windows_Start.bat file and remove set HF_HUB_ENABLE_HF_TRANSFER=1 normally this speeds up download significantly but some people having problem randomly
Furkan Gözükara
2025-01-04 23:22:18 +0000 UTCwell the models are from nvidia itself and sadly their pipe working with those
Furkan Gözükara
2025-01-04 23:21:36 +0000 UTCstucked in that for hours , any solution ? F:\Sanav9>cd Sana F:\Sanav9\Sana>call venv\Scripts\activate.bat * Running on local URL: http://127.0.0.1:7860 To create a public link, set `share=True` in `launch()`. 2025-01-05 02:21:34 - [Sana] - INFO - Sampler flow_dpm-solver, flow_shift: 3.0 2025-01-05 02:21:34 - [Sana] - INFO - Inference with torch.float16, PAG guidance layer: [8] Initial memory state: GPU Memory used: 0.00MB (Reserved: 0.00MB) Initial memory state: RAM Memory used: 616.58MB [DC-AE] Loading model from mit-han-lab/dc-ae-f32c32-sana-1.0 model.safetensors: 1%|▍ | 10.5M/1.25G [00:27<53:56, 383kB/s]
Sher Lock
2025-01-04 18:38:02 +0000 UTCThe installation downloads Sana_1600_1024px.pth - perplexiy.ai says .pth can be unsafe and that "This risk is similar to that of executable files like .exe". Is it not possible for it to use .safetensors files which are safe and don't have that risk?
cool1
2025-01-04 18:24:22 +0000 UTCi see. well FLUX is the king. I would say sana 2k and SD 3.5 are close. Sana 2k is able to generate 4 mega pixel images really fast. I made a tutorial and editing it. best way is comparing each model for your case and decide which one fitting most
Furkan Gözükara
2024-12-30 21:10:19 +0000 UTCI guess he meant what the took actually does. Like why SANA instead of FLUX or SD? What is the differences? What are the strengths and weaknesses of each:)
Mehmet Atakan Çavuşlu
2024-12-30 20:43:43 +0000 UTCnot that level but fast for 4 mega pixel
Furkan Gözükara
2024-12-29 19:06:06 +0000 UTCHow is the model quality? Is it as good as Flux?
Umut Hasanoglu
2024-12-29 17:31:13 +0000 UTCthanks for info.
Furkan Gözükara
2024-12-28 19:25:14 +0000 UTChello. you have to follow requirements video exactly. this is just once. i see you didnt follow since your python installed into AppData https://youtu.be/DrhUHnYfwC0
Furkan Gözükara
2024-12-28 19:24:17 +0000 UTCSame as Trellis_V4 this installer not working for me i will check another way : Cloning into 'Sana'... remote: Enumerating objects: 657, done. remote: Counting objects: 100% (216/216), done. remote: Compressing objects: 100% (118/118), done. remote: Total 657 (delta 137), reused 106 (delta 98), pack-reused 441 (from 1) Receiving objects: 100% (657/657), 5.46 MiB | 19.76 MiB/s, done. Resolving deltas: 100% (289/289), done. Python launcher is available. Generating Python 3.10 VENV Traceback (most recent call last): File "Z:\NVIDIA_Sana_v8\Download_Models.py", line 3, in from huggingface_hub import snapshot_download ModuleNotFoundError: No module named 'huggingface_hub' Virtual environment made and installed properly Appuyez sur une touche pour continuer... Z:\NVIDIA_Sana_v8>cd Sana Z:\NVIDIA_Sana_v8\Sana>call venv\Scripts\activate.bat Traceback (most recent call last): File "C:\Users\Arckheens\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, File "C:\Users\Arckheens\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 86, in _run_code exec(code, run_globals) File "Z:\NVIDIA_Sana_v8\Sana\app\secourses_app.py", line 13, in import gradio as gr ModuleNotFoundError: No module named 'gradio' Appuyez sur une touche pour continuer...
