Update README.md
Browse files
README.md
CHANGED
|
@@ -78,7 +78,21 @@ Use `GHIBSKY style` to invoke the model’s unique aesthetic. It’s best to sta
|
|
| 78 |
|
| 79 |
If you are getting too realistic images try adding `painting` to your prompt, for example: `GHIBSKY style painting`
|
| 80 |
|
| 81 |
-
## Training Details
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
|
| 83 |
- **Trained Using**: [ostris/flux-dev-lora-trainer](https://replicate.com/ostris/flux-dev-lora-trainer/train)
|
| 84 |
- **Number of Images**: 35
|
|
|
|
| 78 |
|
| 79 |
If you are getting too realistic images try adding `painting` to your prompt, for example: `GHIBSKY style painting`
|
| 80 |
|
| 81 |
+
## Training Details (14.11.2024)
|
| 82 |
+
|
| 83 |
+
- **Trained Using**: [ostris/flux-dev-lora-trainer](https://replicate.com/ostris/flux-dev-lora-trainer/train)
|
| 84 |
+
- **Number of Images**: 80
|
| 85 |
+
- **Trigger Word**: GHIBSKY
|
| 86 |
+
- **Auto-captioning**: Enabled
|
| 87 |
+
- **Auto-captioning Prefix**: “GHIBSKY style, ”
|
| 88 |
+
- **Auto-captioning Suffix**: “”
|
| 89 |
+
- **Training Steps**: 1600
|
| 90 |
+
- **Learning Rate**: 0.0004
|
| 91 |
+
- **Batch Size**: 5
|
| 92 |
+
- **LoRA Rank**: 16
|
| 93 |
+
- **Caption Dropout Rate**: 0.2
|
| 94 |
+
|
| 95 |
+
## Training Details (previous version)
|
| 96 |
|
| 97 |
- **Trained Using**: [ostris/flux-dev-lora-trainer](https://replicate.com/ostris/flux-dev-lora-trainer/train)
|
| 98 |
- **Number of Images**: 35
|