Add instructions for installing in ComfyUI Portable

[GalaxyTimeMachine]

Example:

For those running ComfyUI Portable:

[You may need to install Visual Code, with the desktop c++ tools first]

Run these commands from within your portable folder:

.\python_embeded\python.exe -m pip install -U torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

.\python_embeded\python.exe -m pip install -U packaging wheel ninja setuptools
 
git clone https://github.com/Lightricks/LTX-Video-Q8-Kernels.git 

cd .\LTX-Video-Q8-Kernels\

..\python_embeded\python.exe setup.py install

---

[phr00t]

These are the steps I took, which led to this error: #2

[LIQUIDMIND111]

it worked BETTER but now i get this error:

cannot access local variable 'self_attn_func' where it is not associated with a value

[LIQUIDMIND111]

!!! Exception during processing !!! cannot access local variable 'self_attn_func' where it is not associated with a value
Traceback (most recent call last):
File "E:\ComfyUI_NVME\ComfyUI_windows_portable\ComfyUI\execution.py", line 347, 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 "E:\ComfyUI_NVME\ComfyUI_windows_portable\ComfyUI\execution.py", line 222, 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 "E:\ComfyUI_NVME\ComfyUI_windows_portable\ComfyUI\execution.py", line 194, in _map_node_over_list
process_inputs(input_dict, i)
File "E:\ComfyUI_NVME\ComfyUI_windows_portable\ComfyUI\execution.py", line 183, in process_inputs
results.append(getattr(obj, func)(**inputs))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_NVME\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-LTXVideo\q8_nodes.py", line 52, in patch
patcher(transformer, use_fp8_attention)
File "E:\ComfyUI_NVME\ComfyUI_windows_portable\python_embeded\Lib\site-packages\q8_kernels\integration\patch_transformer.py", line 24, in patch_comfyui_native_transformer
attn_forward = cross_attn_forward(use_fp8_attention)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_NVME\ComfyUI_windows_portable\python_embeded\Lib\site-packages\q8_kernels\integration\comfyui_native.py", line 71, in cross_attn_forward
self_attn, cross_attn, is_out_tuple = get_attention_func(use_fp8_attention)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_NVME\ComfyUI_windows_portable\python_embeded\Lib\site-packages\q8_kernels\integration\utils.py", line 104, in get_attention_func
(self_attn_func, self_attn_memory_layout),
^^^^^^^^^^^^^^
UnboundLocalError: cannot access local variable 'self_attn_func' where it is not associated with a value

[LIQUIDMIND111]

I'm sorry to hear you're encountering this error. The message:

pgsql
Copy
Edit
UnboundLocalError: cannot access local variable 'self_attn_func' where it is not associated with a value
indicates that within the get_attention_func function in the utils.py file of the q8_kernels package, the variable self_attn_func is being referenced before it has been assigned a value. This typically happens when a variable is used inside a function before it's defined, often due to missing conditions or incorrect control flow.

Given that you're using the latest versions of both PyTorch (2.6.0+cu126) and CUDA (12.6), it's possible that the LTXVideo-Q8-Kernels package hasn't been fully tested or updated to ensure compatibility with these versions. Newer versions of PyTorch and CUDA can introduce changes that affect how certain operations or functions behave, potentially leading to such errors.

Recommended Steps:

Check for Updates: Visit the LTXVideo-Q8-Kernels GitHub repository to see if there have been any recent commits or releases that address compatibility with PyTorch 2.6.0 and CUDA 12.6.

Review Open Issues: Look through the Issues section of the repository to see if other users have reported similar problems. There might be ongoing discussions or suggested workarounds.

Consider Downgrading: If no updates or solutions are available, you might consider downgrading to a more stable and widely supported version of PyTorch and CUDA. For instance, PyTorch 2.2.2 with CUDA 12.1 (cu121) has been noted for its stability and compatibility with various extensions. To do this:

Uninstall the current versions:

bash
Copy
Edit
E:\ComfyUI_NVME\ComfyUI_windows_portable\python_embeded\python.exe -m pip uninstall torch torchvision torchaudio -y
Install the recommended versions:

bash
Copy
Edit
E:\ComfyUI_NVME\ComfyUI_windows_portable\python_embeded\python.exe -m pip install torch==2.2.2 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
Reinstall the LTXVideo-Q8-Kernels package following the provided instructions.

Contact Maintainers: If the issue persists, consider opening a new issue on the LTXVideo-Q8-Kernels GitHub repository with detailed information about your setup and the error message. The maintainers might provide a fix or further guidance.

Please ensure you back up your current environment before making significant changes, such as downgrading packages, to prevent any unintended disruptions.