kacperwikiel commited on
Commit
a4d0864
·
verified ·
1 Parent(s): e37bde3

Update experiment measurements and status

Browse files
reports/configs/cosmopedia-1337.toml ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [model]
2
+ vocab = 12288
3
+ n_layer = 19
4
+ n_embd = 768
5
+ n_head = 12
6
+ block = 1024
7
+ ffn_mult = 2.8125
8
+ [train]
9
+ seed = 1337
10
+ micro_batch = 64
11
+ global_batch = 512
12
+ steps = 1000
13
+ warmup = 50
14
+ lr = 0.0001
15
+ min_lr = 0.00001
16
+ muon_lr = 0.0033333333333333335
17
+ checkpoint_every = 500
18
+ validate_every = 500
19
+ log_every = 20
20
+ compile = true
21
+ [data]
22
+ train = "/data/ablation/mixtures/cosmopedia/train.bin"
23
+ validation = "/data/final/val.bin"
24
+ sha256 = "c3ae1aa5444b2965cb1d31c4225845ae8bc491b22be8b37d13c913cb6c067e60"
reports/eval/control-1337-1000.json ADDED
The diff for this file is too large to render. See raw diff