Update README.md
Browse files
README.md
CHANGED
|
@@ -4,11 +4,20 @@ pipeline_tag: text-generation
|
|
| 4 |
|
| 5 |
Collection of resources and models for storytelling and roleplay. Updated December 2024.
|
| 6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
# ⚒️ Base models
|
| 8 |
|
| 9 |
- Llama 3 (8B) - the OG
|
| 10 |
- [Mistral-Nemo-Base-2407](https://huggingface.co/mistralai/Mistral-Nemo-Base-2407) (12B)
|
| 11 |
-
-
|
|
|
|
| 12 |
|
| 13 |
# 🤖 Instruct models
|
| 14 |
|
|
|
|
| 4 |
|
| 5 |
Collection of resources and models for storytelling and roleplay. Updated December 2024.
|
| 6 |
|
| 7 |
+
**Current favorite**: [mistralai/Mistral-Small-24B-Instruct-2501](https://huggingface.co/mistralai/Mistral-Small-24B-Instruct-2501)
|
| 8 |
+
|
| 9 |
+
Some notes on best usage:
|
| 10 |
+
- some people prefer base models over instruct models, but base models are too unruly
|
| 11 |
+
- in general, roleplay finetunes I find to be braindamaged
|
| 12 |
+
- you also don't want to "overparameterize" by writing too long a prompt
|
| 13 |
+
- Conclusion: use original instruct models with short prompts
|
| 14 |
+
|
| 15 |
# ⚒️ Base models
|
| 16 |
|
| 17 |
- Llama 3 (8B) - the OG
|
| 18 |
- [Mistral-Nemo-Base-2407](https://huggingface.co/mistralai/Mistral-Nemo-Base-2407) (12B)
|
| 19 |
+
- Qwen2.5
|
| 20 |
+
- Mistral Small
|
| 21 |
|
| 22 |
# 🤖 Instruct models
|
| 23 |
|