Hugging Face's logo

adamkarvonen
/
saebench_gemma-2-2b_width-2pow12_date-0108

Model card Files Community
adamkarvonen commited on
Commit
1ca8ce1
·
verified ·
1 Parent(s): 98acdfe

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_10/ae.pt +3 -0
  2. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_10/config.json +32 -0
  3. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_10/eval_results.json +1 -0
  4. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_11/ae.pt +3 -0
  5. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_11/config.json +32 -0
  6. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_11/eval_results.json +1 -0
  7. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_6/ae.pt +3 -0
  8. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_6/config.json +32 -0
  9. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_6/eval_results.json +1 -0
  10. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_7/ae.pt +3 -0
  11. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_7/config.json +32 -0
  12. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_7/eval_results.json +1 -0
  13. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_8/ae.pt +3 -0
  14. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_8/config.json +32 -0
  15. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_8/eval_results.json +1 -0
  16. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_9/ae.pt +3 -0
  17. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_9/config.json +32 -0
  18. BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_9/eval_results.json +1 -0
  19. GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_0/ae.pt +3 -0
  20. GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_1/ae.pt +3 -0
  21. GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_2/ae.pt +3 -0
  22. GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_3/ae.pt +3 -0
  23. GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt +3 -0
  24. GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt +3 -0
  25. JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_0/ae.pt +3 -0
  26. JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_1/ae.pt +3 -0
  27. JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_2/ae.pt +3 -0
  28. JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_3/ae.pt +3 -0
  29. JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt +3 -0
  30. JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt +3 -0
  31. PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_0/ae.pt +3 -0
  32. PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_1/ae.pt +3 -0
  33. PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_2/ae.pt +3 -0
  34. PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_3/ae.pt +3 -0
  35. PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt +3 -0
  36. PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt +3 -0
  37. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_0/ae.pt +3 -0
  38. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_1/ae.pt +3 -0
  39. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_2/ae.pt +3 -0
  40. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_3/ae.pt +3 -0
  41. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt +3 -0
  42. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt +3 -0
  43. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_0/ae.pt +3 -0
  44. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_122/ae.pt +3 -0
  45. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_244/ae.pt +3 -0
  46. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_2441/ae.pt +3 -0
  47. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_24414/ae.pt +3 -0
  48. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_386/ae.pt +3 -0
  49. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_772/ae.pt +3 -0
  50. StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_7720/ae.pt +3 -0
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_10/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7e56dca54623be6b87ef79fa7675fa038455848308fa0bf7b3615c24f868732
3
+ size 75525142
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_10/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "BatchTopKTrainer",
4
+ "dict_class": "BatchTopKSAE",
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
+ "top_k_aux": 1152,
13
+ "seed": 0,
14
+ "activation_dim": 2304,
15
+ "dict_size": 4096,
16
+ "k": 320,
17
+ "device": "cuda:0",
18
+ "layer": 12,
19
+ "lm_name": "google/gemma-2-2b",
20
+ "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_10",
21
+ "submodule_name": "resid_post_layer_12"
