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
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
|
import { h, Component } from 'preact'
import { bindActionCreators } from 'redux'
import { connect } from 'react-redux'
import util from '../../../util'
const { lerp, norm, randint, randrange } = util
import * as samplernnActions from '../samplernn.actions'
import Group from '../../../common/group.component'
import Slider from '../../../common/slider.component'
import Select from '../../../common/select.component'
import Button from '../../../common/button.component'
import { FileList } from '../../../common/fileList.component'
import TextInput from '../../../common/textInput.component'
class SampleRNNGraph extends Component {
constructor(props){
super()
props.actions.load_graph()
}
render(){
this.refs = {}
return (
<div className='app lossGraph'>
<div className='heading'>
<h3>SampleRNN Loss Graph</h3>
<canvas ref={(ref) => this.refs['canvas'] = ref} />
</div>
</div>
)
}
componentDidUpdate(){
const { lossReport, results } = this.props.samplernn
if (! lossReport || ! results) return
const canvas = this.refs.canvas
canvas.width = window.innerWidth
canvas.height = window.innerHeight
canvas.style.width = canvas.width + 'px'
canvas.style.height = canvas.height + 'px'
const ctx = canvas.getContext('2d')
const w = canvas.width
const h = canvas.height
ctx.clearRect(0,0,w,h)
const resultsByDate = results.map(file => {
if (!file.name.match(/^exp:/)) return null
const dataset = file.name.split("-")[3].split(":")[1]
return [
+new Date(file.date),
dataset
]
}).filter(a => !!a).sort((a,b) => a[0]-a[1])
const keys = Object.keys(lossReport).filter(k => !!lossReport[k].length)
let scaleMax = 0
let scaleMin = Infinity
let epochsMax = 0
keys.forEach(key => {
const loss = lossReport[key]
epochsMax = Math.max(loss.length, epochsMax)
loss.forEach((a) => {
const v = parseFloat(a.training_loss)
if (! v) return
scaleMax = Math.max(v, scaleMax)
scaleMin = Math.min(v, scaleMin)
})
})
// scaleMax *= 10
console.log(scaleMax, scaleMin, epochsMax)
scaleMax = 3
scaleMin = 0
const margin = 0
const wmin = 0
const wmax = w/2
const hmin = 0
const hmax = h/2
const epochsScaleFactor = 1 // 3/2
ctx.save()
let X, Y
for (var ii = 0; ii < epochsMax; ii++) {
X = lerp((ii)/(epochsMax/(epochsScaleFactor))*(epochsScaleFactor), wmin, wmax)
ctx.strokeStyle = 'rgba(0,0,0,0.3)'
ctx.beginPath(0, 0)
ctx.moveTo(X, 0)
ctx.lineTo(X, h)
ctx.lineWidth = 0.5
// ctx.stroke()
if ( ((ii+1) % 6) === 0 ) {
ctx.lineWidth = 0.5
ctx.stroke()
const fontSize = 12
ctx.font = 'italic ' + (fontSize) + 'px "Georgia"'
ctx.fillStyle = 'rgba(0,12,28,0.6)'
ctx.fillText(ii/5*6, X + (8), h - ((fontSize + 4)))
}
}
for (var ii = scaleMin; ii < scaleMax; ii += 1) {
Y = lerp(ii/scaleMax, hmin, hmax)
// ctx.strokeStyle = 'rgba(255,255,255,1.0)'
ctx.beginPath(0, 0)
ctx.moveTo(0, (h-Y))
ctx.lineTo(w, (h-Y))
ctx.lineWidth = 1
// ctx.stroke()
// if ( (ii % 1) < 0.1) {
// ctx.strokeStyle = 'rgba(255,255,255,1.0)'
ctx.lineWidth = 2
ctx.setLineDash([4, 4])
ctx.stroke()
ctx.stroke()
ctx.stroke()
ctx.setLineDash([1,1])
const fontSize = 12
ctx.font = 'italic ' + (fontSize) + 'px "Georgia"'
ctx.fillStyle = 'rgba(0,12,28,0.6)'
ctx.fillText(ii.toFixed(1), w-50, (h-Y) + fontSize + (10))
// }
}
ctx.lineWidth = 1
ctx.restore()
const min_date = resultsByDate[0][0]
const max_date = resultsByDate[resultsByDate.length-1][0]
resultsByDate.forEach(pair => {
const date = pair[0]
const key = pair[1]
const loss = lossReport[key]
if (!key || !loss || !loss.length) return
const vf = parseFloat(loss[loss.length-1].training_loss) || 0
const vg = parseFloat(loss[0].training_loss) || 5
// console.log(vf)
const vv = 1 - norm(vf, scaleMin, scaleMax/2)
ctx.lineWidth = (1-norm(vf, scaleMin, scaleMax)) * 4
// ctx.lineWidth = norm(date, min_date, max_date) * 3
console.log(date, min_date, max_date)
ctx.strokeStyle = 'rgba(' + [randrange(30,190), randrange(30,150), randrange(60,120)].join(',') + ',' + 0.8+ ')'
let begun = false
loss.forEach((a, i) => {
const v = parseFloat(a.training_loss)
if (! v) return
const x = lerp((i)/(epochsMax/(epochsScaleFactor))*(epochsScaleFactor), wmin, wmax)
const y = lerp(norm(v, scaleMin, scaleMax), hmax, hmin)
if (! begun) {
begun = true
ctx.beginPath(0,0)
ctx.moveTo(x,y)
} else {
ctx.lineTo(x,y)
// ctx.stroke()
}
})
ctx.stroke()
const i = loss.length-1
const v = parseFloat(loss[i].training_loss)
const x = lerp((i)/(epochsMax/(epochsScaleFactor))*(epochsScaleFactor), wmin, wmax)
const y = lerp(norm(v, scaleMin, scaleMax), hmax, hmin)
const fontSize = 9
ctx.font = 'italic ' + (fontSize) + 'px "Georgia"'
ctx.fillStyle = 'rgba(0,12,28,0.6)'
ctx.fillText(key, x + 4, y + fontSize/2)
})
}
}
const mapStateToProps = state => ({
samplernn: state.module.samplernn,
})
const mapDispatchToProps = (dispatch, ownProps) => ({
actions: bindActionCreators(samplernnActions, dispatch),
})
export default connect(mapStateToProps, mapDispatchToProps)(SampleRNNGraph)
|
