bpbradle commited on
Commit
bd26461
·
verified ·
1 Parent(s): 544aa35

upload dec/25M/100Mtok/postprefixlm_pct100.json

Browse files
dec/25M/100Mtok/postprefixlm_pct100.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "dec",
3
+ "arch": {
4
+ "dim": 448,
5
+ "num_encoder_layers": 8,
6
+ "num_decoder_layers": 4
7
+ },
8
+ "arch_label": "25M",
9
+ "non_emb_params": 28913024,
10
+ "pretrain_ckpt": "pct100.pt",
11
+ "sft_objective": "prefixlm",
12
+ "sft_train_loss": 4.557777759986488,
13
+ "sft_train_time_s": 40.39173698425293,
14
+ "sft_n_steps": 152,
15
+ "sft_mfu_pct": 0.9538706760557323,
16
+ "sft_tokens_target": 9994240,
17
+ "sft_tokens_actual": 9994240,
18
+ "sft_lr": 0.0002,
19
+ "sft_batch_size": 8,
20
+ "sft_grad_accum": 4,
21
+ "eval_loss_held_out": 4.704791829077394,
22
+ "eval_n_batches": 100,
23
+ "eval_n_valid_tokens": 1494358,
24
+ "eval_elapsed_s": 2.4015464782714844,
25
+ "wallclock_s": 49.830878257751465
26
+ }