Update README.md
Browse files
README.md
CHANGED
@@ -1,11 +1,7 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
-
|
5 |
-
|
6 |
-
Self-RAG is trained on our instruction-following corpora with interleaving passages and reflection tokens using the standard next-token prediction objective, enabling efficient and stable learning with fine-grained feedback.
|
7 |
-
At inference, we leverage reflection tokens covering diverse aspects of generations to sample the best output aligning users' preferences.
|
8 |
-
See full descriptions in See full descriptions in [our paper](https://arxiv.org/abs/2310.11511).
|
9 |
|
10 |
## Usage
|
11 |
Here, we show an easy way to quickly download our model from HuggingFace and run with `vllm` with pre-given passages. Make sure to install dependencies listed at [self-rag/requirements.txt](https://github.com/AkariAsai/self-rag/blob/main/requirements.txt).
|
@@ -53,11 +49,6 @@ or
|
|
53 |
If you have additional input.
|
54 |
You can insert paragraphs anywhere after `### Response:\n"`, but make sure to mark paragraphs as paragraph tokens (i.e., `<paragraph>{0}</paragraph>`).
|
55 |
|
56 |
-
## Training details
|
57 |
-
Our training data is available at the HuggingFace dataset [selfrag_train_data](https://huggingface.co/datasets/selfrag/selfrag_train_data).
|
58 |
-
See our official repository for the training details.
|
59 |
-
We used 8 A100 40GB for training on the Stability HPC server.
|
60 |
-
|
61 |
## Citation and contact
|
62 |
If you use this model, please cite our work:
|
63 |
```
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
4 bit quantization of: https://huggingface.co/selfrag/selfrag_llama2_7b
|
|
|
|
|
|
|
|
|
5 |
|
6 |
## Usage
|
7 |
Here, we show an easy way to quickly download our model from HuggingFace and run with `vllm` with pre-given passages. Make sure to install dependencies listed at [self-rag/requirements.txt](https://github.com/AkariAsai/self-rag/blob/main/requirements.txt).
|
|
|
49 |
If you have additional input.
|
50 |
You can insert paragraphs anywhere after `### Response:\n"`, but make sure to mark paragraphs as paragraph tokens (i.e., `<paragraph>{0}</paragraph>`).
|
51 |
|
|
|
|
|
|
|
|
|
|
|
52 |
## Citation and contact
|
53 |
If you use this model, please cite our work:
|
54 |
```
|