SuperbEmphasis commited on
Commit
11a0d90
·
verified ·
1 Parent(s): d94ce11

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +10 -0
README.md CHANGED
@@ -18,6 +18,16 @@ tags:
18
  ![image/png](https://cdn-uploads.huggingface.co/production/uploads/65bc2496ce846f8aa90aacbe/rf8TNQYh39MuwA72gcHcl.png)
19
  2x12B version of Velvet Eclipse 4x12B
20
 
 
 
 
 
 
 
 
 
 
 
21
  I have been wanting a better model for RP on a 24GB Nvidia card. And there are some great models out there, but I wanted something that I knew I could quantize to Q4, have a great context size, have a very fast response, and would provide some dynamic content. The total is around 30B, but since there are 2/3 models active, the response is quite fast!
22
 
23
  This is using 2x Mistral Nemo finetunes, each with a separate purpose.
 
18
  ![image/png](https://cdn-uploads.huggingface.co/production/uploads/65bc2496ce846f8aa90aacbe/rf8TNQYh39MuwA72gcHcl.png)
19
  2x12B version of Velvet Eclipse 4x12B
20
 
21
+ You may notice the name change... This may or may not have happened due to me SSHing into my GPU virtual machine from my cell phone's SSH client... Whoops. But I decided to keep it this way because I kinda like it.
22
+
23
+
24
+ | Model | Description |
25
+ | ----- | ----------- |
26
+ | [Velvet Eclipse 2x12B](https://huggingface.co/SuperbEmphasis/Viloet-Eclipse-2x12B-v0.2-MINI) | A slimmer model with the ERP and RP experts.|
27
+ | [Velvet Eclipse 2x12B Reasoning](https://huggingface.co/SuperbEmphasis/Viloet-Eclipse-2x12B-v0.2-MINI-Reasoning) | A slimmer model with the ERP and the Reasoning Experts |
28
+ | [Velvet Eclipse 4x12B Reasoning](https://huggingface.co/SuperbEmphasis/Velvet-Eclipse-4x12B-v0.2) | Full 4x12B Parameter Velvet Eclipse |
29
+
30
+
31
  I have been wanting a better model for RP on a 24GB Nvidia card. And there are some great models out there, but I wanted something that I knew I could quantize to Q4, have a great context size, have a very fast response, and would provide some dynamic content. The total is around 30B, but since there are 2/3 models active, the response is quite fast!
32
 
33
  This is using 2x Mistral Nemo finetunes, each with a separate purpose.