demo: bump deps (#107)
* demo: Bump some React versions not using the latest because some things are broken like props and react-scripts using TypeScript. * demo: Add message about resetting account on error. * demo/nb: Correct test to have increase gasUsed max. * demo: Avoid some BN usage when deploying models. I'm not sure why this broke. * demo/ncc: Handle sparse version with centroids as arrays.
This commit is contained in:
Justin D. Harris
GitHub
Родитель
dfcbe24551
Коммит
9919db8b96
|
|
@ -86,11 +86,14 @@ yarn blockchain
|
|||
```
|
||||
|
||||
Do once:
|
||||
* Add http://localhost:7545 to MetaMask.
|
||||
* Copy the first private key output.
|
||||
* In your browser, you will need to add http://localhost:7545 as a custom RPC, you can use the MetaMask extension for this.
|
||||
* If required, you can set the chain ID to: 0xDeCA10B
|
||||
* Copy the first private key output by the above command.
|
||||
* Use that private key to create a new account in MetaMask.
|
||||
|
||||
## Server (Optional)
|
||||
## Server
|
||||
This is used by default in development mode but not in production.
|
||||
|
||||
If you want to store meta-data in a local database file instead of just within the browser, then start the server in one terminal.
|
||||
This step allows you to see models listed when you open the dashboard for the first time.
|
||||
|
||||
|
|
|
|||
|
|
@ -3,6 +3,7 @@
|
|||
# Service Storage
|
||||
# We do not want to enable the data storage service because we do not want to manage that data storage.
|
||||
# Default is enabled.
|
||||
# Disable the data store for a simple production environment.
|
||||
REACT_APP_ENABLE_SERVICE_DATA_STORE=false
|
||||
|
||||
# Online Safety
|
||||
|
|
@ -11,4 +12,5 @@ REACT_APP_ENABLE_SERVICE_DATA_STORE=false
|
|||
# Configure in: client/src/safety/config.ts
|
||||
|
||||
# Default is disabled.
|
||||
# Enable online safety for a simple production environment.
|
||||
REACT_APP_ENABLE_ONLINE_SAFETY=true
|
||||
|
|
|
|||
|
|
@ -1,4 +1,7 @@
|
|||
# See descriptions in `.env`.
|
||||
|
||||
REACT_APP_ENABLE_SERVICE_DATA_STORE=false
|
||||
# Use the data storage because it is helpful in development.
|
||||
REACT_APP_ENABLE_SERVICE_DATA_STORE=true
|
||||
|
||||
# Disable online safety because it makes development more difficult.
|
||||
REACT_APP_ENABLE_ONLINE_SAFETY=false
|
||||
|
|
|
|||
|
|
@ -43,7 +43,7 @@ module.exports = async function (deployer) {
|
|||
|
||||
console.log(`Deploying IMDB model with ${weights.length} weights.`);
|
||||
const intercept = convertNum(model.intercept || model.bias, web3, toFloat);
|
||||
const learningRate = convertNum(0.5, web3, toFloat);
|
||||
const learningRate = convertNum(model.learningRate, web3, toFloat);
|
||||
|
||||
console.log(`Deploying DataHandler.`);
|
||||
return deployer.deploy(DataHandler64).then(dataHandler => {
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
{
|
||||
"name": "decai-demo-client",
|
||||
"version": "1.3.0",
|
||||
"version": "1.4.0",
|
||||
"license": "MIT",
|
||||
"private": true,
|
||||
"proxy": "http://localhost:5387/",
|
||||
|
|
@ -14,8 +14,8 @@
|
|||
"@types/jest": "^24.0.23",
|
||||
