Add info: Improved Text Encoder for Flux.1 / CLIP-L
Browse files
README.md
CHANGED
|
@@ -3,6 +3,14 @@ license: mit
|
|
| 3 |
datasets:
|
| 4 |
- SPRIGHT-T2I/spright_coco
|
| 5 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
## Update 11/AUG/2024:
|
| 7 |
|
| 8 |
New Best-Performing CLIP ViT-L/14 'GmP-smooth' model added (simply download the files named *BEST*!):
|
|
|
|
| 3 |
datasets:
|
| 4 |
- SPRIGHT-T2I/spright_coco
|
| 5 |
---
|
| 6 |
+
## Update 03/SEP/2024:
|
| 7 |
+
|
| 8 |
+
Improved CLIP-L for use with Flux.1 (improved TEXT and prompt detail adherence). Trained with high temperature of 0.1 + tinkering. Otherwise, same GmP-smooth-labels code as previous model fine-tune; same dataset (see below for link to GitHub).
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
----
|
| 14 |
## Update 11/AUG/2024:
|
| 15 |
|
| 16 |
New Best-Performing CLIP ViT-L/14 'GmP-smooth' model added (simply download the files named *BEST*!):
|