Commit
•
8901457
1
Parent(s):
a4f8599
Update README.md
Browse files
README.md
CHANGED
|
@@ -9,6 +9,8 @@ This is an exmaple of character substitution watermarking in a language model. A
|
|
| 9 |
|
| 10 |
This is as a LORA finetuning of https://huggingface.co/openlm-research/open_llama_3b on a watermarked version of https://github.com/tloen/alpaca-lora/blob/main/alpaca_data_cleaned_archive.json. for 2 epochs and 800 total steps.
|
| 11 |
|
|
|
|
|
|
|
| 12 |
@software{openlm2023openllama,
|
| 13 |
author = {Geng, Xinyang and Liu, Hao},
|
| 14 |
title = {OpenLLaMA: An Open Reproduction of LLaMA},
|
|
|
|
| 9 |
|
| 10 |
This is as a LORA finetuning of https://huggingface.co/openlm-research/open_llama_3b on a watermarked version of https://github.com/tloen/alpaca-lora/blob/main/alpaca_data_cleaned_archive.json. for 2 epochs and 800 total steps.
|
| 11 |
|
| 12 |
+
The script I wrote to modify datasets by substituting characters between two user-defined strings has been included. (cswm.py "character substitution watermarking")
|
| 13 |
+
|
| 14 |
@software{openlm2023openllama,
|
| 15 |
author = {Geng, Xinyang and Liu, Hao},
|
| 16 |
title = {OpenLLaMA: An Open Reproduction of LLaMA},
|