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
ac3f3f6
·
verified ·
1 Parent(s): 05996ff

add samelayer_separate llama32-3b-dpo

Browse files
Files changed (12) hide show
  1. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/activations/activations.pt +3 -0
  2. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/checkpoints/final.pt +3 -0
  3. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/metrics/training_metrics.json +80 -0
  4. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/run_meta.json +27 -0
  5. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/activations/activations.pt +3 -0
  6. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/checkpoints/final.pt +3 -0
  7. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/metrics/training_metrics.json +80 -0
  8. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/run_meta.json +27 -0
  9. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/activations/activations.pt +3 -0
  10. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/checkpoints/final.pt +3 -0
  11. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/metrics/training_metrics.json +80 -0
  12. v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/run_meta.json +27 -0
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/activations/activations.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e567a5b1219368fd8730d3198e552ac6d7afc0ba08ba5d63cc5a4e78c0578c4
3
+ size 1501941637
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/checkpoints/final.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:076821cbed26607f60d0844f749d688fe47dde412e24f87f956c2362a81e915c
3
+ size 3624487706
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/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
+ 0.052449775645503925,
10
+ 0.037926328770263945,
11
+ 0.036544836599812544,
12
+ 0.035776839331140384
13
+ ],
14
+ "val_loss": [
15
+ 0.039111704492444145,
16
+ 0.037578351171540965,
17
+ 0.03687173914144801,
18
+ 0.03671000236694101
19
+ ],
20
+ "train_fve_base": [
21
+ 0.5379605463771742,
22
+ 0.6575857759635488,
23
+ 0.6831412057882155,
24
+ 0.69748012764163
25
+ ],
26
+ "train_fve_aligned": [
27
+ 0.4444856291546574,
28
+ 0.5546883888155553,
29
+ 0.5776039686247542,
30
+ 0.5901877406530052
31
+ ],
32
+ "val_fve_base": [
33
+ 0.6378958727676831,
34
+ 0.6633234838540641,
35
+ 0.6747234282693314,
36
+ 0.6785684455751748
37
+ ],
38
+ "val_fve_aligned": [
39
+ 0.5375414487579107,
40
+ 0.5646158928022335,
41
+ 0.5784453237868105,
42
+ 0.5823782566954329
43
+ ],
44
+ "dead_neurons": [
45
+ 0.9047749383854574,
46
+ 0.9567796474162835,
47
+ 0.9561833339132765,
48
+ 0.9556857882294819
49
+ ],
50
+ "l0_base": [
51
+ 199.68282983070637,
52
+ 178.48982049036778,
53
+ 186.42144629305312,
54
+ 191.7795351722125
55
+ ],
56
+ "l0_aligned": [
57
+ 182.6806589316988,
58
+ 164.18897767075308,
59
+ 180.96196366024517,
60
+ 188.89198409223584
61
+ ],
62
+ "self_recon": [
63
+ 0.028287402302531284,
64
+ 0.022075222441964658,
65
+ 0.020765139392408028,
66
+ 0.020039599961875754
67
+ ],
68
+ "cross_recon": [
69
+ 0.03152601383148023,
70
+ 0.026536955127913354,
71
+ 0.025530520920090973,
72
+ 0.024951663259710405
73
+ ],
74
+ "sparsity": [
75
+ 0.011551967584699289,
76
+ 0.005236324004424433,
77
+ 0.005567488607706396,
78
+ 0.005756573833936202
79
+ ]
80
+ }
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L10__aligned_L12/run_meta.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": "separate",
6
+ "source_activations": "/home/ubuntu/interp-alignment-crosscoder/v1/assembled_activations/llama32-3b-dpo/activations/activations.pt",
7
+ "base_activations_path": "/home/ubuntu/interp-alignment-crosscoder/v1/base_activations/llama32-3b-union/activations/base_activations.pt",
8
+ "source_layers": [
9
+ 12,
10
+ 13,
11
+ 14
12
+ ],
13
+ "base_source_layers": [
14
+ 10,
15
+ 11,
16
+ 12,
17
+ 13,
18
+ 14,
19
+ 23,
20
+ 24,
21
+ 25,
22
+ 26
23
+ ],
24
+ "layer": 12,
25
+ "base_layer": 10,
26
+ "aligned_layer": 12
27
+ }
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/activations/activations.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6565198ed1d9073f9e9719bd93e82094d37c85bafc903996f9322ef0bf5106a
3
+ size 1501941637
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/checkpoints/final.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba6799bf60c5428835b62920f0b93f368a593581253490d2e590dc2650d8444f
3
+ size 3624487706
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/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
+ 0.06034684562006334,
10
+ 0.04285239914765987,
11
+ 0.041244965962200085,
12
+ 0.040367874976354114
13
+ ],
14
+ "val_loss": [
15
+ 0.04431246055280351,
16
+ 0.04267683139568224,
17
+ 0.041941710109017906,
18
+ 0.04177696655714075
