OmAlve commited on
Commit
032c570
·
verified ·
1 Parent(s): 76ae1a8

Llama-3.2-3B base + CPT on Marathi Wikipedia (1500 steps) — LoRA adapter

Browse files
Files changed (1) hide show
  1. trainer_state.json +139 -0
trainer_state.json ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.5133470225872689,
6
+ "eval_steps": 500,
7
+ "global_step": 1500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.034223134839151265,
14
+ "grad_norm": 0.4172394275665283,
15
+ "learning_rate": 0.00019800000000000002,
16
+ "loss": 1.1892,
17
+ "step": 100
18
+ },
19
+ {
20
+ "epoch": 0.06844626967830253,
21
+ "grad_norm": 0.41112303733825684,
22
+ "learning_rate": 0.00019754247929410995,
23
+ "loss": 1.0497,
24
+ "step": 200
25
+ },
26
+ {
27
+ "epoch": 0.1026694045174538,
28
+ "grad_norm": 0.4096732437610626,
29
+ "learning_rate": 0.00019019402315000752,
30
+ "loss": 1.0129,
31
+ "step": 300
32
+ },
33
+ {
34
+ "epoch": 0.13689253935660506,
35
+ "grad_norm": 0.34457647800445557,
36
+ "learning_rate": 0.00017832286206772186,
37
+ "loss": 1.0343,
38
+ "step": 400
39
+ },
40
+ {
41
+ "epoch": 0.17111567419575632,
42
+ "grad_norm": 0.4272993206977844,
43
+ "learning_rate": 0.00016252426563357055,
44
+ "loss": 0.9878,
45
+ "step": 500
46
+ },
47
+ {
48
+ "epoch": 0.2053388090349076,
49
+ "grad_norm": 0.383188933134079,
50
+ "learning_rate": 0.00014359044144195547,
51
+ "loss": 1.0277,
52
+ "step": 600
53
+ },
54
+ {
55
+ "epoch": 0.23956194387405885,
56
+ "grad_norm": 0.41364046931266785,
57
+ "learning_rate": 0.00012247081049860883,
58
+ "loss": 0.9947,
59
+ "step": 700
60
+ },
61
+ {
62
+ "epoch": 0.2737850787132101,
63
+ "grad_norm": 0.4872409999370575,
64
+ "learning_rate": 0.00010022439928692875,
65
+ "loss": 0.9475,
66
+ "step": 800
67
+ },
68
+ {
69
+ "epoch": 0.3080082135523614,
70
+ "grad_norm": 0.4362848103046417,
71
+ "learning_rate": 7.796673575799224e-05,
72
+ "loss": 0.9456,
73
+ "step": 900
74
+ },
75
+ {
76
+ "epoch": 0.34223134839151265,
77
+ "grad_norm": 0.44856351613998413,
78
+ "learning_rate": 5.681391210104916e-05,
79
+ "loss": 0.9583,
80
+ "step": 1000
81
+ },
82
+ {
83
+ "epoch": 0.37645448323066394,
84
+ "grad_norm": 0.348608136177063,
85
+ "learning_rate": 3.782661922075803e-05,
86
+ "loss": 0.914,
87
+ "step": 1100
88
+ },
89
+ {
90
+ "epoch": 0.4106776180698152,
91
+ "grad_norm": 0.39325082302093506,
92
+ "learning_rate": 2.195695926616702e-05,
93
+ "loss": 0.9214,
94
+ "step": 1200
95
+ },
96
+ {
97
+ "epoch": 0.44490075290896647,
98
+ "grad_norm": 0.5586904883384705,
99
+ "learning_rate": 1.0000703253328414e-05,
100
+ "loss": 0.9282,
101
+ "step": 1300
102
+ },
103
+ {
104
+ "epoch": 0.4791238877481177,
105
+ "grad_norm": 0.3972143530845642,
106
+ "learning_rate": 2.5573877837031137e-06,
107
+ "loss": 0.9231,
108
+ "step": 1400
109
+ },
110
+ {
111
+ "epoch": 0.5133470225872689,
112
+ "grad_norm": 0.4245308041572571,
113
+ "learning_rate": 2.5177551682409317e-10,
114
+ "loss": 0.8962,
115
+ "step": 1500
116
+ }
117
+ ],
118
+ "logging_steps": 100,
119
+ "max_steps": 1500,
120
+ "num_input_tokens_seen": 0,
121
+ "num_train_epochs": 1,
122
+ "save_steps": 1500,
123
+ "stateful_callbacks": {
124
+ "TrainerControl": {
125
+ "args": {
126
+ "should_epoch_stop": false,
127
+ "should_evaluate": false,
128
+ "should_log": false,
129
+ "should_save": true,
130
+ "should_training_stop": true
131
+ },
132
+ "attributes": {}
133
+ }
134
+ },
135
+ "total_flos": 3.770604684680233e+17,
136
+ "train_batch_size": 8,
137
+ "trial_name": null,
138
+ "trial_params": null
139
+ }