instruction-pretrain
commited on
Commit
•
0ecc04f
1
Parent(s):
457cad7
Update README.md
Browse files
README.md
CHANGED
@@ -16,6 +16,7 @@ We explore supervised multitask pre-training by proposing ***Instruction Pre-Tra
|
|
16 |
|
17 |
**************************** **Updates** ****************************
|
18 |
* 2024/9/20: Our paper has been accepted by EMNLP 2024 main conference🎉
|
|
|
19 |
* 2024/8/29: Updated [guidelines](https://huggingface.co/instruction-pretrain/medicine-Llama3-8B) on evaluating any 🤗Huggingface models on the domain-specific tasks
|
20 |
* 2024/7/31: Updated pre-training suggestions in the `Advanced Usage` section of [instruction-synthesizer](https://huggingface.co/instruction-pretrain/instruction-synthesizer)
|
21 |
* 2024/7/15: We scaled up the pre-trained tokens from 100B to 250B, with the number of synthesized instruction-response pairs reaching 500M. The performance trend on downstream tasks throughout the pre-training process:
|
@@ -224,7 +225,7 @@ text_ids = tokenizer(text, add_special_tokens=False, **kwargs).input_ids
|
|
224 |
## Citation
|
225 |
If you find our work helpful, please cite us:
|
226 |
|
227 |
-
Instruction Pre-Training (EMNLP 2024)
|
228 |
```bibtex
|
229 |
@article{cheng2024instruction,
|
230 |
title={Instruction Pre-Training: Language Models are Supervised Multitask Learners},
|
@@ -234,7 +235,7 @@ Instruction Pre-Training (EMNLP 2024)
|
|
234 |
}
|
235 |
```
|
236 |
|
237 |
-
[Adapt LLM to Domains](https://huggingface.co/papers/2309.09530)
|
238 |
```bibtex
|
239 |
@inproceedings{
|
240 |
cheng2024adapting,
|
|
|
16 |
|
17 |
**************************** **Updates** ****************************
|
18 |
* 2024/9/20: Our paper has been accepted by EMNLP 2024 main conference🎉
|
19 |
+
* 2024/9/11: Updated [FAQ on continual pre-training from Llama3](https://huggingface.co/instruction-pretrain/instruction-synthesizer)
|
20 |
* 2024/8/29: Updated [guidelines](https://huggingface.co/instruction-pretrain/medicine-Llama3-8B) on evaluating any 🤗Huggingface models on the domain-specific tasks
|
21 |
* 2024/7/31: Updated pre-training suggestions in the `Advanced Usage` section of [instruction-synthesizer](https://huggingface.co/instruction-pretrain/instruction-synthesizer)
|
22 |
* 2024/7/15: We scaled up the pre-trained tokens from 100B to 250B, with the number of synthesized instruction-response pairs reaching 500M. The performance trend on downstream tasks throughout the pre-training process:
|
|
|
225 |
## Citation
|
226 |
If you find our work helpful, please cite us:
|
227 |
|
228 |
+
[Instruction Pre-Training](https://huggingface.co/papers/2406.14491) (EMNLP 2024)
|
229 |
```bibtex
|
230 |
@article{cheng2024instruction,
|
231 |
title={Instruction Pre-Training: Language Models are Supervised Multitask Learners},
|
|
|
235 |
}
|
236 |
```
|
237 |
|
238 |
+
[Adapt LLM to Domains](https://huggingface.co/papers/2309.09530) (ICLR 2024)
|
239 |
```bibtex
|
240 |
@inproceedings{
|
241 |
cheng2024adapting,
|