ChuckMcSneed commited on
Commit
16ff62a
·
verified ·
1 Parent(s): 5af3aaa

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -2
README.md CHANGED
@@ -10,7 +10,7 @@ language:
10
  - ru
11
  ---
12
  ![logo-gembo-1.1.png](logo-gembo-1.1.png)
13
- This is like [Gembo v1](https://huggingface.co/ChuckMcSneed/Gembo-v1-70b), but with 6-7% more human data. Does perform a bit worse on the benches(who cares), but should be able to write in more diverse styles(Absolutely worth it, IMO. See [waxwing-styles.txt](waxwing-styles.txt), try adding "challenging, provocative, wry, cynical, perceptive, concise" to your prompt instead of what you usually use and see the difference). Mainly made for RP, but should be okay as an assistant. Turned out quite good, considering the amount of LORAs I merged into it.
14
 
15
  # Observations
16
  - GPTisms and repetition: put temperature and rep. pen. higher, make GPTisms stop sequences
@@ -19,12 +19,15 @@ This is like [Gembo v1](https://huggingface.co/ChuckMcSneed/Gembo-v1-70b), but w
19
  - Fairly neutral, can be violent if you ask it really good, Goliath is a bit better at it
20
  - Has a bit of optimism baked in, but it's not very severe, maybe a tiny bit more than in v1?
21
  - Don't put too many style tags, here less is better
22
- - Unlike v1, 1.1 knows much better when to stop
23
  - Needs more wrangling than v1, but once you get it going it's good
24
  - Sometimes can't handle '
25
  - Moderately intelligent
26
  - Quite creative
27
 
 
 
 
28
  # Naming
29
  Internal name of this model was euryale-guano-saiga-med-janboros-kim-wing-lima-wiz-tony-d30-s40, but I decided to keep it short, and since it was iteration G in my files, I called it "Gembo".
30
 
 
10
  - ru
11
  ---
12
  ![logo-gembo-1.1.png](logo-gembo-1.1.png)
13
+ This is like [Gembo v1](https://huggingface.co/ChuckMcSneed/Gembo-v1-70b), but with 6-7% more human data. Does perform a bit worse on the benches(who cares? I do.), but should be able to write in more diverse styles(See [waxwing-styles.txt](waxwing-styles.txt), tested it with v1, v1 does it better.). Mainly made for RP, but should be okay as an assistant. Turned out quite good, considering the amount of LORAs I merged into it.
14
 
15
  # Observations
16
  - GPTisms and repetition: put temperature and rep. pen. higher, make GPTisms stop sequences
 
19
  - Fairly neutral, can be violent if you ask it really good, Goliath is a bit better at it
20
  - Has a bit of optimism baked in, but it's not very severe, maybe a tiny bit more than in v1?
21
  - Don't put too many style tags, here less is better
22
+ - Unlike v1, 1.1 knows a bit better when to stop
23
  - Needs more wrangling than v1, but once you get it going it's good
24
  - Sometimes can't handle '
25
  - Moderately intelligent
26
  - Quite creative
27
 
28
+ # Worth over v1?
29
+ Meh. I prefer hyperslop over this "humanized" one. Maybe I've been poisoned by slop.
30
+
31
  # Naming
32
  Internal name of this model was euryale-guano-saiga-med-janboros-kim-wing-lima-wiz-tony-d30-s40, but I decided to keep it short, and since it was iteration G in my files, I called it "Gembo".
33