Update README.md
Browse files
README.md
CHANGED
@@ -17,15 +17,11 @@ tags:
|
|
17 |
((MantiCore3E+VicunaCocktail)+[SuperCOT+[StorytellingV2+(SuperHOTProtoType-8192ctx+Metharme)]])
|
18 |
|
19 |
This model is the result of an experimental use of LoRAs on language models and model merges that are not the base HuggingFace-format LLaMA model they were intended for.
|
|
|
20 |
|
21 |
-
|
22 |
|
23 |
-
|
24 |
-
Determine if applying LoRAs and stacking LoRAs onto merged models bypasses the zero-sum result of weight-sum model merging.
|
25 |
-
|
26 |
-
The desired result is to additively apply desired features without paradoxically watering down a model's effective behavior.
|
27 |
-
|
28 |
-
Results are very promising. Further tests are required.
|
29 |
|
30 |
Releasing Chimera as-is; Alpaca instruct verified working, Vicuna instruct formats should work.
|
31 |
If using KoboldAI or Text-Generation-WebUI, recommend switching between Godlike and Storywriter presets and adjusting output length + instructions in memory.
|
|
|
17 |
((MantiCore3E+VicunaCocktail)+[SuperCOT+[StorytellingV2+(SuperHOTProtoType-8192ctx+Metharme)]])
|
18 |
|
19 |
This model is the result of an experimental use of LoRAs on language models and model merges that are not the base HuggingFace-format LLaMA model they were intended for.
|
20 |
+
The desired outcome is to additively apply desired features without paradoxically watering down a model's effective behavior.
|
21 |
|
22 |
+
Potential limitations - LoRAs applied on top of each other may intercompete.
|
23 |
|
24 |
+
Subjective results are very promising. Further tests are required to include isolated tests and benchmarking (exact method not determined at this time).
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
Releasing Chimera as-is; Alpaca instruct verified working, Vicuna instruct formats should work.
|
27 |
If using KoboldAI or Text-Generation-WebUI, recommend switching between Godlike and Storywriter presets and adjusting output length + instructions in memory.
|