Hugging Face's logo

ctigges
/
pythia-70m-deduped__res-sm_processed

Model card Files Community
ctigges commited on
Commit
4de021e
·
verified ·
1 Parent(s): bb155ab

Upload folder using huggingface_hub

Browse files
Files changed (14) hide show
  1. 0-res-sm/cfg.json +1 -21
  2. 0-res-sm/sparsity.safetensors +3 -0
  3. 1-res-sm/cfg.json +1 -21
  4. 1-res-sm/sparsity.safetensors +3 -0
  5. 2-res-sm/cfg.json +1 -21
  6. 2-res-sm/sparsity.safetensors +3 -0
  7. 3-res-sm/cfg.json +1 -21
  8. 3-res-sm/sparsity.safetensors +3 -0
  9. 4-res-sm/cfg.json +1 -21
  10. 4-res-sm/sparsity.safetensors +3 -0
  11. 5-res-sm/cfg.json +1 -21
  12. 5-res-sm/sparsity.safetensors +3 -0
  13. e-res-sm/cfg.json +1 -21
  14. e-res-sm/sparsity.safetensors +3 -0
0-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
- }
 
1
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.0.hook_resid_post", "hook_layer": 0, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbb5e02251657a3d3176d1059be408d40743923280b21349e281e6607eef9ac2
3
+ size 131152
1-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.1.hook_resid_post", "hook_layer": 1, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab4d2869a4215fb21c2232f086f2447f520ed553b7f4a38c2688cd5317893ca7
3
+ size 131152
2-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
- }
 
1
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.2.hook_resid_post", "hook_layer": 2, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5eb79aac97f83d2de383d4cd2d701fea9077c7dee85a6e8668d323d23918cf2b
3
+ size 131152
3-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
- }
 
1
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.3.hook_resid_post", "hook_layer": 3, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5992a0d5a944b0090147a083bfd53d802e1757708644d19378122f3e2ef67878
3
+ size 131152
4-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
- }
 
1
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.4.hook_resid_post", "hook_layer": 4, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3de19ed9e6a674abe9afd61ef2851926212cb87105241ec32409ced3ed4676dc
3
+ size 131152
5-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
- }
 
1
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.5.hook_resid_post", "hook_layer": 5, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94356d7e0bc1268bcbf3fb432f2bf5a3be98187b81069b65b83652834163d393
3
+ size 131152
e-res-sm/cfg.json CHANGED
@@ -1,21 +1 @@
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
- }
 
1
+ {"architecture": "standard", "d_in": 512, "d_sae": 32768, "dtype": "torch.float32", "device": "cuda", "model_name": "pythia-70m-deduped", "hook_name": "blocks.0.hook_resid_pre", "hook_layer": 0, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": null, "prepend_bos": false, "dataset_path": "EleutherAI/the_pile_deduplicated", "dataset_trust_remote_code": true, "context_size": 128, "normalize_activations": "none"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e-res-sm/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91075cc05f2b2962ab89ac30ba55f9deb606a79fe50c67301fa82dfad86e5e8c
3
+ size 131152