Update README.md
Browse files
README.md
CHANGED
@@ -8,14 +8,15 @@ language:
|
|
8 |
|
9 |
**She's back!**
|
10 |
|
11 |
-
Stheno's Sister Model,
|
|
|
12 |
```
|
13 |
- Same Dataset used as Stheno v3.2 -> See notes there.
|
14 |
- LoRA Fine-Tune -> FFT is simply too expensive.
|
15 |
-
- Trained
|
16 |
```
|
17 |
|
18 |
-
Testing Notes
|
19 |
```
|
20 |
- Better prompt adherence.
|
21 |
- Better anatomy / spatial awareness.
|
@@ -29,7 +30,7 @@ Testing Notes
|
|
29 |
|
30 |
---
|
31 |
|
32 |
-
Recommended Sampler Settings
|
33 |
```
|
34 |
Temperature - 1.17
|
35 |
min_p - 0.075
|
|
|
8 |
|
9 |
**She's back!**
|
10 |
|
11 |
+
Stheno's Sister Model, designed to impress.
|
12 |
+
|
13 |
```
|
14 |
- Same Dataset used as Stheno v3.2 -> See notes there.
|
15 |
- LoRA Fine-Tune -> FFT is simply too expensive.
|
16 |
+
- Trained over 8x H100 SXMs and then some more afterwards.
|
17 |
```
|
18 |
|
19 |
+
**Testing Notes**
|
20 |
```
|
21 |
- Better prompt adherence.
|
22 |
- Better anatomy / spatial awareness.
|
|
|
30 |
|
31 |
---
|
32 |
|
33 |
+
**Recommended Sampler Settings**:
|
34 |
```
|
35 |
Temperature - 1.17
|
36 |
min_p - 0.075
|