Update README.md
Browse files
README.md
CHANGED
|
@@ -87,6 +87,12 @@ model = AutoModel.from_pretrained("fc91/phi3-mini-instruct-full_ethics-lora")
|
|
| 87 |
|
| 88 |
["hendrycks/ethics"](https://huggingface.co/datasets/hendrycks/ethics)
|
| 89 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
### Training Procedure
|
| 91 |
|
| 92 |
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
|
@@ -133,6 +139,12 @@ Validation Loss = 0.119734
|
|
| 133 |
|
| 134 |
["hendrycks/ethics"](https://huggingface.co/datasets/hendrycks/ethics)
|
| 135 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
#### Factors
|
| 137 |
|
| 138 |
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
|
|
|
| 87 |
|
| 88 |
["hendrycks/ethics"](https://huggingface.co/datasets/hendrycks/ethics)
|
| 89 |
|
| 90 |
+
The following subsets of the above dataset were leveraged:
|
| 91 |
+
-commonsense (10k random samples)
|
| 92 |
+
-deontology (10k random samples)
|
| 93 |
+
-justice (10k random samples)
|
| 94 |
+
-utilitarianism (10k random samples)
|
| 95 |
+
|
| 96 |
### Training Procedure
|
| 97 |
|
| 98 |
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
|
|
|
| 139 |
|
| 140 |
["hendrycks/ethics"](https://huggingface.co/datasets/hendrycks/ethics)
|
| 141 |
|
| 142 |
+
The following subsets of the above dataset were leveraged:
|
| 143 |
+
-commonsense (2.5k random samples)
|
| 144 |
+
-deontology (2.5k random samples)
|
| 145 |
+
-justice (2.5k random samples)
|
| 146 |
+
-utilitarianism (2.5k random samples)
|
| 147 |
+
|
| 148 |
#### Factors
|
| 149 |
|
| 150 |
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|