Access is denied

[AndreySemjonov]

I get such error when try to generate image ...
Can't understand where problem is.

--- Starting Generation ---
Model: full, Res: 1024x1024, Steps: 50, CFG: 5.0, Shift: 3.0, Seed: 251542427621138
Using standard sequence lengths: CLIP-L: 77, OpenCLIP: 150, T5: 256, Llama: 256
Ensuring pipe on: cuda:0 (Offload NOT enabled)
Executing pipeline inference...
torch.nn.functional.scaled_dot_product_attention does not support output_attentions=True. Falling back to eager attention. This warning can be removed using the argument attn_implementation="eager" when loading the model.
0%| | 0/50 [00:00<?, ?it/s]using SageAttention
0%| | 0/50 [00:01<?, ?it/s]
!!! ERROR during execution: [WinError 5] Access is denied
Traceback (most recent call last):
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hidreamsampler.py", line 679, in generate
pipeline_output = pipe(
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\utils_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\pipelines\hidream_image\pipeline_hidream_image.py", line 755, in call
noise_pred = self.transformer(
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1739, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1750, in _call_impl
return forward_call(*args, **kwargs)
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\models\transformers\transformer_hidream_image.py", line 460, in forward
hidden_states, initial_encoder_hidden_states = block(
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1739, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1750, in _call_impl
return forward_call(*args, **kwargs)
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\models\transformers\transformer_hidream_image.py", line 222, in forward
return self.block(
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1739, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1750, in _call_impl
return forward_call(*args, **kwargs)
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\models\transformers\transformer_hidream_image.py", line 168, in forward
attn_output_i, attn_output_t = self.attn1(
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1739, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\module.py", line 1750, in _call_impl
return forward_call(*args, **kwargs)
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\models\attention.py", line 89, in forward
return self.processor(
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\models\attention_processor.py", line 128, in call
hidden_states = attention(query, key, value)
File "D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hi_diffusers\models\attention_processor.py", line 45, in attention
hidden_states = sageattn(query, key, value, tensor_layout="NHD", is_causal=False)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\sageattention\core.py", line 105, in sageattn
q_int8, q_scale, k_int8, k_scale = per_block_int8(q, k, sm_scale=sm_scale, tensor_layout=tensor_layout)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\sageattention\quant_per_block.py", line 63, in per_block_int8
quant_per_block_int8_kernel[grid](
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\runtime\jit.py", line 345, in
return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\runtime\jit.py", line 607, in run
device = driver.active.get_current_device()
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\runtime\driver.py", line 23, in getattr
self._initialize_obj()
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\runtime\driver.py", line 20, in _initialize_obj
self._obj = self._init_fn()
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\runtime\driver.py", line 9, in _create_driver
return actives0
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\backends\nvidia\driver.py", line 412, in init
self.utils = CudaUtils() # TODO: make static
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\backends\nvidia\driver.py", line 90, in init
mod = compile_module_from_src(Path(os.path.join(dirname, "driver.c")).read_text(), "cuda_utils")
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\backends\nvidia\driver.py", line 67, in compile_module_from_src
so = _build(name, src_path, tmpdir, library_dirs(), include_dir, libraries)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\site-packages\triton\runtime\build.py", line 79, in _build
ret = subprocess.check_call(cc_cmd)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 364, in check_call
retcode = call(*popenargs, **kwargs)
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 345, in call
with Popen(*popenargs, **kwargs) as p:
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 966, in init
self._execute_child(args, executable, preexec_fn, close_fds,
File "C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 1435, in _execute_child
hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
PermissionError: [WinError 5] Access is denied

[ansorensen]

Not 100% sure, just starting to learn how this all works myself. It looks like the call starts with:

"D:\stablediffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_HiDream-Sampler\hidreamsampler.py"

And crashes when it tries to access your user-installed python packages.

"C:\Users\andre\AppData\Local\Programs\Python\Python310\lib\subprocess.py"

I think ComfyUI may have its own python virtual environment (it would be in ComfyUI.venv I think?) that does not have access to your User-installed python (in Users\andre\AppData...).

However, it's possible in the HiDream installation to install HiDream dependencies in your User python, instead of the special ComfyUI virtual environment.

What I would try is something like:

• Uninstall HiDream-Sampler (delete custom-modules folder)
• redownload HiDream from github
• make sure you're use the ComfyUI virtual environment
• do the pip install again (within the ComfyUI virtual environment)
• install the triton wheel
• install the weights again

Making sure your using the right virtual environment in Git Bash is something like

$source <path to>/ComfUI/.venv/scripts/activate

If you're using GitBash, you should see some indication that you're in a virtual environment after you do that.

[ansorensen]

I ended up figuring out my access denied and missing module errors, detailing it in: #119

If you have a windows portable version (i.e. a ComfyUI that uses a python virtual environment instead of the globally installed python), then you're probably running into those issues. I used the windows installer for ComfyUI and found it had these issues.

I had to make sure all the python commands (including pip) were run from the virtual python (/ComfyUI/.venv/Scripts/python) and update Pip using that python so it was pointed at the virtual python environment, and not my global installation. Details in the ticket linked above.