Add files using upload-large-folder tool
Browse files- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_0/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_0/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_1/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_1/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_2/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_2/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_3/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_3/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_4/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_4/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_5/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_5/config.json +29 -0
gemma-2-2b_standard_new_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 sha256:243365f839b4182319013232b2475750ad92068bcd4467950c91383dc5590d32
|
3 |
+
size 1208232616
|
gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_0/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"trainer": {
|
3 |
+
"dict_class": "AutoEncoder",
|
4 |
+
"trainer_class": "StandardTrainerAprilUpdate",
|
5 |
+
"activation_dim": 2304,
|
6 |
+
"dict_size": 16384,
|
7 |
+
"lr": 0.0003,
|
8 |
+
"l1_penalty": 0.012,
|
9 |
+
"warmup_steps": 1000,
|
10 |
+
"sparsity_warmup_steps": 5000,
|
11 |
+
"steps": 244140,
|
12 |
+
"decay_start": 195312,
|
13 |
+
"seed": 0,
|
14 |
+
"device": "cuda:2",
|
15 |
+
"layer": 12,
|
16 |
+
"lm_name": "google/gemma-2-2b",
|
17 |
+
"wandb_name": "StandardTrainerNew-google/gemma-2-2b-resid_post_layer_12_trainer_0",
|
18 |
+
"submodule_name": "resid_post_layer_12"
|
19 |
+
},
|
20 |
+
"buffer": {
|
21 |
+
"d_submodule": 2304,
|
22 |
+
"io": "out",
|
23 |
+
"n_ctxs": 244,
|
24 |
+
"ctx_len": 1024,
|
25 |
+
"refresh_batch_size": 128,
|
26 |
+
"out_batch_size": 2048,
|
27 |
+
"device": "cuda:2"
|
28 |
+
}
|
29 |
+
}
|
gemma-2-2b_standard_new_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 sha256:57c34c16b2c40c0f5e02c00caf6cdc012e65681c77bb677634dcca519068a407
|
3 |
+
size 1208232616
|
gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_1/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"trainer": {
|
3 |
+
"dict_class": "AutoEncoder",
|
4 |
+
"trainer_class": "StandardTrainerAprilUpdate",
|
5 |
+
"activation_dim": 2304,
|
6 |
+
"dict_size": 16384,
|
7 |
+
"lr": 0.0003,
|
8 |
+
"l1_penalty": 0.015,
|
9 |
+
"warmup_steps": 1000,
|
10 |
+
"sparsity_warmup_steps": 5000,
|
11 |
+
"steps": 244140,
|
12 |
+
"decay_start": 195312,
|
13 |
+
"seed": 0,
|
14 |
+
"device": "cuda:2",
|
15 |
+
"layer": 12,
|
16 |
+
"lm_name": "google/gemma-2-2b",
|
17 |
+
"wandb_name": "StandardTrainerNew-google/gemma-2-2b-resid_post_layer_12_trainer_1",
|
18 |
+
"submodule_name": "resid_post_layer_12"
|
19 |
+
},
|
20 |
+
"buffer": {
|
21 |
+
"d_submodule": 2304,
|
22 |
+
"io": "out",
|
23 |
+
"n_ctxs": 244,
|
24 |
+
"ctx_len": 1024,
|
25 |
+
"refresh_batch_size": 128,
|
26 |
+
"out_batch_size": 2048,
|
27 |
+
"device": "cuda:2"
|
28 |
+
}
|
29 |
+
}
|
gemma-2-2b_standard_new_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 sha256:2500aa1f28537c0e3de0b9533a47f2665183816c7652130d1439fff270f3c284
|
3 |
+
size 1208232616
|
gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_2/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"trainer": {
|
3 |
+
"dict_class": "AutoEncoder",
|
4 |
+
"trainer_class": "StandardTrainerAprilUpdate",
|
5 |
+
"activation_dim": 2304,
|
6 |
+
"dict_size": 16384,
|
7 |
+
"lr": 0.0003,
|
8 |
+
"l1_penalty": 0.02,
|
9 |
+
"warmup_steps": 1000,
|
10 |
+
"sparsity_warmup_steps": 5000,
|
11 |
+
"steps": 244140,
|
12 |
+
"decay_start": 195312,
|
13 |
+
"seed": 0,
|
14 |
+
"device": "cuda:2",
|
15 |
+
"layer": 12,
|
16 |
+
"lm_name": "google/gemma-2-2b",
|
17 |
+
