Update README.md
Browse files
README.md
CHANGED
|
@@ -25,6 +25,13 @@ Rest in peace 4070.
|
|
| 25 |
---
|
| 26 |
|
| 27 |
### Our Work:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
**[Gemma3NPC models](https://huggingface.co/collections/npcLM/gemma3npc-all)**
|
| 29 |
|
| 30 |
- `Gemma3NPC-it` - Specialized for NPC dialogues(Recommended)
|
|
@@ -32,20 +39,45 @@ Rest in peace 4070.
|
|
| 32 |
- `Gemma3NPC` - General Roleplaying
|
| 33 |
- `Gemma3NPC-filtered` - filtered model, more censored
|
| 34 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
**[GemmaReLe](https://huggingface.co/collections/npcLM/rele)**
|
| 36 |
|
| 37 |
- A model finetuned specifically for ReLe, performance requires further evaluation.
|
| 38 |
|
| 39 |
-
**[Datasets](https://huggingface.co/collections/npcLM/finetuning-datasets)**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
|
| 41 |
-
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
| 43 |
- SonnetRolePlay
|
| 44 |
- ReLe_Synthetic_v1
|
| 45 |
|
| 46 |
-
**[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
-
**[Demo Video](https://youtu.be/
|
| 49 |
|
| 50 |
---
|
| 51 |
|
|
|
|
| 25 |
---
|
| 26 |
|
| 27 |
### Our Work:
|
| 28 |
+
|
| 29 |
+
**[Gemma4NPC models](https://huggingface.co/collections/npcLM/gemma4npc)**
|
| 30 |
+
Our newest version of the GemmaNPC series, using the new Gemma 4 models.
|
| 31 |
+
- `Gemma4NPC-E4B` - larger (Recommended)
|
| 32 |
+
- `Gemma4NPC-E2B` - for edge devices
|
| 33 |
+
Both have vision and audio functining properly.
|
| 34 |
+
|
| 35 |
**[Gemma3NPC models](https://huggingface.co/collections/npcLM/gemma3npc-all)**
|
| 36 |
|
| 37 |
- `Gemma3NPC-it` - Specialized for NPC dialogues(Recommended)
|
|
|
|
| 39 |
- `Gemma3NPC` - General Roleplaying
|
| 40 |
- `Gemma3NPC-filtered` - filtered model, more censored
|
| 41 |
|
| 42 |
+
**[GemmaThink models](https://huggingface.co/collections/npcLM/gemmathink)**
|
| 43 |
+
|
| 44 |
+
- `gemma-3-1b-it-thinking-32k-grpo-merged` - The **complete** model trained on the 32k row reasoning dataset(recommended)
|
| 45 |
+
- `gemma-3-1b-it-thinking-20k-sft-rl` - Trained on the 20k row reasoning dataset
|
| 46 |
+
|
| 47 |
**[GemmaReLe](https://huggingface.co/collections/npcLM/rele)**
|
| 48 |
|
| 49 |
- A model finetuned specifically for ReLe, performance requires further evaluation.
|
| 50 |
|
| 51 |
+
**[Finetuning Datasets](https://huggingface.co/collections/npcLM/finetuning-datasets)**
|
| 52 |
+
The datasets used to train our models.
|
| 53 |
+
- PIPPA (processed, includes original & filtered)
|
| 54 |
+
- NPC-Dialogue (v1 & v2)
|
| 55 |
+
- SonnetRolePlay
|
| 56 |
+
- ReLe_Synthetic_v1
|
| 57 |
|
| 58 |
+
**[RP Datasets](https://huggingface.co/collections/npcLM/rp-datasets)**
|
| 59 |
+
Individual roleplaying datasets.
|
| 60 |
+
- PIPPA (original & filtered)
|
| 61 |
+
- NPC-Dialogue (v1 & v2)
|
| 62 |
+
- NPC-Quest-Dialogue
|
| 63 |
- SonnetRolePlay
|
| 64 |
- ReLe_Synthetic_v1
|
| 65 |
|
| 66 |
+
**[Reasoning Datasets](https://huggingface.co/collections/npcLM/reasoning-datasets)
|
| 67 |
+
Synthetically generated reasoning data using Qwen3-14b.
|
| 68 |
+
- code-thinking
|
| 69 |
+
- Creative-Writing-Thinking
|
| 70 |
+
- sciqa-thinking
|
| 71 |
+
- brainstorming-thinking
|
| 72 |
+
- gsm8k-thinking
|
| 73 |
+
- databricks-thinking
|
| 74 |
+
- cnn-summarization
|
| 75 |
+
|
| 76 |
+
Gemma3NPC **[Technical Blog](https://huggingface.co/blog/chimbiwide/gemma3npc)**
|
| 77 |
+
|
| 78 |
+
Gemma3NPC **[Demo Video](https://youtu.be/N8-XZQ8C6kc?si=X48VmhDfSukgpSLT)**
|
| 79 |
|
| 80 |
+
GemmaThink **[Demo Video](https://youtu.be/MrwM3-S6DVs?si=cKUUkWeWQ-Z_ffu-)**
|
| 81 |
|
| 82 |
---
|
| 83 |
|