Spaces:
Running
Running
Update cache
Browse files- cache/models/TinyStories-1Layer-21M_ccb_attn_preproj_2023-08-18_23-08-47/cb_acts.pkl +1 -1
- cache/models/TinyStories-1Layer-21M_ccb_attn_preproj_2023-08-18_23-08-47/info.txt +1 -1
- cache/models/TinyStories-1Layer-21M_vcb_mlp_2023-08-19_02-15-56/info.txt +1 -1
- cache/models/TinyStories-33M_ccb_attn_preproj_2023-08-27_00-48-47/cb_acts.pkl +1 -1
- cache/models/TinyStories-33M_ccb_attn_preproj_2023-08-27_00-48-47/info.txt +1 -1
cache/models/TinyStories-1Layer-21M_ccb_attn_preproj_2023-08-18_23-08-47/cb_acts.pkl
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1048577327
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:99027fe6e75643df8e1ed3f295dc3b48713370717ada2169adfe4aafb4d932ce
|
3 |
size 1048577327
|
cache/models/TinyStories-1Layer-21M_ccb_attn_preproj_2023-08-18_23-08-47/info.txt
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
num_codes: 25000
|
2 |
cb_at: attn_preproj
|
3 |
-
|
4 |
n_layers: 1
|
5 |
n_heads: 16
|
6 |
model_name: roneneldan/TinyStories-1Layer-21M
|
|
|
1 |
num_codes: 25000
|
2 |
cb_at: attn_preproj
|
3 |
+
gcb: True
|
4 |
n_layers: 1
|
5 |
n_heads: 16
|
6 |
model_name: roneneldan/TinyStories-1Layer-21M
|
cache/models/TinyStories-1Layer-21M_vcb_mlp_2023-08-19_02-15-56/info.txt
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
num_codes: 50000
|
2 |
cb_at: mlp
|
3 |
-
|
4 |
n_layers: 1
|
5 |
n_heads: 16
|
6 |
model_name: roneneldan/TinyStories-1Layer-21M
|
|
|
1 |
num_codes: 50000
|
2 |
cb_at: mlp
|
3 |
+
gcb: False
|
4 |
n_layers: 1
|
5 |
n_heads: 16
|
6 |
model_name: roneneldan/TinyStories-1Layer-21M
|
cache/models/TinyStories-33M_ccb_attn_preproj_2023-08-27_00-48-47/cb_acts.pkl
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4194308945
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ba92eee6fae68ea426da5ca7be6a3d4d67e43a8b125f1910dd58f4266ff54c61
|
3 |
size 4194308945
|
cache/models/TinyStories-33M_ccb_attn_preproj_2023-08-27_00-48-47/info.txt
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
num_codes: 10000
|
2 |
cb_at: attn_preproj
|
3 |
-
|
4 |
n_layers: 4
|
5 |
n_heads: 16
|
6 |
model_name: roneneldan/TinyStories-33M
|
|
|
1 |
num_codes: 10000
|
2 |
cb_at: attn_preproj
|
3 |
+
gcb: True
|
4 |
n_layers: 4
|
5 |
n_heads: 16
|
6 |
model_name: roneneldan/TinyStories-33M
|