blob: 06e761e252488b2edd28cef09d13edfcf315ecfe (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
|
# @package _global_
defaults:
- override /mapping: null
- override /criterion: ctc
- override /callbacks: htr
- override /datamodule: iam_lines
- override /network: conformer
- override /model: null
- override /lr_schedulers: null
- override /optimizers: null
epochs: &epochs 999
num_classes: &num_classes 57
max_output_len: &max_output_len 89
summary: [[1, 56, 1024]]
mapping: &mapping
mapping:
_target_: text_recognizer.data.mappings.EmnistMapping
callbacks:
stochastic_weight_averaging:
_target_: pytorch_lightning.callbacks.StochasticWeightAveraging
swa_epoch_start: 0.75
swa_lrs: 1.0e-5
annealing_epochs: 10
annealing_strategy: cos
device: null
optimizers:
radam:
_target_: torch.optim.RAdam
lr: 3.0e-4
betas: [0.9, 0.999]
weight_decay: 0
eps: 1.0e-8
parameters: network
lr_schedulers:
network:
_target_: torch.optim.lr_scheduler.ReduceLROnPlateau
mode: min
factor: 0.5
patience: 10
threshold: 1.0e-4
threshold_mode: rel
cooldown: 0
min_lr: 1.0e-5
eps: 1.0e-8
verbose: false
interval: epoch
monitor: val/loss
datamodule:
batch_size: 8
num_workers: 12
train_fraction: 0.9
pin_memory: true
<< : *mapping
model:
_target_: text_recognizer.models.conformer.LitConformer
<<: *mapping
max_output_len: *max_output_len
start_token: <s>
end_token: <e>
pad_token: <p>
blank_token: <b>
trainer:
_target_: pytorch_lightning.Trainer
stochastic_weight_avg: true
auto_scale_batch_size: binsearch
auto_lr_find: false
gradient_clip_val: 0.5
fast_dev_run: false
gpus: 1
precision: 16
max_epochs: *epochs
terminate_on_nan: true
weights_summary: null
limit_train_batches: 1.0
limit_val_batches: 1.0
limit_test_batches: 1.0
resume_from_checkpoint: null
accumulate_grad_batches: 1
overfit_batches: 0
|