"@types/murmurhash-js": "^1.0.3",
|
||||
"@types/node": "^12.12.14",
|
||||
"@types/react": "^16.9.14",
|
||||
"@types/react-dom": "^16.9.4",
|
||||
"@types/react": "^16.9.55",
|
||||
"@types/react-dom": "^16.9.9",
|
||||
"axios": "^0.19.0",
|
||||
"blueimp-load-image": "^2.24.0",
|
||||
"canvas": "^2.6.0",
|
||||
|
|
@ -24,18 +24,17 @@
|
|||
"moment": "^2.24.0",
|
||||
"murmurhash-js": "^1.0.0",
|
||||
"notistack": "^0.9.7",
|
||||
"react": "^16.12.0",
|
||||
"react-addons-update": "^15.6.2",
|
||||
"react-dom": "^16.12.0",
|
||||
"react": "^16.14.0",
|
||||
"react-dom": "^16.14.0",
|
||||
"react-dropzone": "^10.2.1",
|
||||
"react-router-dom": "^5.1.2",
|
||||
"react-scripts": "^3.2.0",
|
||||
"react-scripts": "^3.4.4",
|
||||
"react-spinners": "^0.6.1",
|
||||
"serve": "^11.2.0",
|
||||
"truffle": "^5.1.12",
|
||||
"truffle": "^5.1.50",
|
||||
"typescript": "^3.7.3",
|
||||
"web3": "^1.2.6",
|
||||
"web3-eth-contract": "^1.2.6"
|
||||
"web3": "^1.3.0",
|
||||
"web3-eth-contract": "^1.3.0"
|
||||
},
|
||||
"scripts": {
|
||||
"start": "bash deploy_client.sh",
|
||||
|
|
|
|||
|
|
@ -1097,8 +1097,8 @@ class Model extends React.Component {
|
|||
}
|
||||
})
|
||||
.on('error', err => {
|
||||
console.error(err);
|
||||
this.notify("Error adding data. See the console for details.", { variant: 'error' })
|
||||
console.error(err)
|
||||
this.notify("Error adding data. See the console for details. If you were using a private network, it might help to \"Reset Account\" in your browser or extension (MetaMask).", { variant: 'error' })
|
||||
});
|
||||
});
|
||||
}
|
||||
|
|
|
|||
|
|
@ -16,7 +16,7 @@ import DeleteIcon from '@material-ui/icons/Delete';
|
|||
import { withSnackbar } from 'notistack';
|
||||
import PropTypes from 'prop-types';
|
||||
import React from 'react';
|
||||
import update from 'react-addons-update';
|
||||
import update from 'immutability-helper';
|
||||
import { checkStorages } from '../components/storageSelector';
|
||||
import { getNetworkType } from '../getWeb3';
|
||||
import { OnlineSafetyValidator } from '../safety/validator';
|
||||
|
|
@ -237,7 +237,7 @@ class ModelList extends React.Component {
|
|||
{serviceStorageEnabled ? " or if they are listed on a centralized database" : ""}.
|
||||
</Typography>
|
||||
<Typography component="p">
|
||||
You can deploy your own model <Link href='/addModel'>here</Link> or use an already deployed model by filling in the information <Link href='/addDeployedModel'>here</Link>.
|
||||
You can deploy your own model <Link href='/add'>here</Link> or use an already deployed model by filling in the information <Link href='/addDeployedModel'>here</Link>.
|
||||
</Typography>
|
||||
</div>
|
||||
{this.state.loadingModels ?
|
||||
|
|
|
|||
|
|
@ -1,7 +1,9 @@
|
|||
const _toFloat = 1E9;
|
||||
|
||||
export function convertNum(num, web3, toFloat = _toFloat) {
|
||||
return web3.utils.toBN(Math.round(num * toFloat));
|
||||
export function convertNum(num, _web3, toFloat = _toFloat) {
|
||||
// We used to convert to BN here
|
||||
// but it caused problems and inconsistencies when running tests vs. running in the browser (using MetaMask).
