54 lines
2.2 KiB
Plaintext
54 lines
2.2 KiB
Plaintext
|
Just place the Mordant-Prompt-Enhancer file in your ComfyUI custom_nodes folder.
|
|||
|
|
You can see in the example workflow how to connect everything. And the default settings should be good, and you can experiment after.
|
|||
|
|
The node looks for the models in the /models/text_encoders forlders, so place the mordant gguf there.
|
|||
|
|
|
|||
|
|
What does the mode do?
|
|||
|
|
|
|||
|
|
Enhancer is your daily prompt enhancer mode, general is more of a llm.
|
|||
|
|
|
|||
|
|
enable_enhancement toggle, basically turns the node on and off, for when you want to just use your own un-altered prompt.
|
|||
|
|
|
|||
|
|
Everything else is fairly obvious, if not, you should just leave it as is. Keep_loaded is if you have enough vram to keep the model loaded after generation, otherwise, by default the node unloads it.
|
|||
|
|
|
|||
|
|
|
|||
|
|
**Installation**
|
|||
|
|
Place the node file in ComfyUI's `custom_nodes/` folder. The only additional dependency is `llama-cpp-python`.
|
|||
|
|
|
|||
|
|
**CPU‑only** (no GPU offloading):
|
|||
|
|
```bat
|
|||
|
|
# Windows portable
|
|||
|
|
python_embeded\python.exe -m pip install llama-cpp-python
|
|||
|
|
```
|
|||
|
|
```bash
|
|||
|
|
# Linux / venv / Mac
|
|||
|
|
pip install llama-cpp-python
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
**GPU / CUDA — recommended** (pre‑built wheels, no compiler required):
|
|||
|
|
```bat
|
|||
|
|
# Windows portable — replace cu121 with your CUDA version
|
|||
|
|
python_embeded\python.exe -m pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121
|
|||
|
|
```
|
|||
|
|
```bash
|
|||
|
|
# Linux / venv — replace cu121 with your CUDA version
|
|||
|
|
pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121
|
|||
|
|
```
|
|||
|
|
> Supported versions: `cu121` · `cu122` · `cu124` · `cu125` — run `nvidia-smi` to check yours (top‑right of output).
|
|||
|
|
|
|||
|
|
<details>
|
|||
|
|
<summary>Build from source (advanced — only if wheels fail)</summary>
|
|||
|
|
|
|||
|
|
Requires CUDA Toolkit, CMake, and C++ build tools (Visual Studio on Windows).
|
|||
|
|
|
|||
|
|
```bat
|
|||
|
|
# Windows
|
|||
|
|
set CMAKE_ARGS=-DGGML_CUDA=on
|
|||
|
|
python_embeded\python.exe -m pip install llama-cpp-python --force-reinstall --no-cache-dir
|
|||
|
|
```
|
|||
|
|
```bash
|
|||
|
|
# Linux
|
|||
|
|
CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python --force-reinstall --no-cache-dir
|
|||
|
|
```
|
|||
|
|
</details>
|
|||
|
|
|
|||
|
|
After restarting ComfyUI, the node appears under `text/llm` as **Mordant Prompt Enhancer**. Just pick your model and start composing.
|