Balledk commited on
Commit
574ab91
·
verified ·
1 Parent(s): be7bc73

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +54 -0
README.md ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: creativeml-openrail-m
3
+ base_model: black-forest-labs/FLUX.1-dev
4
+ tags:
5
+ - lora
6
+ - flux
7
+ - diffusers
8
+ - ai-toolkit
9
+ - z-image-turbo
10
+ datasets:
11
+ - cowl1
12
+ - cowl2
13
+ instance_prompt: cowl neck
14
+ ---
15
+
16
+ # Model Card: Flux LoRA - cowl (Z-Image De-Turbo)
17
+
18
+ This is a LoRA model trained on a curated dataset of high-quality images (1024x1024) using the [Ostris - AI Toolkit](https://github.com/ostris/ai-toolkit).
19
+
20
+ **Architecture:** Specifically optimized for **Z-Image De-Turbo (De-Distilled)**.
21
+
22
+ ## Training Settings
23
+
24
+ | Parameter | Value |
25
+ |---|---|
26
+ | **Trigger Word** | `cowl neck` |
27
+ | **Model Architecture** | Z-Image De-Turbo (De-Distilled) |
28
+ | **Batch Size** | 2 |
29
+ | **Rank (Dimension)** | 48 |
30
+ | **Precision** | float8 (Transformer & Text Encoder) |
31
+ | **Save Frequency** | Every 200 steps |
32
+ | **Total Training Steps** | 4000 |
33
+
34
+ ## Prompting Strategy & "Mix & Match"
35
+ This LoRA was trained using **granular, tag-based descriptive captions**. This approach "de-couples" specific attributes, allowing you to freely combine features from across the dataset.
36
+
37
+ * **Modular Control:** You are not limited to the training images. You can mix attributes—for example, taking a `silk` texture from one style, a `deep drape` from another, and setting the color to `emerald green`.
38
+ * **Tag-Based Precision:** Since every detail was tagged, the model performs best when you describe the specific materials (satin, jersey, wool), finishes, and colors you want to see.
39
+
40
+ **Example Prompt:**
41
+ > `cowl neck, emerald green color, satin material, deep draped folds, elegant aesthetic, studio lighting`
42
+
43
+ ## Usage
44
+ - **Inference:** Use settings compatible with Z-Image Turbo/De-Turbo.
45
+ - **LoRA Strength:** 0.6 - 1.0 (Start at 0.8).
46
+ - **Precision:** Trained in **float8**. Ensure your environment (ComfyUI, Diffusers, etc.) supports this for optimal results.
47
+
48
+ ## Training Infrastructure
49
+ - **Toolkit:** Ostris AI Toolkit
50
+ - **Dataset Size:** [Indsæt antal] images (1024x1024)
51
+ - **Training Method:** LoRA
52
+
53
+ ---
54
+ *Model card generated for the cowl series.*