adamkarvonen commited on
Commit
8308b9e
·
verified ·
1 Parent(s): 420cb0a

Add files using upload-large-folder tool

Browse files
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_0/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98c239e931c5717251468c524b1a9cf159aa78b0ad8cfcdcf666f0de383e5ca7
3
+ size 1208232982
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_0/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "TopKTrainer",
4
+ "dict_class": "AutoEncoderTopK",
5
+ "lr": 0.0003,
6
+ "steps": 244140,
7
+ "auxk_alpha": 0.03125,
8
+ "warmup_steps": 1000,
9
+ "decay_start": 195312,
10
+ "threshold_beta": 0.999,
11
+ "threshold_start_step": 1000,
12
+ "seed": 0,
13
+ "activation_dim": 2304,
14
+ "dict_size": 16384,
15
+ "k": 20,
16
+ "device": "cuda:0",
17
+ "layer": 12,
18
+ "lm_name": "google/gemma-2-2b",
19
+ "wandb_name": "TopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_0",
20
+ "submodule_name": "resid_post_layer_12"
21
+ },
22
+ "buffer": {
23
+ "d_submodule": 2304,
24
+ "io": "out",
25
+ "n_ctxs": 244,
26
+ "ctx_len": 1024,
27
+ "refresh_batch_size": 4,
28
+ "out_batch_size": 2048,
29
+ "device": "cuda:0"
30
+ }
31
+ }
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_1/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c39c2db6962ede49eee6cf940f2760c6206d1a86048ef6fe5dc4b8012f4a718
3
+ size 1208232982
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_1/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "TopKTrainer",
4
+ "dict_class": "AutoEncoderTopK",
5
+ "lr": 0.0003,
6
+ "steps": 244140,
7
+ "auxk_alpha": 0.03125,
8
+ "warmup_steps": 1000,
9
+ "decay_start": 195312,
10
+ "threshold_beta": 0.999,
11
+ "threshold_start_step": 1000,
12
+ "seed": 0,
13
+ "activation_dim": 2304,
14
+ "dict_size": 16384,
15
+ "k": 40,
16
+ "device": "cuda:0",
17
+ "layer": 12,
18
+ "lm_name": "google/gemma-2-2b",
19
+ "wandb_name": "TopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_1",
20
+ "submodule_name": "resid_post_layer_12"
21
+ },
22
+ "buffer": {
23
+ "d_submodule": 2304,
24
+ "io": "out",
25
+ "n_ctxs": 244,
26
+ "ctx_len": 1024,
27
+ "refresh_batch_size": 4,
28
+ "out_batch_size": 2048,
29
+ "device": "cuda:0"
30
+ }
31
+ }
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_2/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f340937a88c08fb8f4000fdd60566869b37542e134dac5e5b26e83eaf62f7a46
3
+ size 1208232982
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_2/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "TopKTrainer",
4
+ "dict_class": "AutoEncoderTopK",
5
+ "lr": 0.0003,
6
+ "steps": 244140,
7
+ "auxk_alpha": 0.03125,
8
+ "warmup_steps": 1000,
9
+ "decay_start": 195312,
10
+ "threshold_beta": 0.999,
11
+ "threshold_start_step": 1000,
12
+ "seed": 0,
13
+ "activation_dim": 2304,
14
+ "dict_size": 16384,
15
+ "k": 80,
16
+ "device": "cuda:0",
17
+ "layer": 12,
18
+ "lm_name": "google/gemma-2-2b",
19
+ "wandb_name": "TopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_2",
20
+ "submodule_name": "resid_post_layer_12"
21
+ },
22
+ "buffer": {
23
+ "d_submodule": 2304,
24
+ "io": "out",
25
+ "n_ctxs": 244,
26
+ "ctx_len": 1024,
27
+ "refresh_batch_size": 4,
28