|
||||
return Math.round(num * toFloat);
|
||||
}
|
||||
|
||||
export function convertToHex(num, web3, toFloat = _toFloat) {
|
||||
|
|
|
|||
|
|
@ -2,35 +2,32 @@ import assert from 'assert'
|
|||
import Web3 from 'web3'
|
||||
import { convertNum } from '../../float-utils'
|
||||
import { ModelDeployer } from '../deploy-model'
|
||||
import { CentroidInfo, DensePerceptronModel, NaiveBayesModel, NearestCentroidModel, SparseCentroidInfo, SparseNearestCentroidModel, SparsePerceptronModel } from '../model-interfaces'
|
||||
import { CentroidInfo, DensePerceptronModel, Model, NaiveBayesModel, NearestCentroidModel, SparseCentroidInfo, SparseNearestCentroidModel, SparsePerceptronModel } from '../model-interfaces'
|
||||
|
||||
declare const web3: Web3
|
||||
|
||||
function assertEqualNumbers(actual: any, expected: any, message?: string | Error): void {
|
||||
if (message) {
|
||||
message += "\n"
|
||||
}
|
||||
if (web3.utils.isBN(actual)) {
|
||||
if (web3.utils.isBN(expected)) {
|
||||
if (message === undefined) {
|
||||
message = `actual: ${actual} (${typeof actual})\nexpected: ${expected} (${typeof expected})`
|
||||
}
|
||||
message = `${message || ""}actual: ${actual} (BN)\nexpected: ${expected} (BN)`
|
||||
return assert(actual.eq(expected), message)
|
||||
} else {
|
||||
const expectedBN = web3.utils.toBN(expected)
|
||||
if (message === undefined) {
|
||||
message = `actual: ${actual} (${typeof actual})\nexpected: ${expected} (${typeof expected}) => BN: ${expectedBN}`
|
||||
}
|
||||
message = `${message || ""}actual: ${actual} (BN)\nexpected: ${expected} (${typeof expected}) => BN: ${expectedBN}`
|
||||
return assert(actual.eq(expectedBN), message)
|
||||
}
|
||||
} else if (web3.utils.isBN(expected)) {
|
||||
const actualBN = web3.utils.toBN(actual)
|
||||
if (message === undefined) {
|
||||
message = `actual: ${actual} (${typeof actual}) => BN: ${actualBN}\nexpected: ${expected} (${typeof expected})`
|
||||
}
|
||||
message = `${message || ""}actual: ${actual} (${typeof actual}) => BN: ${actualBN}\nexpected: ${expected} (BN)`
|
||||
return assert(actualBN.eq(expected), message)
|
||||
} else {
|
||||
if (typeof actual === 'string') {
|
||||
actual = parseInt(actual)
|
||||
}
|
||||
return assert.equal(actual, expected, message)
|
||||
return assert.strictEqual(actual, expected, message)
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -68,8 +65,9 @@ describe("ModelDeployer", () => {
|
|||
account,
|
||||
})
|
||||
|
||||
assertEqualNumbers(await m.methods.smoothingFactor().call(), convertNum(model.smoothingFactor, web3), "smoothingFactor")
|
||||
for (let i = 0; i < model.classifications.length; ++i) {
|
||||
assert.equal(await m.methods.classifications(i).call(), model.classifications[i])
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), model.classifications[i])
|
||||
assertEqualNumbers(await m.methods.getNumSamples(i).call(), model.classCounts[i])
|
||||
for (const [featureIndex, count] of model.featureCounts[i]) {
|
||||
assertEqualNumbers(await m.methods.getFeatureCount(i, featureIndex).call(), count)
|
||||
|
|
@ -83,8 +81,8 @@ describe("ModelDeployer", () => {
|
|||
const model = new NearestCentroidModel(
|
||||
'dense nearest centroid classifier',
|
||||
{
|
||||
"AA": new CentroidInfo([-1, -1], 2),
|
||||
"BB": new CentroidInfo([+1, +1], 2),
|
||||
"AA": new CentroidInfo([-1, -1, 4.88, -8.44, -3], 2),
|
||||
"BB": new CentroidInfo([+1, -1.8, 9.07, 3, -3], 2),
|
||||
}
|
||||
)
|
||||
const m = await deployer.deployModel(
|
||||
|
|
@ -96,20 +94,21 @@ describe("ModelDeployer", () => {
|
|||
let i = -1
|
||||
for (let [classification, centroidInfo] of Object.entries(model.centroids)) {
|
||||
++i
|
||||
assert.equal(await m.methods.classifications(i).call(), classification)
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), classification)
|
||||
assertEqualNumbers(await m.methods.getNumSamples(i).call(), centroidInfo.dataCount)
|
||||
for (let j = 0; j < centroidInfo.centroid.length; ++j) {
|
||||
assertEqualNumbers(await m.methods.getCentroidValue(i, j).call(), convertNum(centroidInfo.centroid[j], web3))
|
||||
const actual = await m.methods.getCentroidValue(i, j).call()
|
||||
assertEqualNumbers(actual, convertNum(centroidInfo.centroid[j], web3), `centroid value for class ${i}[${j}]`)
|
||||
}
|
||||
}
|
||||
})
|
||||
|
||||
it("should deploy sparse Nearest Centroid", async () => {
|
||||
// Values should all be positive since the representation is sparse.
|
||||
const model = new SparseNearestCentroidModel(
|
||||
'sparse nearest centroid classifier',
|
||||
{
|
||||
"AA": new SparseCentroidInfo({ '0': 0, '1': +1, '7': 1 }, 2),
|
||||
// Values should all be positive since the representation is sparse.
