Hugging Face's logo

canrager
/
lm_sae

Model card Files Community
canrager commited on
Commit
f553951
•
1 Parent(s): 169f303

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_0/config.json +2 -1
  2. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_1/config.json +2 -1
  3. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_2/config.json +2 -1
  4. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_3/config.json +2 -1
  5. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_4/config.json +2 -1
  6. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_5/config.json +2 -1
  7. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_0/config.json +2 -1
  8. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_1/config.json +2 -1
  9. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_2/config.json +2 -1
  10. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_3/config.json +2 -1
  11. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_4/config.json +2 -1
  12. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_5/config.json +2 -1
  13. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_0/config.json +2 -1
  14. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_1/config.json +2 -1
  15. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_2/config.json +2 -1
  16. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_3/config.json +2 -1
  17. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_4/config.json +2 -1
  18. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_5/config.json +2 -1
  19. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_0/config.json +2 -1
  20. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_1/config.json +2 -1
  21. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_2/config.json +2 -1
  22. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_3/config.json +2 -1
  23. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_4/config.json +2 -1
  24. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_5/config.json +2 -1
  25. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_0/config.json +2 -1
  26. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_1/config.json +2 -1
  27. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_2/config.json +2 -1
  28. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_3/config.json +2 -1
  29. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_4/config.json +2 -1
  30. gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_5/config.json +2 -1
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_0/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
- "submodule_name": "resid_post_layer_11"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
+ "submodule_name": "resid_post_layer_11",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_1/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
- "submodule_name": "resid_post_layer_11"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
+ "submodule_name": "resid_post_layer_11",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_2/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
- "submodule_name": "resid_post_layer_11"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
+ "submodule_name": "resid_post_layer_11",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_3/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
- "submodule_name": "resid_post_layer_11"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
+ "submodule_name": "resid_post_layer_11",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_4/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
- "submodule_name": "resid_post_layer_11"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
+ "submodule_name": "resid_post_layer_11",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_11/trainer_5/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
- "submodule_name": "resid_post_layer_11"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 11,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_11",
15
+ "submodule_name": "resid_post_layer_11",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_0/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
- "submodule_name": "resid_post_layer_15"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
+ "submodule_name": "resid_post_layer_15",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_1/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
- "submodule_name": "resid_post_layer_15"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
+ "submodule_name": "resid_post_layer_15",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_2/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
- "submodule_name": "resid_post_layer_15"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
+ "submodule_name": "resid_post_layer_15",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_3/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
- "submodule_name": "resid_post_layer_15"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
+ "submodule_name": "resid_post_layer_15",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_4/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
- "submodule_name": "resid_post_layer_15"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
+ "submodule_name": "resid_post_layer_15",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_15/trainer_5/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
- "submodule_name": "resid_post_layer_15"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 15,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_15",
15
+ "submodule_name": "resid_post_layer_15",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_0/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
- "submodule_name": "resid_post_layer_19"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
+ "submodule_name": "resid_post_layer_19",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_1/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
- "submodule_name": "resid_post_layer_19"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
+ "submodule_name": "resid_post_layer_19",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_2/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
- "submodule_name": "resid_post_layer_19"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
+ "submodule_name": "resid_post_layer_19",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_3/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
- "submodule_name": "resid_post_layer_19"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
+ "submodule_name": "resid_post_layer_19",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_4/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
- "submodule_name": "resid_post_layer_19"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
+ "submodule_name": "resid_post_layer_19",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_19/trainer_5/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
- "submodule_name": "resid_post_layer_19"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 19,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_19",
15
+ "submodule_name": "resid_post_layer_19",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_0/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
- "submodule_name": "resid_post_layer_3"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
+ "submodule_name": "resid_post_layer_3",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_1/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
- "submodule_name": "resid_post_layer_3"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
+ "submodule_name": "resid_post_layer_3",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_2/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
- "submodule_name": "resid_post_layer_3"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
+ "submodule_name": "resid_post_layer_3",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_3/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
- "submodule_name": "resid_post_layer_3"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
+ "submodule_name": "resid_post_layer_3",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_4/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
- "submodule_name": "resid_post_layer_3"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
+ "submodule_name": "resid_post_layer_3",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_3/trainer_5/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
- "submodule_name": "resid_post_layer_3"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 3,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_3",
15
+ "submodule_name": "resid_post_layer_3",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_0/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
- "submodule_name": "resid_post_layer_7"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
+ "submodule_name": "resid_post_layer_7",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_1/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
- "submodule_name": "resid_post_layer_7"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
+ "submodule_name": "resid_post_layer_7",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_2/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
- "submodule_name": "resid_post_layer_7"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
+ "submodule_name": "resid_post_layer_7",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_3/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
- "submodule_name": "resid_post_layer_7"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
+ "submodule_name": "resid_post_layer_7",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_4/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
- "submodule_name": "resid_post_layer_7"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
+ "submodule_name": "resid_post_layer_7",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,
gemma-2-2b_sweep_standard_ctx128_ef8_0824/resid_post_layer_7/trainer_5/config.json CHANGED
@@ -12,7 +12,8 @@
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
- "submodule_name": "resid_post_layer_7"
 
16
  },
17
  "buffer": {
18
  "d_submodule": 2304,
 
12
  "layer": 7,
13
  "lm_name": "google/gemma-2-2b",
14
  "wandb_name": "StandardTrainer-google/gemma-2-2b-resid_post_layer_7",
15
+ "submodule_name": "resid_post_layer_7",
16
+ "steps": "48828"
17
  },
18
  "buffer": {
19
  "d_submodule": 2304,