Hugging Face's logo

adamkarvonen
/
temp

Model card Files Community
adamkarvonen commited on
Commit
04904c9
·
verified ·
1 Parent(s): 08d00c9

Add files using upload-large-folder tool

Browse files
Files changed (10) hide show
  1. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_1/ae.pt +3 -0
  2. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_1/config.json +29 -0
  3. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_2/ae.pt +3 -0
  4. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_2/config.json +29 -0
  5. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_3/ae.pt +3 -0
  6. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_3/config.json +29 -0
  7. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_4/ae.pt +3 -0
  8. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_4/config.json +29 -0
  9. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_5/ae.pt +3 -0
  10. Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_5/config.json +29 -0
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_1/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0456e8f89a7d8e8b0c9f459266cd5abeca34e88aa0dfe25d99f115cc35bae4fd
3
+ size 100733608
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_1/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "dict_class": "AutoEncoder",
4
+ "trainer_class": "StandardTrainerAprilUpdate",
5
+ "activation_dim": 768,
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:0",
15
+ "layer": 8,
16
+ "lm_name": "EleutherAI/pythia-160m-deduped",
17
+ "wandb_name": "StandardTrainerNew-EleutherAI/pythia-160m-deduped-resid_post_layer_8_trainer_7",
18
+ "submodule_name": "resid_post_layer_8"
19
+ },
20
+ "buffer": {
21
+ "d_submodule": 768,
22
+ "io": "out",
23
+ "n_ctxs": 244,
24
+ "ctx_len": 1024,
25
+ "refresh_batch_size": 32,
26
+ "out_batch_size": 2048,
27
+ "device": "cuda:0"
28
+ }
29
+ }
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_2/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc63701f51a3105252744651a04ed847cb660b8e15816278c0e1f2237a31ff50
3
+ size 100733608
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_2/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "dict_class": "AutoEncoder",
4
+ "trainer_class": "StandardTrainerAprilUpdate",
5
+ "activation_dim": 768,
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:0",
15
+ "layer": 8,
16
+ "lm_name": "EleutherAI/pythia-160m-deduped",
17
+ "wandb_name": "StandardTrainerNew-EleutherAI/pythia-160m-deduped-resid_post_layer_8_trainer_8",
18
+ "submodule_name": "resid_post_layer_8"
19
+ },
20
+ "buffer": {
21
+ "d_submodule": 768,
22
+ "io": "out",
23
+ "n_ctxs": 244,
24
+ "ctx_len": 1024,
25
+ "refresh_batch_size": 32,
26
+ "out_batch_size": 2048,
27
+ "device": "cuda:0"
28
+ }
29
+ }
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_3/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67e8a04d84460f93f395ba4ff34e01c3326c43f5110610bcdae09b8e6b8fdcdd
3
+ size 100733608
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_3/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "dict_class": "AutoEncoder",
4
+ "trainer_class": "StandardTrainerAprilUpdate",
5
+ "activation_dim": 768,
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:0",
15
+ "layer": 8,
16
+ "lm_name": "EleutherAI/pythia-160m-deduped",
17
+ "wandb_name": "StandardTrainerNew-EleutherAI/pythia-160m-deduped-resid_post_layer_8_trainer_9",
18
+ "submodule_name": "resid_post_layer_8"
19
+ },
20
+ "buffer": {
21
+ "d_submodule": 768,
22
+ "io": "out",
23
+ "n_ctxs": 244,
24
+ "ctx_len": 1024,
25
+ "refresh_batch_size": 32,
26
+ "out_batch_size": 2048,
27
+ "device": "cuda:0"
28
+ }
29
+ }
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_4/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49031f0f47dd040414fc44026679814976a6d1d884fb8884348da9fad02bac7a
3
+ size 100733608
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_4/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "dict_class": "AutoEncoder",
4
+ "trainer_class": "StandardTrainerAprilUpdate",
5
+ "activation_dim": 768,
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:0",
15
+ "layer": 8,
16
+ "lm_name": "EleutherAI/pythia-160m-deduped",
17
+ "wandb_name": "StandardTrainerNew-EleutherAI/pythia-160m-deduped-resid_post_layer_8_trainer_10",
18
+ "submodule_name": "resid_post_layer_8"
19
+ },
20
+ "buffer": {
21
+ "d_submodule": 768,
22
+ "io": "out",
23
+ "n_ctxs": 244,
24
+ "ctx_len": 1024,
25
+ "refresh_batch_size": 32,
26
+ "out_batch_size": 2048,
27
+ "device": "cuda:0"
28
+ }
29
+ }
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_5/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7354597a3c4f34d60e64502165ca3b8df5550bc6522ca46681b9b4e75e9a1fe9
3
+ size 100733608
Standard_pythia-160m-deduped__0108-finetune_2pow14/resid_post_layer_8/trainer_5/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "dict_class": "AutoEncoder",
4
+ "trainer_class": "StandardTrainerAprilUpdate",
5
+ "activation_dim": 768,
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:0",
15
+ "layer": 8,
16
+ "lm_name": "EleutherAI/pythia-160m-deduped",
17
+ "wandb_name": "StandardTrainerNew-EleutherAI/pythia-160m-deduped-resid_post_layer_8_trainer_11",
18
+ "submodule_name": "resid_post_layer_8"
19
+ },
20
+ "buffer": {
21
+ "d_submodule": 768,
22
+ "io": "out",
23
+ "n_ctxs": 244,
24
+ "ctx_len": 1024,
25
+ "refresh_batch_size": 32,
26
+ "out_batch_size": 2048,
27
+ "device": "cuda:0"
28
+ }
29
+ }