---
license: mit
---
Model versions
* v1 : part 3 (trained on 30,000 data) of trial v1
* v2 : part 5 (trained on about 50,000 data) of trial v9
* v3 : part 3 (trained on about 30,000 data) of trial v35
* v4 : ensemble of 4 models (trainied on about 50,000 data) from trial v47.
* v5 : ensemble of 4 models (trainied over about 100,000 data) from trial v72 to v75.

Check the following document for details.

* [CLIP Text Deprojector](https://github.com/nanashi161382/unstable_diffusion/wiki/CLIP-Text-Deprojector)

Here are other additional resources (all of which are written in Japanese).
* https://note.com/tomo161382/n/nfded9d8e4cfa  --  v1
* https://note.com/tomo161382/n/n936f43e4147b  --  v2
* https://note.com/tomo161382/n/nf627c74ceb72
* https://note.com/tomo161382/n/nf4e87191a7ee  --  v3
* https://note.com/tomo161382/n/ncbd459c3997c
* https://note.com/tomo161382/n/n0c4e276b5958
* https://note.com/tomo161382/n/n4597114e750f  --  v4
* https://note.com/tomo161382/n/nc5aaeee4dbf4
* https://note.com/tomo161382/n/n3d1d058c25e8  --  v5
* https://note.com/tomo161382/n/n235ecc71aa15
* https://note.com/tomo161382/n/nf2df40c881ca
* https://note.com/tomo161382/n/na51afab6ee70
* https://note.com/tomo161382/n/n9f3e06db6e69
* https://note.com/tomo161382/n/n2ccd08cd26fd
* https://note.com/tomo161382/n/nf980b007212c
* https://note.com/tomo161382/n/n7aa3ca17c721
* https://note.com/tomo161382/n/n8c437b24d251
* https://note.com/tomo161382/n/ncf523bc1d4cb