Add files using upload-large-folder tool
Browse files- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_1/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_1/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_2/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_2/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_3/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_4/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_4/config.json +29 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_5/ae.pt +3 -0
- gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_5/config.json +29 -0
gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/resid_post_layer_12/trainer_1/ae.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5fd68698fbd20d33338660e5672e91956aa8fea450a06eff892ed374c49733a8
|
| 3 |
+
size 302066344
|
gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/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-2pow14_date-0107-finetune/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:b66f9787a60cf3a2cb46b2fe61c85bc011926d66c4fbbdcb728a99e928293559
|
| 3 |
+
size 302066344
|
gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/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-2pow14_date-0107-finetune/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-2pow14_date-0107-finetune/resid_post_layer_12/trainer_4/ae.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0bd9dec2f958efc6345454a238b21e4827afda0eddbc5a366271488e8f5de861
|
| 3 |
+
size 302066344
|
gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/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-2pow14_date-0107-finetune/resid_post_layer_12/trainer_5/ae.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:235a399a5f82616cc986db91792c1b722ec9509a823995c9e1175f611ccdf024
|
| 3 |
+
size 302066344
|
gemma-2-2b_standard_new_width-2pow14_date-0107-finetune/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 |
+
}
|