Hugging Face's logo

Datasets:
MInAlA
/
crosscoder-multilayer-split-activations

Tags:
mechanistic-interpretability
crosscoder
activations
sparse-autoencoder
License:
Dataset card Files
xet
 Community
1
ishangarg183 commited on
Commit
7a5b70c
·
verified ·
1 Parent(s): e0dc03e

add samelayer_reduce llama32-3b-dpo

Browse files
Files changed (4) hide show
  1. v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/activations/activations.pt +3 -0
  2. v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/checkpoints/final.pt +3 -0
  3. v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/metrics/training_metrics.json +80 -0
  4. v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/run_meta.json +34 -0
v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/activations/activations.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:164738c2656dcba45b0197a921767da5dff5587679ca6e96c3ef9a6d5d5f91d6
3
+ size 1501941701
v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/checkpoints/final.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05ca5d1733ad7267b99642a174dc3cdecdc7dac87cf90e1704c8cf7a50d8fc9e
3
+ size 3624487706
v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/metrics/training_metrics.json ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epochs": [
3
+ 1,
4
+ 2,
5
+ 3,
6
+ 4
7
+ ],
8
+ "train_loss": [
9
+ 0.04851673393832684,
10
+ 0.028567250171035823,
11
+ 0.026409504524123856,
12
+ 0.025370360229771106
13
+ ],
14
+ "val_loss": [
15
+ 0.030632317358754692,
16
+ 0.028031306682612882,
17
+ 0.026968675787458245,
18
+ 0.026772269335482757
19
+ ],
20
+ "train_fve_base": [
21
+ 0.6179440687500241,
22
+ 0.7526046327520238,
23
+ 0.778157336247016,
24
+ 0.7906076979442291
25
+ ],
26
+ "train_fve_aligned": [
27
+ 0.6100056713182747,
28
+ 0.7455868242570551,
29
+ 0.7708290390500672,
30
+ 0.7827828095728935
31
+ ],
32
+ "val_fve_base": [
33
+ 0.729565389493373,
34
+ 0.759788408953482,
35
+ 0.7725035086352163,
36
+ 0.7754090562540823
37
+ ],
38
+ "val_fve_aligned": [
39
+ 0.7226130859389979,
40
+ 0.7538367856235404,
41
+ 0.766779824389213,
42
+ 0.7696343095514787
43
+ ],
44
+ "dead_neurons": [
45
+ 0.9130961950313815,
46
+ 0.9554086537717451,
47
+ 0.9543492382058468,
48
+ 0.9534787365517557
49
+ ],
50
+ "l0_base": [
51
+ 219.79082749562173,
52
+ 208.65767294220666,
53
+ 211.514375364857,
54
+ 212.9011420023351
55
+ ],
56
+ "l0_aligned": [
57
+ 216.24501970227672,
58
+ 206.60232413893755,
59
+ 210.75957749562173,
60
+ 212.3771161704612
61
+ ],
62
+ "self_recon": [
63
+ 0.026408582905136698,
64
+ 0.01717368742560895,
65
+ 0.015432445818457211,
66
+ 0.014596288345245264
67
+ ],
68
+ "cross_recon": [
69
+ 0.026462274951500393,
70
+ 0.017178485409769306,
71
+ 0.015436923448311712,
72
+ 0.014601193326027924
73
+ ],
74
+ "sparsity": [
75
+ 0.011523240911478033,
76
+ 0.004522168440434383,
77
+ 0.004802289213093571,
78
+ 0.004933594420478313
79
+ ]
80
+ }
v3-samelayer-seed-99/samelayer_reduce/llama32-3b-dpo/L12-14/run_meta.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "meta-llama/Llama-3.2-3B-Instruct",
3
+ "aligned_run_id": "llama32-3b-dpo",
4
+ "position": "last_prompt",
5
+ "multi_type": "reduce",
6
+ "source_activations": "/home/igarg/research/interp-alignment/interp_utils/crosscoder/results-multi-v1/v1/assembled_activations/llama32-3b-dpo/activations/activations.pt",
7
+ "base_activations_path": "/home/igarg/research/interp-alignment/interp_utils/crosscoder/results-multi-v1/v1/assembled_activations/llama32-3b-dpo/activations/activations.pt",
8
+ "source_layers": [
9
+ 12,
10
+ 13,
11
+ 14
12
+ ],
13
+ "base_source_layers": [
14
+ 12,
15
+ 13,
16
+ 14
17
+ ],
18
+ "layers": [
19
+ 12,
20
+ 13,
21
+ 14
22
+ ],
23
+ "base_layers": [
24
+ 12,
25
+ 13,
26
+ 14
27
+ ],
28
+ "aligned_layers": [
29
+ 12,
30
+ 13,
31
+ 14
32
+ ],
33
+ "layer_reduction": "mean"
34
+ }