Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: Mihaiii/Pallas-0.5-LASER-0.4
|
3 |
+
inference: false
|
4 |
+
license: other
|
5 |
+
license_name: yi-license
|
6 |
+
license_link: https://huggingface.co/01-ai/Yi-34B/blob/main/LICENSE
|
7 |
+
metrics:
|
8 |
+
- accuracy
|
9 |
+
---
|
10 |
+
|
11 |
+
This model has a [LASER](https://pratyushasharma.github.io/laser/) intervention on [Mihaiii/Pallas-0.5-LASER-0.4](https://huggingface.co/Mihaiii/Pallas-0.5-LASER-0.4) .
|
12 |
+
|
13 |
+
Configs used:
|
14 |
+
|
15 |
+
- lnum: 54
|
16 |
+
- lnames: mlp (meaning: ["mlp.gate_proj.weight", "mlp.up_proj.weight", "mlp.down_proj.weight"])
|
17 |
+
- rate: 8
|
18 |
+
- dataset: bigbench (subset: causal_judgement)
|
19 |
+
- intervention type: rank-reduction
|
20 |
+
|
21 |
+
|
22 |
+
|Name|Validation acc (higher is better)|Validation logloss (lower is better)|Test acc (higher is better)|Test logloss (lower is better)|
|
23 |
+
|---|---|---|---|---|
|
24 |
+
|Pallas-0.5|55.263|1.650|60.526|1.463|
|
25 |
+
|Pallas-0.5-LASER-0.1|55.263|1.639|61.184|1.451|
|
26 |
+
|Pallas-0.5-LASER-0.2|55.263|1.646|61.184|1.458|
|
27 |
+
|Pallas-0.5-LASER-0.3|55.263|1.575|61.842|1.382|
|
28 |
+
|Pallas-0.5-LASER-0.4|55.263|1.525|61.842|1.326|
|
29 |
+
|Pallas-0.5-LASER-0.5|55.263|1.484|61.842|1.297|
|
30 |
+
|
31 |
+
|
32 |
+
In order to replicate on a single A100, you can use [my branch](https://github.com/Mihaiii/laser/tree/allow-Yi-on-one-A100) (the original code will throw OOM for 34b models).
|