Update README.md
Browse files
README.md
CHANGED
@@ -86,6 +86,12 @@ The control-vectors in this repo were created as an experiment by combining two
|
|
86 |
|
87 |
I also tried using 3 different values for the `--regularisation_factor` option; `1.0` (the default), `0.5` and `0.0`:
|
88 |
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
```sh
|
90 |
python create_control_vectors.py --model_id ./Mistral-Large-Instruct-2407 --output_path Mistral-Large-Instruct-2407_alignment_ --system_prompt_file data/system_messages_alignment.json --prompt_file data/prompts.txt --num_prompt_sample 500 --regularisation_factor 1.0
|
91 |
python create_control_vectors.py --model_id ./Mistral-Large-Instruct-2407 --output_path Mistral-Large-Instruct-2407_alignment_ --system_prompt_file data/system_messages_alignment.json --prompt_file data/prompts.txt --num_prompt_sample 500 --regularisation_factor 0.5
|
@@ -98,8 +104,4 @@ python create_control_vectors.py --model_id ./c4ai-command-r-plus --output_path
|
|
98 |
python create_control_vectors.py --model_id ./c4ai-command-r-plus --output_path Mistral-Large-Instruct-2407_alignment_ --system_prompt_file data/system_messages_alignment.json --prompt_file data/prompts.txt --num_prompt_sample 500 --regularisation_factor 0.0
|
99 |
```
|
100 |
|
101 |
-
- [regularisation_factor = 0.0](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1.1-EXPERIMENTAL/tree/main/regularisation_factor%20%3D%200.0)
|
102 |
-
- [regularisation_factor = 0.5](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1.1-EXPERIMENTAL/tree/main/regularisation_factor%20%3D%200.5)
|
103 |
-
- [regularisation_factor = 1.0](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1.1-EXPERIMENTAL/tree/main/regularisation_factor%20%3D%201.0)
|
104 |
-
|
105 |
Please leave feedback if you find these useful and I will consider running the same on the other 9 models in the main [creative-writing-control-vectors-v2.1](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1) repo.
|
|
|
86 |
|
87 |
I also tried using 3 different values for the `--regularisation_factor` option; `1.0` (the default), `0.5` and `0.0`:
|
88 |
|
89 |
+
- [regularisation_factor = 0.0](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1.1-EXPERIMENTAL/tree/main/regularisation_factor%20%3D%200.0)
|
90 |
+
- [regularisation_factor = 0.5](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1.1-EXPERIMENTAL/tree/main/regularisation_factor%20%3D%200.5)
|
91 |
+
- [regularisation_factor = 1.0](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1.1-EXPERIMENTAL/tree/main/regularisation_factor%20%3D%201.0)
|
92 |
+
|
93 |
+
Created using:
|
94 |
+
|
95 |
```sh
|
96 |
python create_control_vectors.py --model_id ./Mistral-Large-Instruct-2407 --output_path Mistral-Large-Instruct-2407_alignment_ --system_prompt_file data/system_messages_alignment.json --prompt_file data/prompts.txt --num_prompt_sample 500 --regularisation_factor 1.0
|
97 |
python create_control_vectors.py --model_id ./Mistral-Large-Instruct-2407 --output_path Mistral-Large-Instruct-2407_alignment_ --system_prompt_file data/system_messages_alignment.json --prompt_file data/prompts.txt --num_prompt_sample 500 --regularisation_factor 0.5
|
|
|
104 |
python create_control_vectors.py --model_id ./c4ai-command-r-plus --output_path Mistral-Large-Instruct-2407_alignment_ --system_prompt_file data/system_messages_alignment.json --prompt_file data/prompts.txt --num_prompt_sample 500 --regularisation_factor 0.0
|
105 |
```
|
106 |
|
|
|
|
|
|
|
|
|
107 |
Please leave feedback if you find these useful and I will consider running the same on the other 9 models in the main [creative-writing-control-vectors-v2.1](https://huggingface.co/jukofyork/creative-writing-control-vectors-v2.1) repo.
|