MEGA AMIGA
2024-12-28 02:48:55 +0000 UTCSome of those might be in the Kohya SS help or the help for the image generator you're using. You could also ask an AI (eg. ChatGPT, perplexity.ai or a local AI). Here's some info on those based on asking perplexity.AI BF16 (Brain Float 16 - I think they say it's called that because of Google Brain) is a 16-bit floating-point format with 8 bits for the exponent and 8 bits for the significand. "It offers better stability during training than FP16 and can represent a much larger numerical range". I'm not sure how accurate that is though. But it's what it uses in Kohya training for one of the high quality presets and I think it might work faster on certain GPUS. I might be wrong. FP16 is a 16-bit floating-point format with 5 bits for the exponent and 11 bits for the significand. It can represent numbers between -65K and +65K. Float (FP32) Float refers to 32-bit floating-point precision. It's the standard precision used in most computations and offers higher accuracy than FP16 or BF16 CFG Scale (Classifier-Free Guidance Scale) "CFG Scale is used for classifier-free guidance, which affects the influence of the negative prompt during image generation. A higher CFG scale typically results in stronger adherence to the prompt" (I think it means positive prompt here). Though when asked, perplexity also agrees it affects the positive prompt "The CFG scale primarily controls how closely the image generation process adheres to the positive prompt". And it says "Impact on Negative Prompts. The CFG scale also influences the effect of negative prompts, but in a less intuitive way: At lower CFG values, the negative prompt has a stronger influence on the generated image. As the CFG scale increases, the impact of the negative prompt decreases" VAE (Variational Autoencoder) VAE is a neural network architecture used in Stable Diffusion models for encoding and decoding images. It's crucial for converting between image and latent space representations.
cool1
2024-12-27 21:53:25 +0000 UTCV8 zip file working on kaggle very fast. 2k model not working though
Furkan Gözükara
2024-12-27 13:57:16 +0000 UTCstill not working let me update a fix will reply you
Furkan Gözükara
2024-12-27 13:08:30 +0000 UTCUh. Sorry! I used v5. Will try v7
Philipp Capetian
2024-12-27 12:41:55 +0000 UTChi your error is requirements please follow this video every step : https://youtu.be/DrhUHnYfwC0
Furkan Gözükara
2024-12-27 12:40:34 +0000 UTChi fixed with v7 zip file
Furkan Gözükara
2024-12-27 12:40:01 +0000 UTChi fixed with v7 zip file
Furkan Gözükara
2024-12-27 12:39:54 +0000 UTChi fixed with v7 zip file
Furkan Gözükara
2024-12-27 12:39:52 +0000 UTCwith V7 updated the kaggle notebook. i am testing atm
Furkan Gözükara
2024-12-27 12:39:19 +0000 UTCHello Mehmet Welcome. these topics are really deep topics. So I don't plan to cover them atm sadly :( But as you follow tutorials historically, you will get used to and start understanding more each term. also you can read this excellent article : https://stable-diffusion-art.com/how-stable-diffusion-work/ still i am trying to show all the relevant information in tutorials that will help you in your generations and trainings
Furkan Gözükara
2024-12-27 12:36:22 +0000 UTCHello Furkan, I want to learn more about these AI stuff. I can perfectly follow and really like your tutorials. But I still lack info on basic stuff like what is VAE what is fp8 or fp16 etc. What is cfg scale etc. Do you plan to make a tutorial on these parameters especially on sd and flux? So we can fine tune our models better? I feel like i am blindly following your instructions and not knowing what button does what? Hope you understand what I meant, I really like your guides. Huge thanks!
Mehmet Atakan Çavuşlu
2024-12-27 12:16:06 +0000 UTCI install perfect, but when trying to start i got the same error msg, no cuda gpu "No CUDA runtime is found, using CUDA_HOME='C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4' ..not sure, but could be i dont have the folder "c:\Program Files\ i have the german win11 it is c:\Programme\
Nedo Braun
2024-12-27 10:18:54 +0000 UTCThe kaggle notebook gives out access errors while loading: Cannot access gated repo for url https://huggingface.co/google/shieldgemma-2b/resolve/main/config.json. Access to model google/shieldgemma-2b is restricted. You must have access to it and be authenticated to access it. Please log in.
Philipp Capetian
2024-12-27 10:15:40 +0000 UTCSame for me, RTX3070, CUDA 12.4 and Python 3.10.11 also
Lee
2024-12-27 10:11:52 +0000 UTCYour latest "Sana_v5.zip" one-click installer is throwing an error: RuntimeError: No CUDA GPUs are available while I have an RTX 4090, CUDA 12.4, and Python 3.10.11. Please check this error and provide a solution.
Anshul Gupta
2024-12-27 09:14:32 +0000 UTCHi they are posted here with links description and titles : https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Patreon-Posts-Index.md#patreon-exclusive-content sorted by last updated / published date
Furkan Gözükara
2024-12-17 21:44:21 +0000 UTCI would love to see with every new one-click installer post, also a short description what the tool does and also what it doesn't do. I have basically zero information about SANA but I expect it to be important if you have created installers for it. Same would be true for every other app you have prepared. And thanks a lot for your hard works on these!