22
+ },
23
+ "buffer": {
24
+ "d_submodule": 2304,
25
+ "io": "out",
26
+ "n_ctxs": 244,
27
+ "ctx_len": 1024,
28
+ "refresh_batch_size": 4,
29
+ "out_batch_size": 2048,
30
+ "device": "cuda:0"
31
+ }
32
+ }
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_10/eval_results.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"l2_loss": 48.85125, "l1_loss": 1669.4, "l0": 321.7892822265625, "frac_variance_explained": 0.8830859375, "cossim": 0.93751953125, "l2_ratio": 0.9375, "relative_reconstruction_bias": 1.0002734375, "loss_original": 2.152919921875, "loss_reconstructed": 2.195830078125, "loss_zero": 12.4375, "frac_recovered": 0.99533203125, "frac_alive": 0.999755859375, "hyperparameters": {"n_inputs": 200, "context_length": 1024}}
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_11/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:788072b25d7b0eee3e51ea5615ef25aaca4482e9af5f3ea33b62e4e743960335
3
+ size 75525142
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_11/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "BatchTopKTrainer",
4
+ "dict_class": "BatchTopKSAE",
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
+ "top_k_aux": 1152,
13
+ "seed": 0,
14
+ "activation_dim": 2304,
15
+ "dict_size": 4096,
16
+ "k": 640,
17
+ "device": "cuda:0",
18
+ "layer": 12,
19
+ "lm_name": "google/gemma-2-2b",
20
+ "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_11",
21
+ "submodule_name": "resid_post_layer_12"
22
+ },
23
+ "buffer": {
24
+ "d_submodule": 2304,
25
+ "io": "out",
26
+ "n_ctxs": 244,
27
+ "ctx_len": 1024,
28
+ "refresh_batch_size": 4,
29
+ "out_batch_size": 2048,
30
+ "device": "cuda:0"
31
+ }
32
+ }
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_11/eval_results.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"l2_loss": 38.55, "l1_loss": 2861.84, "l0": 643.082568359375, "frac_variance_explained": 0.9279296875, "cossim": 0.9608984375, "l2_ratio": 0.96087890625, "relative_reconstruction_bias": 1.00046875, "loss_original": 2.152919921875, "loss_reconstructed": 2.16953125, "loss_zero": 12.4375, "frac_recovered": 0.99826171875, "frac_alive": 0.971435546875, "hyperparameters": {"n_inputs": 200, "context_length": 1024}}
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_6/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd123d1f222b07a7153d5fc3871b40c79944d6a21592a5aaf18cdf6448ea082f
3
+ size 75525142
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_6/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "BatchTopKTrainer",
4
+ "dict_class": "BatchTopKSAE",
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
+ "top_k_aux": 1152,
13
+ "seed": 0,
14
+ "activation_dim": 2304,
15
+ "dict_size": 4096,
16
+ "k": 20,
17
+ "device": "cuda:0",
18
+ "layer": 12,
19
+ "lm_name": "google/gemma-2-2b",
20
+ "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_6",
21
+ "submodule_name": "resid_post_layer_12"
22
+ },
23
+ "buffer": {
24
+ "d_submodule": 2304,
25
+ "io": "out",
26
+ "n_ctxs": 244,
27
+ "ctx_len": 1024,
28
+ "refresh_batch_size": 4,
29
+ "out_batch_size": 2048,
30
+ "device": "cuda:0"
31
+ }
32
+ }
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_6/eval_results.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"l2_loss": 71.1775, "l1_loss": 287.09, "l0": 19.922880859375, "frac_variance_explained": 0.74666015625, "cossim": 0.865703125, "l2_ratio": 0.8649609375, "relative_reconstruction_bias": 1.0004296875, "loss_original": 2.152919921875, "loss_reconstructed": 2.55654296875, "loss_zero": 12.4375, "frac_recovered": 0.9603515625, "frac_alive": 1.0, "hyperparameters": {"n_inputs": 200, "context_length": 1024}}
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_7/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4040775dd6a9059c152c05d266fe006157b31297152227a072b4d51408c53809
3
+ size 75525142
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_7/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "BatchTopKTrainer",
4
+ "dict_class": "BatchTopKSAE",
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
+ "top_k_aux": 1152,
13
+ "seed": 0,
14
+ "activation_dim": 2304,
15
+ "dict_size": 4096,
16
+ "k": 40,
17
+ "device": "cuda:0",
18
+ "layer": 12,
19
+ "lm_name": "google/gemma-2-2b",
20
+ "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_7",
21
+ "submodule_name": "resid_post_layer_12"
22
+ },
23
+ "buffer": {
24
+ "d_submodule": 2304,
25
+ "io": "out",
26
+ "n_ctxs": 244,
27
+ "ctx_len": 1024,
28
+ "refresh_batch_size": 4,
29
+ "out_batch_size": 2048,
30
+ "device": "cuda:0"
31
+ }
32
+ }
