forked from erogol/seg-torch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
opts.lua
49 lines (42 loc) · 2.11 KB
/
opts.lua
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
local opts = {}
lapp = require 'pl.lapp'
function opts.parse(arg)
local opt = lapp [[
Command line options:
Training Related:
-l,--learningRate (default 5e-4) learning rate
-d,--learningRateDecaySteps (default 5) number of epochs to reduce LR by 0.1
-w,--weightDecay (default 2e-4) L2 penalty on the weights
-m,--momentum (default 0.9) momentum
-b,--batchSize (default 10) batch size
--maxepoch (default 300) maximum number of training epochs
--plot plot training/testing error in real-time
--manualSeed (default 0), Manually set RNG seed
--checkpoint (default false) Save model per epoch
Device Related:
-t,--nThreads (default 8) number of threads for data loader
-i,--devid (default 1) device ID (if using CUDA)
--nGPU (default 1) number of GPUs you want to train on
--save (default /media/) save trained model here
Dataset Related:
--channels (default 3)
--datapath (default /media/Dataset/)
dataset location
--dataset (default nil) dataset type
--cachepath (default /media/) cache directory to save the loaded dataset
--imHeight (default 360) image height
--imWidth (default 480) image width
--maskHeight (default 45) mask height
--maskWidth (default 60) mask width
Model Related:
--model (default models/encoder.lua)
Path of a model
--encoder (default /media/Models/CamVid/enc/model-100.net)
pretrained encoder for which you want to train your decoder
Saving/Displaying Information:
--confusion (default skip)
skip: skip confusion, all: test+train, test : test only
]]
return opt
end
return opts