|
||||
"AA": new SparseCentroidInfo({ '0': 0, '1': +1, '7': 1.5, }, 2),
|
||||
"BB": new SparseCentroidInfo({ '0': +1, '1': 0, '5': 0.5 }, 2),
|
||||
}
|
||||
)
|
||||
|
|
@ -122,17 +121,50 @@ describe("ModelDeployer", () => {
|
|||
let i = -1
|
||||
for (let [classification, centroidInfo] of Object.entries(model.centroids)) {
|
||||
++i
|
||||
assert.equal(await m.methods.classifications(i).call(), classification)
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), classification)
|
||||
assertEqualNumbers(await m.methods.getNumSamples(i).call(), centroidInfo.dataCount)
|
||||
for (const [featureIndex, value] of Object.entries(centroidInfo.centroid)) {
|
||||
assertEqualNumbers(await m.methods.getCentroidValue(i, featureIndex).call(), convertNum(value, web3))
|
||||
assertEqualNumbers(await m.methods.getCentroidValue(i, featureIndex).call(), convertNum(value, web3), `centroid value for class ${i}[${featureIndex}]`)
|
||||
}
|
||||
}
|
||||
})
|
||||
|
||||
it("should deploy sparse Nearest Centroid with array centroids", async () => {
|
||||
// This shouldn't happen but it could if a model gets exported from the Python code
|
||||
// and the type is set correctly.
|
||||
const model = {
|
||||
type: 'sparse nearest centroid classifier',
|
||||
centroids: {
|
||||
"AA": {
|
||||
centroid: [0, 1, 1.5, 2, 87.88],
|
||||
dataCount: 2
|
||||
},
|
||||
"BB": {
|
||||
centroid: [1, 0, 0.5, 3.787],
|
||||
dataCount: 2
|
||||
},
|
||||
}
|
||||
}
|
||||
const m = await deployer.deployModel(
|
||||
model as Model,
|
||||
{
|
||||
account,
|
||||
})
|
||||
|
||||
let i = -1
|
||||
for (let [classification, centroidInfo] of Object.entries(model.centroids)) {
|
||||
++i
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), classification)
|
||||
assertEqualNumbers(await m.methods.getNumSamples(i).call(), centroidInfo.dataCount)
|
||||
for (const [featureIndex, value] of Object.entries(centroidInfo.centroid)) {
|
||||
assertEqualNumbers(await m.methods.getCentroidValue(i, featureIndex).call(), convertNum(value, web3), `centroid value for class ${i}[${featureIndex}]`)
|
||||
}
|
||||
}
|
||||
})
|
||||
|
||||
it("should deploy dense Perceptron", async () => {
|
||||
const classifications = ["A", "B"]
|
||||
const weights = [1, -1]
|
||||
const weights = [1, -1, 2.33, -8.66]
|
||||
const intercept = 0
|
||||
const m = await deployer.deployModel(
|
||||
new DensePerceptronModel(
|
||||
|
|
@ -146,7 +178,7 @@ describe("ModelDeployer", () => {
|
|||
})
|
||||
|
||||
for (let i = 0; i < classifications.length; ++i) {
|
||||
assert.equal(await m.methods.classifications(i).call(), classifications[i])
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), classifications[i])
|
||||
}
|
||||
for (let i = 0; i < weights.length; ++i) {
|
||||
assertEqualNumbers(await m.methods.weights(i).call(), convertNum(weights[i], web3))
|
||||
|
|
@ -156,8 +188,8 @@ describe("ModelDeployer", () => {
|
|||
|
||||
it("should deploy sparse Perceptron", async () => {
|
||||
const classifications = ["AA", "BB"]
|
||||
const weights = [2, -2]
|
||||
const sparseWeights = { '4': 7, '11': 8, }
|
||||
const weights = [2, -2, 2.44, -7.55, 0.537080412, 2000, -23232.32]
|
||||
const sparseWeights = null
|
||||
const intercept = 3
|
||||
const m = await deployer.deployModel(
|
||||
new SparsePerceptronModel(
|
||||
|
|
@ -171,14 +203,52 @@ describe("ModelDeployer", () => {
|
|||
})
|
||||
|
||||
for (let i = 0; i < classifications.length; ++i) {
|
||||
assert.equal(await m.methods.classifications(i).call(), classifications[i])
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), classifications[i])
