Upload pretrained SAE model. Hook point: blocks.3.hook_mlp_out. Language Model Name: gpt2
Browse files
.gitattributes
CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
+
blocks.3.hook_mlp_out/sae_weights.safetensor filter=lfs diff=lfs merge=lfs -text
|
blocks.3.hook_mlp_out/hyperparams.json
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"hook_point_in": "blocks.3.hook_mlp_out",
|
3 |
+
"hook_point_out": "blocks.3.hook_mlp_out",
|
4 |
+
"use_decoder_bias": false,
|
5 |
+
"apply_decoder_bias_to_pre_encoder": true,
|
6 |
+
"decoder_bias_init_method": "geometric_median",
|
7 |
+
"expansion_factor": 32,
|
8 |
+
"d_model": 768,
|
9 |
+
"d_sae": 24576,
|
10 |
+
"norm_activation": "token-wise",
|
11 |
+
"decoder_exactly_unit_norm": false,
|
12 |
+
"use_glu_encoder": false,
|
13 |
+
"l1_coefficient": 0.00012,
|
14 |
+
"lp": 1,
|
15 |
+
"use_ghost_grads": true
|
16 |
+
}
|
blocks.3.hook_mlp_out/lm_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "gpt2",
|
3 |
+
"model_from_pretrained_path": null,
|
4 |
+
"cache_dir": null,
|
5 |
+
"d_model": 768,
|
6 |
+
"local_files_only": false
|
7 |
+
}
|
blocks.3.hook_mlp_out/sae_weights.safetensor
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:70a60affc950adf2d76fafeee437a4205b9292d2915792bdea5edcd96ec04475
|
3 |
+
size 151290320
|