Can't import SageAttention: No module named 'triton' is:issue

[Xxredz]

I keep receiving this error when I try to generate an image. How can I fix this?

If anyone could provide easy to understand steps, I would really appreciate it.

Just so you know, I click python.exe in my ComfyUI_windows_portable > python_embeded folder and it detects Python 3.12.7

Error:

FETCH DATA from: C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager-main\extension-node-map.json [DONE]
got prompt
Loading text encoder model (clipL) from: C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\models\clip\clip-vit-large-patch14
Text encoder to dtype: torch.float16
Loading tokenizer (clipL) from: C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\models\clip\clip-vit-large-patch14
Loading text encoder model (llm) from: C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\models\LLM\llava-llama-3-8b-text-encoder-tokenizer
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 4/4 [00:10<00:00, 2.64s/it]
Text encoder to dtype: torch.float16
Loading tokenizer (llm) from: C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\models\LLM\llava-llama-3-8b-text-encoder-tokenizer
llm prompt attention_mask shape: torch.Size([1, 161]), masked tokens: 24
clipL prompt attention_mask shape: torch.Size([1, 77]), masked tokens: 25
!!! Exception during processing !!! Can't import SageAttention: No module named 'triton'
Traceback (most recent call last):
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-HunyuanVideoWrapper\nodes.py", line 150, in loadmodel
from sageattention import sageattn_varlen
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\sageattention_init_.py", line 1, in
from .core import sageattn, sageattn_varlen
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\sageattention\core.py", line 2, in
import triton
ModuleNotFoundError: No module named 'triton'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 323, in execute
output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 198, in get_output_data
return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 169, in _map_node_over_list
process_inputs(input_dict, i)
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 158, in process_inputs
results.append(getattr(obj, func)(**inputs))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Nebula PC\Desktop\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-HunyuanVideoWrapper\nodes.py", line 152, in loadmodel
raise ValueError(f"Can't import SageAttention: {str(e)}")
ValueError: Can't import SageAttention: No module named 'triton'

[tavyscrolls]

If using linux: pip install triton
Since you're on windows, you'll have to find a fork that supports windows

[scottmudge]

People just need to sticky this:

https://github.com/woct0rdho/triton-windows/releases

Also, this question belongs on the ComfyUI-HunyuanVideoWrapper project issue tracker, not this one. I don't think this official repo even uses SageAttention.

[tommywright]

I have the same as OP. I have Triton installed and Sageattention. I used Triton 3.1.0-cp312 as I'm using Python 3.12. I can go into Python and do an 'import sageattention' or 'import triton' and see no errors. I'll go back and see if HunyuanVideo is looking for a particular Triton I guess.

[tommywright]

I ended up following the instructions very carefully here on the triton-windows page (I didn't realize triton is really meant for Linux and the Windows version is kinda a hack). I'm using an installed Cuda so there are specific instructions for that also. Just read carefully and it will work. https://github.com/woct0rdho/triton-windows?tab=readme-ov-file