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_7/eval_results.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"l2_loss": 66.54, "l1_loss": 402.34, "l0": 39.944716796875, "frac_variance_explained": 0.77919921875, "cossim": 0.88330078125, "l2_ratio": 0.88310546875, "relative_reconstruction_bias": 1.00033203125, "loss_original": 2.152919921875, "loss_reconstructed": 2.396982421875, "loss_zero": 12.4375, "frac_recovered": 0.97587890625, "frac_alive": 1.0, "hyperparameters": {"n_inputs": 200, "context_length": 1024}}
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_8/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e609d9ad8b04890412881b51171fd7f30e6f7aeb781bf68d1037ace553e8ff3
3
+ size 75525142
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_8/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "BatchTopKTrainer",
4
+ "dict_class": "BatchTopKSAE",
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
+ "top_k_aux": 1152,
13
+ "seed": 0,
14
+ "activation_dim": 2304,
15
+ "dict_size": 4096,
16
+ "k": 80,
17
+ "device": "cuda:0",
18
+ "layer": 12,
19
+ "lm_name": "google/gemma-2-2b",
20
+ "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_8",
21
+ "submodule_name": "resid_post_layer_12"
22
+ },
23
+ "buffer": {
24
+ "d_submodule": 2304,
25
+ "io": "out",
26
+ "n_ctxs": 244,
27
+ "ctx_len": 1024,
28
+ "refresh_batch_size": 4,
29
+ "out_batch_size": 2048,
30
+ "device": "cuda:0"
31
+ }
32
+ }
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_8/eval_results.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"l2_loss": 61.62375, "l1_loss": 577.64, "l0": 79.8656640625, "frac_variance_explained": 0.81333984375, "cossim": 0.90095703125, "l2_ratio": 0.9005859375, "relative_reconstruction_bias": 1.0005859375, "loss_original": 2.152919921875, "loss_reconstructed": 2.291015625, "loss_zero": 12.4375, "frac_recovered": 0.98591796875, "frac_alive": 1.0, "hyperparameters": {"n_inputs": 200, "context_length": 1024}}
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_9/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a241b985f09e2d3f5541cb3d127950ccc9d6237f533c9e8d107799af1259afa
3
+ size 75525142
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_9/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trainer": {
3
+ "trainer_class": "BatchTopKTrainer",
4
+ "dict_class": "BatchTopKSAE",
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
+ "top_k_aux": 1152,
13
+ "seed": 0,
14
+ "activation_dim": 2304,
15
+ "dict_size": 4096,
16
+ "k": 160,
17
+ "device": "cuda:0",
18
+ "layer": 12,
19
+ "lm_name": "google/gemma-2-2b",
20
+ "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_9",
21
+ "submodule_name": "resid_post_layer_12"
22
+ },
23
+ "buffer": {
24
+ "d_submodule": 2304,
25
+ "io": "out",
26
+ "n_ctxs": 244,
27
+ "ctx_len": 1024,
28
+ "refresh_batch_size": 4,
29
+ "out_batch_size": 2048,
30
+ "device": "cuda:0"
31
+ }
32
+ }
BatchTopKTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_9/eval_results.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"l2_loss": 56.02125, "l1_loss": 883.4, "l0": 159.4246337890625, "frac_variance_explained": 0.847890625, "cossim": 0.91810546875, "l2_ratio": 0.91796875, "relative_reconstruction_bias": 1.00001953125, "loss_original": 2.152919921875, "loss_reconstructed": 2.2319140625, "loss_zero": 12.4375, "frac_recovered": 0.99177734375, "frac_alive": 1.0, "hyperparameters": {"n_inputs": 200, "context_length": 1024}}
GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/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:ba233fc8ba9e4e02fdda1fdd6409da877dfccaa586204e8838f16ab284bbf405
3
+ size 75557910
GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_1/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:887ed43cda633a2935ae7e8a93690b024b4db61eba86cd31f4860177fb7d74bf
3
+ size 75557910
GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/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:d3a1ccf2baa4e2b0beb1067dac78261662b1cae85c812b93e6ca52609a3f8c46
3
+ size 75557910
GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/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:bf6049a877d63d090059682d7f31d4f9abeb567e8363adb2c8f1c0947b0fcdc4
3
+ size 75557910
GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09f96ba30e4bf8f6b590c69e87dd7b8ae0d7be6e280eee717a9bdaca691fa3ec
3
+ size 75557910
GatedSAETrainer_google_gemma-2-2b_ctx1024_0108/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:c8f6ce284848acb59d9229414d93493a81ce64222fd7ecb3d0ca65c5c7986ba6
3
+ size 75557910
JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_0/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a17abc5a1e2602b158758d8fda172bb308618a2b29a82b9e429f41306532e81
3
+ size 75541279
JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_1/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bda0e5d0830cd179667d094b99d53dd44c49830f942806414a37426f0df7622
3
+ size 75541279
JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/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:e79b0318df990fbc84587edd10612be6b8cbb84638e3bc4cb5780c31833ea8ea
3
+ size 75541279
JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_3/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:004eb54008f8506586a93893c6338e054e297ec3c2ff20abecb80cb98fd5a164
3
+ size 75541279
JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:189175ea1c34fe55de1b96480590e8b3dfb79963ec561f961efb0b5892774997
3
+ size 75541279
JumpReluTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:522fd8e36dab064fbfa848f4e0d36ad7ee463b95382eba68282a851b80211ff2
3
+ size 75541279
PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/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:f2f39f941791308e5a38407cc0e6dd9b0d4cb9e51a234a789db6fd43a43f2e7c
3
+ size 75524776
PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/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:cc236b61b19d327d6f8afd38662f28d89a459dfb6754f22f62080767b57f61ad
3
+ size 75524776
PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/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:e2f120429a8042890ad1810ef5a86d8500c3ba738959a200aeda13910a074d12
3
+ size 75524776
PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_3/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69c315b36b04096117a57d910139c9f64be2e28ee15a62be163c7b897bf6768b
3
+ size 75524776
PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/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:e97e13f77ae26ca96a6fbf2ded67661da894acaff5fd9b81869e740fe3d3b10b
3
+ size 75524776
PAnnealTrainer_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2347792806f7bdb1be9925e34fc43abbeb1a3de7bc2b3451fc8fea16db797af5
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_0/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7cfc0683229ea8a10530761fe8924042b2e169afda357e89cf195e5e4c230813
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/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:f4410a2aefd77dcf7c9e3ccb2c3860712fbc5f4c05c0ab0038f3e0ffd01e7161
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_2/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76009c54e89f0e0cd6a8cdd3dbb96dbefdf2203d67c1493260e812c6ae0c8687
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/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:d2f89b968bf24a929123d9eb876bc683d5e127465008912eec2eaca26f3b75c1
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_4/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c58ab3f9c6023a99832bd9d44e1af6e2975d16014f6b9c1e3d24b7ee7132036
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12/trainer_5/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ed200814789373a94bcce1b3d95b8d98f79041d726910f972669b0b46201b50
3
+ size 75524776
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_0/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bbd6a47e1b24bf9267b3a3e38b8055ea373a50f45e1ef446742faf65987454f0
3
+ size 75524792
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_122/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bcab681719adf3e2bd225cc03a7ef8fde7053d447eda80e7807396f4b24b1d5b
3
+ size 75524808
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_244/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57f38fd757927e096ab982173169215966c60fe84bba113a877e6696c2bd4e1d
3
+ size 75524808
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_2441/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c634d8a50c960690b2191e8a161322b8561dc168a60f8afd34b2f3c7449f8dd6
3
+ size 75524880
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_24414/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:897dad9d76414b44a70f7fdabca7e56036956fdac36ffe4f1208ccb79edec791
3
+ size 75525080
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_386/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42ba721d3e5bf5d480d3455043773b67f890ccc6e194614bfce824a5f8c39b43
3
+ size 75524808
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_772/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cc92e67df96094a7037230ef7427e8d077972707e7060b396e6585e558e9bea
3
+ size 75524808
StandardTrainerAprilUpdate_google_gemma-2-2b_ctx1024_0108/resid_post_layer_12_checkpoints/trainer_0_step_7720/ae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1686c1e40fd958abd1357111c7ba1f158cc0e21fbe0c79c6ada356d697cb2196
3
+ size 75524880