Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -17,26 +17,26 @@ base_model:
|
|
| 17 |
|
| 18 |
| Parameter | Value |
|
| 19 |
| :-------- | :---: |
|
| 20 |
-
| **direction_index** |
|
| 21 |
-
| **attn.o_proj.max_weight** | 1.
|
| 22 |
-
| **attn.o_proj.max_weight_position** |
|
| 23 |
-
| **attn.o_proj.min_weight** |
|
| 24 |
-
| **attn.o_proj.min_weight_distance** |
|
| 25 |
-
| **mamba.out_proj.max_weight** | 1.
|
| 26 |
-
| **mamba.out_proj.max_weight_position** |
|
| 27 |
-
| **mamba.out_proj.min_weight** |
|
| 28 |
-
| **mamba.out_proj.min_weight_distance** |
|
| 29 |
-
| **mlp.shared_down_proj.max_weight** |
|
| 30 |
-
| **mlp.shared_down_proj.max_weight_position** |
|
| 31 |
-
| **mlp.shared_down_proj.min_weight** | 0.
|
| 32 |
-
| **mlp.shared_down_proj.min_weight_distance** |
|
| 33 |
|
| 34 |
## Performance
|
| 35 |
|
| 36 |
| Metric | This model | Original model ([ibm-granite/granite-4.0-h-1b](https://huggingface.co/ibm-granite/granite-4.0-h-1b)) |
|
| 37 |
| :----- | :--------: | :---------------------------: |
|
| 38 |
| **KL divergence** | 0.03 | 0 *(by definition)* |
|
| 39 |
-
| **Refusals** |
|
| 40 |
|
| 41 |
-----
|
| 42 |
|
|
|
|
| 17 |
|
| 18 |
| Parameter | Value |
|
| 19 |
| :-------- | :---: |
|
| 20 |
+
| **direction_index** | 23.75 |
|
| 21 |
+
| **attn.o_proj.max_weight** | 1.75 |
|
| 22 |
+
| **attn.o_proj.max_weight_position** | 24.08 |
|
| 23 |
+
| **attn.o_proj.min_weight** | 1.17 |
|
| 24 |
+
| **attn.o_proj.min_weight_distance** | 20.46 |
|
| 25 |
+
| **mamba.out_proj.max_weight** | 1.86 |
|
| 26 |
+
| **mamba.out_proj.max_weight_position** | 31.62 |
|
| 27 |
+
| **mamba.out_proj.min_weight** | 1.29 |
|
| 28 |
+
| **mamba.out_proj.min_weight_distance** | 18.93 |
|
| 29 |
+
| **mlp.shared_down_proj.max_weight** | 1.11 |
|
| 30 |
+
| **mlp.shared_down_proj.max_weight_position** | 36.36 |
|
| 31 |
+
| **mlp.shared_down_proj.min_weight** | 0.16 |
|
| 32 |
+
| **mlp.shared_down_proj.min_weight_distance** | 4.17 |
|
| 33 |
|
| 34 |
## Performance
|
| 35 |
|
| 36 |
| Metric | This model | Original model ([ibm-granite/granite-4.0-h-1b](https://huggingface.co/ibm-granite/granite-4.0-h-1b)) |
|
| 37 |
| :----- | :--------: | :---------------------------: |
|
| 38 |
| **KL divergence** | 0.03 | 0 *(by definition)* |
|
| 39 |
+
| **Refusals** | 7/100 | 93/100 |
|
| 40 |
|
| 41 |
-----
|
| 42 |
|