+ "out_batch_size": 2048,
29
+ "device": "cuda:0"
30
+ }
31
+ }
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_3/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:463730c65d234709cde032f931e5fad382b63db738c1dc3e4ac231f45a0883b1
3
+ size 1208232982
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_3/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "TopKTrainer",
4
+ "dict_class": "AutoEncoderTopK",
5
+ "lr": 0.0003,
6
+ "steps": 244140,
7
+ "auxk_alpha": 0.03125,
8
+ "warmup_steps": 1000,
9
+ "decay_start": 195312,
10
+ "threshold_beta": 0.999,
11
+ "threshold_start_step": 1000,
12
+ "seed": 0,
13
+ "activation_dim": 2304,
14
+ "dict_size": 16384,
15
+ "k": 160,
16
+ "device": "cuda:0",
17
+ "layer": 12,
18
+ "lm_name": "google/gemma-2-2b",
19
+ "wandb_name": "TopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_3",
20
+ "submodule_name": "resid_post_layer_12"
21
+ },
22
+ "buffer": {
23
+ "d_submodule": 2304,
24
+ "io": "out",
25
+ "n_ctxs": 244,
26
+ "ctx_len": 1024,
27
+ "refresh_batch_size": 4,
28
+ "out_batch_size": 2048,
29
+ "device": "cuda:0"
30
+ }
31
+ }
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_4/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:adb685c97d08793593f01a139c140b9988c65a6fa86219b61a5c13f1e832f53d
3
+ size 1208232982
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_4/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "TopKTrainer",
4
+ "dict_class": "AutoEncoderTopK",
5
+ "lr": 0.0003,
6
+ "steps": 244140,
7
+ "auxk_alpha": 0.03125,
8
+ "warmup_steps": 1000,
9
+ "decay_start": 195312,
10
+ "threshold_beta": 0.999,
11
+ "threshold_start_step": 1000,
12
+ "seed": 0,
13
+ "activation_dim": 2304,
14
+ "dict_size": 16384,
15
+ "k": 320,
16
+ "device": "cuda:0",
17
+ "layer": 12,
18
+ "lm_name": "google/gemma-2-2b",
19
+ "wandb_name": "TopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_4",
20
+ "submodule_name": "resid_post_layer_12"
21
+ },
22
+ "buffer": {
23
+ "d_submodule": 2304,
24
+ "io": "out",
25
+ "n_ctxs": 244,
26
+ "ctx_len": 1024,
27
+ "refresh_batch_size": 4,
28
+ "out_batch_size": 2048,
29
+ "device": "cuda:0"
30
+ }
31
+ }
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_5/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62cd93a18e9618a69ff4415cb00d61ac2ba869f014bdc10c3e4075d480cda4e3
3
+ size 1208232982
gemma-2-2b_top_k_width-2pow16_date-0107/resid_post_layer_12/trainer_5/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "TopKTrainer",
4
+ "dict_class": "AutoEncoderTopK",
5
+ "lr": 0.0003,
6
+ "steps": 244140,
7
+ "auxk_alpha": 0.03125,
8
+ "warmup_steps": 1000,
9
+ "decay_start": 195312,
10
+ "threshold_beta": 0.999,
11
+ "threshold_start_step": 1000,
12
+ "seed": 0,
13
+ "activation_dim": 2304,
14
+ "dict_size": 16384,
15
+ "k": 640,
16
+ "device": "cuda:0",
17
+ "layer": 12,
18
+ "lm_name": "google/gemma-2-2b",
19
+ "wandb_name": "TopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_5",
20
+ "submodule_name": "resid_post_layer_12"
21
+ },
22
+ "buffer": {
23
+ "d_submodule": 2304,
24
+ "io": "out",
25
+ "n_ctxs": 244,
26
+ "ctx_len": 1024,
27
+ "refresh_batch_size": 4,
28
+ "out_batch_size": 2048,
29
+ "device": "cuda:0"
30
+ }
31
+ }