31:T628,vegalite({ data: { values: parsedAll }, vconcat: [ { selection: { hparamBrush: { type: "interval" }, }, mark: { type: "rect" }, encoding: { x: { field: otherFieldKeys[0], type: "ordinal" }, y: { field: otherFieldKeys[1], type: "ordinal" }, "color": { "condition": { "selection": "hparamBrush", "aggregate": "count", "field": "*", "type": "quantitative" }, "value": "grey" } } }, { selection: { "grid": { "type": "interval", "resolve": "global", "bind": "scales", "translate": "[mousedown[!event.shiftKey], window:mouseup] > window:mousemove!", "zoom": "wheel![!event.shiftKey]" } }, transform: [ { filter: {and: [{ selection: "hparamBrush" }]}, groupby: { and: otherFieldKeys } } ], mark: { type: lines ? 'line' : 'circle', interpolate: "monotone" }, encoding: { row: { field: rowField, type: "ordinal" }, column: { field: colField, type: "ordinal" }, x: {type: wallClock ? "temporal" : "quantitative", field: wallClock ? 't': 'i', domain: false, scale: wallClock ? 'time' : 'linear' }, y: {field: "lossEMA", type:"quantitative", scale: { type: logScale ? 'log' : 'linear' } }, color: {"field": "logName", "type": "nominal", legend: false }, opacity: {"value": 0.5}, } } ] })32:T66a,parsedAll = { const src = useFileInput ? logFiles : logFilenames.map(url => { const name = url.split('/')[url.split('/').length - 1] return [name, fetch(url).then(resp => resp.text())] }) const parsedIndiv = src.map(async ([name, logstrPromise]) => { const logstr = await logstrPromise const byLine = logstr.trim().split('\n').map(s => s.trim()) const [nClients, nExamples, nSync, nAvg] = name.slice(0, name.length-4).split("_") .map(x => parseInt(x, 10)) const clients = {} byLine.forEach(line => { const parts = line.split(' ') const clientId = parts[0] if(clientId !== undefined && clientId.length === 4 && clients[clientId] === undefined) { clients[clientId] = { loss: [], numUpdates: 0, nClients, nExamples, nSync, nAvg} } if(parts[1] === 'up') { clients[clientId].loss.push([parseFloat(parts[6]), parseFloat(parts[parts.length - 1])]) } else if(parts[1] === 'update') { clients[clientId].numUpdates++ } }) const ret = [] Object.keys(clients).map(clientId => { if(clients[clientId].loss.length == 0) return; let [lossEMA, t0] = clients[clientId].loss[0] clients[clientId].loss.forEach(([y, t], i) => { lossEMA = lossEMA * ema + (1.0 - ema) * y ret.push({...clients[clientId], loss: y, t:t - t0, lossEMA, i: i, clientId: clientId, logName:name}) }) }) return ret }) const parsedData = await Promise.all(parsedIndiv) return parsedData.reduce((a, b) => a.concat(b), []) }33:[] 2f:{"title":"Federated Learning Loss Vis","cells":[{"id":0,"value":"md`# Federated Learning Loss Vis`","pinned":false,"hidden":false,"mode":"ojs"},{"id":253,"value":"viewof rowField = \n html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":257,"value":"viewof colField = \n html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":261,"value":"otherFieldKeys = fields.filter(f => f !== rowField && f !== colField)","pinned":false,"hidden":false,"mode":"ojs"},{"id":362,"value":"viewof ema = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":422,"value":"viewof wallClock = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":443,"value":"viewof lines = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":447,"value":"viewof logScale = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":330,"value":"md`Drag over the hyperparam grid to filter the runs for the multiples chart`","pinned":false,"hidden":false,"mode":"ojs"},{"id":114,"value":"$31","pinned":false,"hidden":false,"mode":"ojs"},{"id":417,"value":"md`Check the below checkbox and upload a folder of logs to visualise them instead`","pinned":false,"hidden":false,"mode":"ojs"},{"id":411,"value":"viewof useFileInput = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":73,"value":"viewof logFilesInput = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":10,"value":"vegalite = require(\"@observablehq/vega-lite@0.1\")","pinned":false,"hidden":false,"mode":"ojs"},{"id":75,"value":"logFiles = Promise.all([...logFilesInput].map(f => Files.text(f)))\n .then(fs => fs.map((f, i) => [logFilesInput[i].name, f]))","pinned":false,"hidden":false,"mode":"ojs"},{"id":116,"value":"$32","pinned":false,"hidden":false,"mode":"ojs"},{"id":390,"value":"logFilenames = {\n const raw = await fetch('https://storage.googleapis.com/learnjs-data/logs/logfilenames.txt').then(r => r.text())\n return raw.trim().split('\\n').map(line => line.trim()).map(fname => `https://storage.googleapis.com/learnjs-data/logs/${fname}`)\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":150,"value":"fields = [\"nClients\", \"nExamples\", \"nSync\", \"nAvg\"]","pinned":false,"hidden":false,"mode":"ojs"},{"id":190,"value":"uniqueValues = {\n return fields.map(key => {\n const uniq = parsedAll.reduce((uniq, vals) => (uniq.add(vals[key]), uniq), new Set)\n const vals = [...uniq].sort((a, b) => a - b)\n return [key, vals]\n }).reduce((obj, [k, v]) => (obj[k] = v, obj), {})\n}","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W33","stdlib":"1","titling":"auto"}