"wandb_name": "StandardTrainerNew-google/gemma-2-2b-resid_post_layer_12_trainer_2",
|
18 |
+
"submodule_name": "resid_post_layer_12"
|
19 |
+
},
|
20 |
+
"buffer": {
|
21 |
+
"d_submodule": 2304,
|
22 |
+
"io": "out",
|
23 |
+
"n_ctxs": 244,
|
24 |
+
"ctx_len": 1024,
|
25 |
+
"refresh_batch_size": 128,
|
26 |
+
"out_batch_size": 2048,
|
27 |
+
"device": "cuda:2"
|
28 |
+
}
|
29 |
+
}
|
gemma-2-2b_standard_new_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 sha256:8bb94b29cbf921038fea94ab85ca66c19f619f47c8139b9925e48d4e6480f6ac
|
3 |
+
size 1208232616
|
gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_3/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"trainer": {
|
3 |
+
"dict_class": "AutoEncoder",
|
4 |
+
"trainer_class": "StandardTrainerAprilUpdate",
|
5 |
+
"activation_dim": 2304,
|
6 |
+
"dict_size": 16384,
|
7 |
+
"lr": 0.0003,
|
8 |
+
"l1_penalty": 0.03,
|
9 |
+
"warmup_steps": 1000,
|
10 |
+
"sparsity_warmup_steps": 5000,
|
11 |
+
"steps": 244140,
|
12 |
+
"decay_start": 195312,
|
13 |
+
"seed": 0,
|
14 |
+
"device": "cuda:2",
|
15 |
+
"layer": 12,
|
16 |
+
"lm_name": "google/gemma-2-2b",
|
17 |
+
"wandb_name": "StandardTrainerNew-google/gemma-2-2b-resid_post_layer_12_trainer_3",
|
18 |
+
"submodule_name": "resid_post_layer_12"
|
19 |
+
},
|
20 |
+
"buffer": {
|
21 |
+
"d_submodule": 2304,
|
22 |
+
"io": "out",
|
23 |
+
"n_ctxs": 244,
|
24 |
+
"ctx_len": 1024,
|
25 |
+
"refresh_batch_size": 128,
|
26 |
+
"out_batch_size": 2048,
|
27 |
+
"device": "cuda:2"
|
28 |
+
}
|
29 |
+
}
|
gemma-2-2b_standard_new_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 sha256:00aa97c3a220aadb85a7e8fec705031c969280c9920250fd95ee34c6d331bf39
|
3 |
+
size 1208232616
|
gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_4/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"trainer": {
|
3 |
+
"dict_class": "AutoEncoder",
|
4 |
+
"trainer_class": "StandardTrainerAprilUpdate",
|
5 |
+
"activation_dim": 2304,
|
6 |
+
"dict_size": 16384,
|
7 |
+
"lr": 0.0003,
|
8 |
+
"l1_penalty": 0.04,
|
9 |
+
"warmup_steps": 1000,
|
10 |
+
"sparsity_warmup_steps": 5000,
|
11 |
+
"steps": 244140,
|
12 |
+
"decay_start": 195312,
|
13 |
+
"seed": 0,
|
14 |
+
"device": "cuda:2",
|
15 |
+
"layer": 12,
|
16 |
+
"lm_name": "google/gemma-2-2b",
|
17 |
+
"wandb_name": "StandardTrainerNew-google/gemma-2-2b-resid_post_layer_12_trainer_4",
|
18 |
+
"submodule_name": "resid_post_layer_12"
|
19 |
+
},
|
20 |
+
"buffer": {
|
21 |
+
"d_submodule": 2304,
|
22 |
+
"io": "out",
|
23 |
+
"n_ctxs": 244,
|
24 |
+
"ctx_len": 1024,
|
25 |
+
"refresh_batch_size": 128,
|
26 |
+
"out_batch_size": 2048,
|
27 |
+
"device": "cuda:2"
|
28 |
+
}
|
29 |
+
}
|
gemma-2-2b_standard_new_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 sha256:79506ebf3b1acb9871297dfa2c23b7892a0f58c41cb71a7e359eb39cd0a60b69
|
3 |
+
size 1208232616
|
gemma-2-2b_standard_new_width-2pow16_date-0107/resid_post_layer_12/trainer_5/config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"trainer": {
|
3 |
+
"dict_class": "AutoEncoder",
|
4 |
+
"trainer_class": "StandardTrainerAprilUpdate",
|
5 |
+
"activation_dim": 2304,
|
6 |
+
"dict_size": 16384,
|
7 |
+
"lr": 0.0003,
|
8 |
+
"l1_penalty": 0.06,
|
9 |
+
"warmup_steps": 1000,
|
10 |
+
"sparsity_warmup_steps": 5000,
|
11 |
+
"steps": 244140,
|
12 |
+
"decay_start": 195312,
|
13 |
+
"seed": 0,
|
14 |
+
"device": "cuda:2",
|
15 |
+
"layer": 12,
|
16 |
+
"lm_name": "google/gemma-2-2b",
|
17 |
+
"wandb_name": "StandardTrainerNew-google/gemma-2-2b-resid_post_layer_12_trainer_5",
|
18 |
+
"submodule_name": "resid_post_layer_12"
|
19 |
+
},
|
20 |
+
"buffer": {
|
21 |
+
"d_submodule": 2304,
|
22 |
+
"io": "out",
|
23 |
+
"n_ctxs": 244,
|
24 |
+
"ctx_len": 1024,
|
25 |
+
"refresh_batch_size": 128,
|
26 |
+
"out_batch_size": 2048,
|
27 |
+
"device": "cuda:2"
|
28 |
+
}
|
29 |
+
}
|