Defalt-404 commited on
Commit
38a656e
1 Parent(s): 5047984

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +26 -13
README.md CHANGED
@@ -19,19 +19,32 @@ GPT-6B_Tuned_small_pile is a GPT-j-6B model trained on 0.1 million example of pi
19
  n_embd: 4096, n_layer: 28, n_positions: 2048
20
 
21
  Tuning Parameters:
22
- val_split_percent: 20,
23
- momentum: 0.9
24
- train_batch_size (eff) : 32
25
- train_micro_batch: 16
26
- gradient_accumulation_steps: 2
27
- gradient_clipping: 0.5
28
- learning_rate: 0.00001
29
- weight_decay: 0.01
30
- lr_schedular: cosine
31
- lr_warmup_steps: 1000
32
- lr_decay: 0.1
33
- lr_decay_step: 2000
34
- mixed_precision: bf16
 
 
 
 
 
 
 
 
 
 
 
 
 
35
 
36
 
37
 
 
19
  n_embd: 4096, n_layer: 28, n_positions: 2048
20
 
21
  Tuning Parameters:
22
+
23
+ val_split_percent: 20,
24
+
25
+ momentum: 0.9
26
+
27
+ train_batch_size (eff) : 32
28
+
29
+ train_micro_batch: 16
30
+
31
+ gradient_accumulation_steps: 2
32
+
33
+ gradient_clipping: 0.5
34
+
35
+ learning_rate: 0.00001
36
+
37
+ weight_decay: 0.01
38
+
39
+ lr_schedular: cosine
40
+
41
+ lr_warmup_steps: 1000
42
+
43
+ lr_decay: 0.1
44
+
45
+ lr_decay_step: 2000
46
+
47
+ mixed_precision: bf16
48
 
49
 
50