Upload folder using huggingface_hub
Browse files- 0-res-sm/cfg.json +21 -0
- 0-res-sm/sae_weights.safetensors +3 -0
- 1-res-sm/cfg.json +21 -0
- 1-res-sm/sae_weights.safetensors +3 -0
- 2-res-sm/cfg.json +21 -0
- 2-res-sm/sae_weights.safetensors +3 -0
- 3-res-sm/cfg.json +21 -0
- 3-res-sm/sae_weights.safetensors +3 -0
- 4-res-sm/cfg.json +21 -0
- 4-res-sm/sae_weights.safetensors +3 -0
- 5-res-sm/cfg.json +21 -0
- 5-res-sm/sae_weights.safetensors +3 -0
- e-res-sm/cfg.json +21 -0
- e-res-sm/sae_weights.safetensors +3 -0
0-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.0.hook_resid_post",
|
4 |
+
"hook_layer": 0,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
0-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1f92ebb7304fc976ae88816c2e65e8599c78deceac46c9b03703f4bb0ce3dba1
|
3 |
+
size 134351168
|
1-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.1.hook_resid_post",
|
4 |
+
"hook_layer": 1,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
1-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0e024597ddb0f7512fa07fe481f9a1356918bbad72675cca2792484f39cf815f
|
3 |
+
size 134351168
|
2-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.2.hook_resid_post",
|
4 |
+
"hook_layer": 2,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
2-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fad08101f96115b812dec4b0f7a2203fff69cb9e96fab6e62140d5db0bfee502
|
3 |
+
size 134351168
|
3-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.3.hook_resid_post",
|
4 |
+
"hook_layer": 3,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
3-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f883903b9e7dc0b340bfa16b3454a07b59f738d4ff66d8fabdeee8e9e0b662b0
|
3 |
+
size 134351168
|
4-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.4.hook_resid_post",
|
4 |
+
"hook_layer": 4,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
4-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:19959fd5a65cafb8f4f080464ccbfbbaea515da059e0745b389f24e6b682ab8c
|
3 |
+
size 134351168
|
5-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.5.hook_resid_post",
|
4 |
+
"hook_layer": 5,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
5-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0c01f32d7c454be8f195d8497db5966ee8856182341b8e2fcaec2f8e57608f79
|
3 |
+
size 134351168
|
e-res-sm/cfg.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "pythia-70m-deduped",
|
3 |
+
"hook_name": "blocks.0.hook_resid_pre",
|
4 |
+
"hook_layer": 0,
|
5 |
+
"hook_head_index": null,
|
6 |
+
"d_in": 512,
|
7 |
+
"d_sae": 32768,
|
8 |
+
"dtype": "torch.float32",
|
9 |
+
"device": "mps",
|
10 |
+
"dataset_path": "EleutherAI/the_pile_deduplicated",
|
11 |
+
"context_size": 128,
|
12 |
+
"architecture": "standard",
|
13 |
+
"activation_fn_str": "relu",
|
14 |
+
"activation_fn_kwargs": null,
|
15 |
+
"apply_b_dec_to_input": true,
|
16 |
+
"finetuning_scaling_factor": false,
|
17 |
+
"sae_lens_training_version": null,
|
18 |
+
"prepend_bos": false,
|
19 |
+
"dataset_trust_remote_code": true,
|
20 |
+
"normalize_activations": "none"
|
21 |
+
}
|
e-res-sm/sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f48bc4ab5266fbd1a9204356ae9a832d42c77785d459213be15cc4f713322800
|
3 |
+
size 134351168
|