Pavel Desort
2024-12-17 17:38:24 +0000 UTCthanks a lot
Furkan Gözükara
2024-11-30 21:52:16 +0000 UTCYou, sir, are awesome! Thank you!
Victor Domingos Alves Tarzariol
2024-11-30 20:06:19 +0000 UTCyep they will do for some scripts. but you can edit and look each file content they are all py files - by the way i use the very best kaspersky internet security no detection
Furkan Gözükara
2024-11-29 13:44:42 +0000 UTCnope not yet. i wouldnt enter that area until one of the major scripts adds support
Furkan Gözükara
2024-11-29 13:44:22 +0000 UTCIs there a tutorial for fine-tuning it with our dataset please ?
Chahrazad Seh
2024-11-29 10:44:46 +0000 UTCWindows defender is flagging this as a virus (Trojan:Script/Wacatac.B!ml) https://www.microsoft.com/en-us/wdsi/threats/malware-encyclopedia-description?name=Trojan%3AScript%2FWacatac.B!ml&threatid=2147735503
Glen Carpenter
2024-11-29 01:21:17 +0000 UTChi message me install logs from discord so i can see
Furkan Gözükara
2024-11-24 12:14:48 +0000 UTCfixed. if such thing happens always look for attachments
Furkan Gözükara
2024-11-24 12:14:31 +0000 UTChi，i have done all of the video. but still the same issue as this.
big dato
2024-11-24 12:12:18 +0000 UTCClicking on the zip is returning an "expired URL" error
Marc Mercuri
2024-11-24 02:04:45 +0000 UTChello. you have to install requirements as shown in this video : https://youtu.be/DrhUHnYfwC0
Furkan Gözükara
2024-11-23 22:59:26 +0000 UTCi didnt try but currently i am using cuda 12.4 as shown in this tutorial : https://youtu.be/DrhUHnYfwC0
Furkan Gözükara
2024-11-23 22:58:58 +0000 UTCi dont know but you can modify app code and try. we have --share it works gradio live share
Furkan Gözükara
2024-11-23 22:58:16 +0000 UTCAny way to run this on local network with --listen flag?
Sushant Bhosale
2024-11-23 21:39:29 +0000 UTCI was using the V2.zip, everything downloaded and installed. I ran the install.bat -> update.bat -> start.bat and when trying to generate an image with default settings it said: ....File "F:\AI\Sana_v2\Sana\venv\lib\site-packages\triton\runtime\build.py", line 53, in _build raise RuntimeError("Failed to find C compiler. Please specify via CC environment variable.") RuntimeError: Failed to find C compiler. Please specify via CC environment variable. When starting it also said: WARNING: Failed to find MSVC.
Lena Davis
2024-11-23 08:27:45 +0000 UTCthis does not work with cuda 11.8?
Hey Ooo
2024-11-23 04:42:38 +0000 UTCnope i don't think there is any UI not even ComfyUI supporting it
Furkan Gözükara
2024-11-23 00:38:20 +0000 UTCcan we download it in swarmui or webui ?
Nabil Boulezaz
2024-11-22 23:53:23 +0000 UTCyes but i don't know how to yet. they have instructions on github
Furkan Gözükara
2024-11-22 21:12:39 +0000 UTCtrue. well this is a nice addition though to our scripts arsenal for who wants to test
Furkan Gözükara
2024-11-22 21:12:27 +0000 UTC:D lets see if community will love it
Furkan Gözükara
2024-11-22 21:11:36 +0000 UTCis it trainable model?
Sipriyani
2024-11-22 20:53:17 +0000 UTCIt looks good, though it says it compresses images more so maybe it could lose some detail on compressing them. Though it says it can do 4K resolution a lot faster than other models I think. One thing about it is it seems to be non-commercial use only, so if so it's worse than Flux Schnell and SD3.5 Large if you want to earn money from it. So even if people install Sana and all other things for it, they might get great images out of it but never earn anything with it which really restricts its usefulness. Maybe we could have trainers fro SD3.5 Large too.
cool1
2024-11-22 20:52:01 +0000 UTCOne thing for sure, it's VERY sharp. Next is to try and fine-tune/train it :p
BecauseReasons
2024-11-22 20:24:57 +0000 UTC