|
||||
}
|
||||
|
||||
assertEqualNumbers(await m.methods.intercept().call(), convertNum(intercept, web3), "intercept:")
|
||||
|
||||
assertEqualNumbers(await m.methods.learningRate().call(), convertNum(0.5, web3), "learningRate:")
|
||||
|
||||
for (let i = 0; i < weights.length; ++i) {
|
||||
assertEqualNumbers(await m.methods.weights(i).call(), convertNum(weights[i], web3))
|
||||
assertEqualNumbers(await m.methods.weights(i).call(), convertNum(weights[i], web3), `weight ${i}:`)
|
||||
}
|
||||
for (const [featureIndex, weight] of Object.entries(sparseWeights)) {
|
||||
assertEqualNumbers(await m.methods.weights(featureIndex).call(), convertNum(weight, web3))
|
||||
})
|
||||
|
||||
it("should deploy sparse Perceptron - with sparseWeights", async () => {
|
||||
const classifications = ["AA", "BB"]
|
||||
const weights = [2, -2, 2.44, -7.55, -3]
|
||||
const sparseWeights = { '8': 7, '11': 8, '12': 8.21, '15': -4.55, '17': -3 }
|
||||
const intercept = 3
|
||||
const m = await deployer.deployModel(
|
||||
new SparsePerceptronModel(
|
||||
'sparse perceptron',
|
||||
classifications,
|
||||
weights, sparseWeights,
|
||||
intercept,
|
||||
),
|
||||
{
|
||||
account,
|
||||
// notify: console.debug,
|
||||
})
|
||||
|
||||
for (let i = 0; i < classifications.length; ++i) {
|
||||
assert.strictEqual(await m.methods.classifications(i).call(), classifications[i])
|
||||
}
|
||||
|
||||
assertEqualNumbers(await m.methods.intercept().call(), convertNum(intercept, web3), "intercept:")
|
||||
|
||||
assertEqualNumbers(await m.methods.learningRate().call(), convertNum(0.5, web3), "learningRate:")
|
||||
|
||||
for (let i = 0; i < weights.length; ++i) {
|
||||
const actual = await m.methods.weights(i).call()
|
||||
assertEqualNumbers(actual, convertNum(weights[i], web3), `weight ${i}:`)
|
||||
}
|
||||
|
||||
for (const [featureIndexKey, weight] of Object.entries(sparseWeights)) {
|
||||
const featureIndex = parseInt(featureIndexKey, 10)
|
||||
const actual = await m.methods.weights(featureIndex).call()
|
||||
assertEqualNumbers(actual, convertNum(weight, web3), `sparseWeight ${featureIndex}:`)
|
||||
}
|
||||
assertEqualNumbers(await m.methods.intercept().call(), convertNum(intercept, web3))
|
||||
})
|
||||
})
|
||||
|
|
|
|||
|
|
@ -17,6 +17,12 @@ async function deployDensePerceptron(model, web3, toFloat) {
|
|||
const weights = convertData(model.weights, web3, toFloat)
|
||||
const intercept = convertNum(model.intercept || model.bias, web3, toFloat)
|
||||
const learningRate = convertNum(model.learningRate || 1, web3, toFloat)
|
||||
|
||||
// TODO Handle feature indices.
|
||||
if (model.featureIndices) {
|
||||
throw new Error("featureIndices are not supported yet.")
|
||||
}
|
||||
|
||||
console.log(` Deploying Dense Perceptron classifier with first ${Math.min(weights.length, weightChunkSize)} weights.`)
|
||||
const classifierContract = await DensePerceptron.new(classifications, weights.slice(0, weightChunkSize), intercept, learningRate)
|
||||
gasUsed += (await web3.eth.getTransactionReceipt(classifierContract.transactionHash)).gasUsed
|
||||
|
|
@ -43,6 +49,20 @@ async function deploySparsePerceptron(model, web3, toFloat) {
|
|||
const weights = convertData(model.weights, web3, toFloat)
|
||||
const intercept = convertNum(model.intercept || model.bias, web3, toFloat)
|
||||
const learningRate = convertNum(model.learningRate || 1, web3, toFloat)
|
||||
const sparseWeights = []
|
||||
|
||||
// TODO Handle feature indices.
|
||||
if (model.featureIndices) {
|
||||
throw new Error("featureIndices are not supported yet.")
