Update README.md

README.md

@@ -10,6 +10,38 @@ tags:
 - realistic
 - lora
 ---
+# Lenovo ChromaRadiance
+
+## 🛠️ Usage Instructions
+
+To use LoRAs with Chroma Radiance, you need to apply a few changes to your ComfyUI files using the modified files provided in this repository:
+
+1. **Replace `lora.py` (Main):**
+   Take the file from the `comfy` folder in this repo and replace the file at:
+   `ComfyUI/comfy/lora.py`
+
+2. **Replace `lora.py` (Weight Adapter):**
+   Take the file from the `weight_adapter` folder in this repo and replace the file at:
+   `ComfyUI/comfy/weight_adapter/lora.py`
+
+### Workflow & Notes
+Please use the workflow provided in this repository for the best results.
+
+> **Observation:** Based on my personal testing, GGUF generation takes a bit longer, but the quality appears to be slightly better. However, it is hard to say for sure as I haven't had enough time for extensive testing yet.
+
+### 💡 Generation Tips
+
+1. **Samplers & Settings:**
+   * **Best Quality:** I achieve the best results using `fully_implicit` samplers (such as `radau_iia_2s` and `gauss-legendre_2s`). For these, **20-30 steps** are sufficient, and you should keep **bong_math** disabled.
+   * **Alternative (Faster):** `dpmpp3m` + `beta` also works quite well with **50 steps** and **bong_math** enabled. It generates faster, but the output tends to be slightly noisier.
+
+2. **LoRA Strength:**
+   * Recommended strength: **0.85 - 1.0**.
+
+3. **Combinations:**
+   * This model can be combined with my **NiceGirls** LoRA.
+   * *Suggested weights:* NiceGirls at **0.6** + Lenovo at **0.8**.
+---
 
 ## Example Generations