19
+ ],
20
+ "train_fve_base": [
21
+ 0.5860457829891835,
22
+ 0.7039440290244504,
23
+ 0.7262249445261068,
24
+ 0.7382202571550448
25
+ ],
26
+ "train_fve_aligned": [
27
+ 0.4542641992883521,
28
+ 0.5618457463632462,
29
+ 0.5823798467385024,
30
+ 0.5941674131715221
31
+ ],
32
+ "val_fve_base": [
33
+ 0.6867500849419239,
34
+ 0.707200338079043,
35
+ 0.7164034440879422,
36
+ 0.718846176619305
37
+ ],
38
+ "val_fve_aligned": [
39
+ 0.545515342220586,
40
+ 0.5684155185185178,
41
+ 0.5799110694081372,
42
+ 0.5835203235062005
43
+ ],
44
+ "dead_neurons": [
45
+ 0.898241551905018,
46
+ 0.9499712402682797,
47
+ 0.9499577242800595,
48
+ 0.9495343382568716
49
+ ],
50
+ "l0_base": [
51
+ 200.82734967892586,
52
+ 175.56724314068884,
53
+ 182.48458479276124,
54
+ 187.18709865732632
55
+ ],
56
+ "l0_aligned": [
57
+ 175.97964098073555,
58
+ 149.4989054290718,
59
+ 164.78272767075308,
60
+ 174.42259559252773
61
+ ],
62
+ "self_recon": [
63
+ 0.0331904827467597,
64
+ 0.025551214000841923,
65
+ 0.024096118005713523,
66
+ 0.02328617341372783
67
+ ],
68
+ "cross_recon": [
69
+ 0.03644425757185834,
70
+ 0.030087274373647627,
71
+ 0.028893894954680772,
72
+ 0.02823165520129674
73
+ ],
74
+ "sparsity": [
75
+ 0.012578659568316865,
76
+ 0.005266275153844957,
77
+ 0.00559128970081738,
78
+ 0.005789039241420774
79
+ ]
80
+ }
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L11__aligned_L13/run_meta.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": "separate",
6
+ "source_activations": "/home/ubuntu/interp-alignment-crosscoder/v1/assembled_activations/llama32-3b-dpo/activations/activations.pt",
7
+ "base_activations_path": "/home/ubuntu/interp-alignment-crosscoder/v1/base_activations/llama32-3b-union/activations/base_activations.pt",
8
+ "source_layers": [
9
+ 12,
10
+ 13,
11
+ 14
12
+ ],
13
+ "base_source_layers": [
14
+ 10,
15
+ 11,
16
+ 12,
17
+ 13,
18
+ 14,
19
+ 23,
20
+ 24,
21
+ 25,
22
+ 26
23
+ ],
24
+ "layer": 13,
25
+ "base_layer": 11,
26
+ "aligned_layer": 13
27
+ }
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/activations/activations.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b0a69b497d4c2503c5db7141539f626300c5b99c2998d05950cc5fd0bca2387
3
+ size 1501941637
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/checkpoints/final.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb1e520bfed879887800b649976c67b5cc55aa9bacd4149ddb2e8e002ade4820
3
+ size 3624487706
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/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
+ 0.06936184684618203,
10
+ 0.04917339026771368,
11
+ 0.04704783792932063,
12
+ 0.045912571551503314
13
+ ],
14
+ "val_loss": [
15
+ 0.051206712986474265,
16
+ 0.04865274728514761,
17
+ 0.047548526264141994,
18
+ 0.047262383498134414
19
+ ],
20
+ "train_fve_base": [
21
+ 0.5766915626784758,
22
+ 0.6999707951645926,
23
+ 0.7245593924516832,
24
+ 0.7373763519424621
25
+ ],
26
+ "train_fve_aligned": [
27
+ 0.4872064327748282,
28
+ 0.6022230053741335,
29
+ 0.625215293542725,
30
+ 0.6388012777276716
31
+ ],
32
+ "val_fve_base": [
33
+ 0.6789751655144217,
34
+ 0.7065545707472956,
35
+ 0.7178483084234267,
36
+ 0.7210081016830124
37
+ ],
38
+ "val_fve_aligned": [
39
+ 0.583134077918467,
40
+ 0.6113991291111053,
41
+ 0.6269694860068915,
42
+ 0.6313064597664079
43
+ ],
44
+ "dead_neurons": [
45
+ 0.9109237759139757,
46
+ 0.9583724516216107,
47
+ 0.9579495170696813,
48
+ 0.957303654082811
49
+ ],
50
+ "l0_base": [
51
+ 211.8811478400467,
52
+ 193.52214681844717,
53
+ 199.11412726211324,
54
+ 202.78714244016345
55
+ ],
56
+ "l0_aligned": [
57
+ 203.33501167542323,
58
+ 188.23648204903677,
59
+ 199.3588733216579,
60
+ 203.99746424401636
61
+ ],
62
+ "self_recon": [
63
+ 0.03815987810027126,
64
+ 0.028675192373359474,
65
+ 0.02677436479597292,
66
+ 0.02570984768749188
67
+ ],
68
+ "cross_recon": [
69
+ 0.04208693216770661,
70
+ 0.0333005928755746,
71
+ 0.03154237282946836,
72
+ 0.03072873527000365
73
+ ],
74
+ "sparsity": [
75
+ 0.014367195467426024,
76
+ 0.007177960527740023,
77
+ 0.007656523775444157,
78
+ 0.007911229476776306
79
+ ]
80
+ }
v3-samelayer-seed-99/samelayer_separate/llama32-3b-dpo/base_L12__aligned_L14/run_meta.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": "separate",
6
+ "source_activations": "/home/ubuntu/interp-alignment-crosscoder/v1/assembled_activations/llama32-3b-dpo/activations/activations.pt",
7
+ "base_activations_path": "/home/ubuntu/interp-alignment-crosscoder/v1/base_activations/llama32-3b-union/activations/base_activations.pt",
8
+ "source_layers": [
9
+ 12,
10
+ 13,
11
+ 14
12
+ ],
13
+ "base_source_layers": [
14
+ 10,
15
+ 11,
16
+ 12,
17
+ 13,
18
+ 14,
19
+ 23,
20
+ 24,
21
+ 25,
22
+ 26
23
+ ],
24
+ "layer": 14,
25
+ "base_layer": 12,
26
+ "aligned_layer": 14
27
+ }