Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language: en
|
| 4 |
+
tags:
|
| 5 |
+
- conversational
|
| 6 |
+
---
|
| 7 |
+
# BART chatbot trained on [LIGHT](https://parl.ai/projects/light/) dataset with [Text Generative Adversarial Imitation Learning](https://arxiv.org/abs/2004.13796)
|
| 8 |
+
|
| 9 |
+
This model is intended to be used with [npc-engine](https://github.com/npc-engine/npc-engine).
|
| 10 |
+
|
| 11 |
+
It was based on [facebook/bart-large](https://huggingface.co/facebook/bart-large). [microsoft/deberta-base](https://huggingface.co/microsoft/deberta-base) was used as an adversarial for GAIL stage.
|
| 12 |
+
|
| 13 |
+
MSE fine-tuning: https://colab.research.google.com/drive/1goFW4X6qyK_vyLSdkanyFf0-zO_rrO1W?usp=sharing
|
| 14 |
+
|
| 15 |
+
GAIL fine-tuning: https://colab.research.google.com/drive/1jxHNWBh9em9FWevo99WBBdSkUSvrUvTx?usp=sharing
|
| 16 |
+
|
| 17 |
+
Packages used during training are not supported so it may not work anymore.
|