|
||||
}
|
||||
|
||||
if (typeof model.sparseWeights === 'object') {
|
||||
for (let [featureIndexKey, weight] of Object.entries(model.sparseWeights)) {
|
||||
const featureIndex = parseInt(featureIndexKey, 10)
|
||||
sparseWeights.push([featureIndex, convertNum(weight, web3, toFloat)])
|
||||
}
|
||||
}
|
||||
|
||||
console.log(` Deploying Sparse Perceptron classifier with first ${Math.min(weights.length, weightChunkSize)} weights...`)
|
||||
const classifierContract = await SparsePerceptron.new(classifications, weights.slice(0, weightChunkSize), intercept, learningRate)
|
||||
let gasUsed = (await web3.eth.getTransactionReceipt(classifierContract.transactionHash)).gasUsed
|
||||
|
|
@ -55,6 +75,14 @@ async function deploySparsePerceptron(model, web3, toFloat) {
|
|||
gasUsed += r.receipt.gasUsed
|
||||
}
|
||||
|
||||
const sparseWeightsChunkSize = Math.round(weightChunkSize / 2)
|
||||
for (let i = 0; i < sparseWeights.length; i += sparseWeightsChunkSize) {
|
||||
const r = await classifierContract.initializeSparseWeights(
|
||||
sparseWeights.slice(i, i + sparseWeightsChunkSize))
|
||||
console.debug(` Added sparse classifier weights [${i},${Math.min(i + sparseWeightsChunkSize, sparseWeights.length)}) out of ${sparseWeights.length}. gasUsed: ${r.receipt.gasUsed}`)
|
||||
gasUsed += r.receipt.gasUsed
|
||||
}
|
||||
|
||||
console.log(` Deployed Sparse Perceptron classifier to ${classifierContract.address}. gasUsed: ${gasUsed}`)
|
||||
|
||||
return {
|
||||
|
|
@ -205,11 +233,14 @@ async function deployNaiveBayes(model, web3, toFloat) {
|
|||
}
|
||||
|
||||
/**
|
||||
* @param model A model object or a string for the path to a JSON model file.
|
||||
* @returns The contract for the model, an instance of `Classifier64`
|
||||
* along with the the total amount of gas used to deploy the model.
|
||||
*/
|
||||
exports.deployModel = async function (path, web3, toFloat = _toFloat) {
|
||||
const model = JSON.parse(fs.readFileSync(path, 'utf8'))
|
||||
exports.deployModel = async function (model, web3, toFloat = _toFloat) {
|
||||
if (typeof model === 'string') {
|
||||
model = JSON.parse(fs.readFileSync(model, 'utf8'))
|
||||
}
|
||||
switch (model.type) {
|
||||
case 'dense perceptron':
|
||||
return deployDensePerceptron(model, web3, toFloat)
|
||||
|
|
|
|||
|
|
@ -5,7 +5,7 @@ import NaiveBayesClassifier from '../contracts/compiled/NaiveBayesClassifier.jso
|
|||
import NearestCentroidClassifier from '../contracts/compiled/NearestCentroidClassifier.json'
|
||||
import SparseNearestCentroidClassifier from '../contracts/compiled/SparseNearestCentroidClassifier.json'
|
||||
import SparsePerceptron from '../contracts/compiled/SparsePerceptron.json'
|
||||
import { convertDataToHex, convertToHex } from '../float-utils'
|
||||
import { convertData, convertNum } from '../float-utils'
|
||||
import { DensePerceptronModel, Model, NaiveBayesModel, NearestCentroidModel, SparseNearestCentroidModel, SparsePerceptronModel } from './model-interfaces'
|
||||
|
||||
export class ModelDeployer {
|
||||
|
|
@ -42,7 +42,7 @@ export class ModelDeployer {
|
|||
const initialFeatureChunkSize = 150
|
||||
const featureChunkSize = 350
|
||||
const { classifications, classCounts, featureCounts, totalNumFeatures } = model
|
||||
const smoothingFactor = convertToHex(model.smoothingFactor || defaultSmoothingFactor, this.web3, toFloat)
|
||||
const smoothingFactor = convertNum(model.smoothingFactor || defaultSmoothingFactor, this.web3, toFloat)
|
||||
|
||||
const ContractInfo = ModelDeployer.modelTypes[model.type]
|
||||
const contract = new this.web3.eth.Contract(ContractInfo.abi, undefined, { from: account })
|
||||
|
|
@ -119,8 +119,8 @@ export class ModelDeployer {
|
|||
for (let [classification, centroidInfo] of Object.entries(model.centroids)) {
|
||||
classifications.push(classification)
|
||||
dataCounts.push(centroidInfo.dataCount)
|
||||
if (Array.isArray(centroidInfo.centroid)) {
|
||||
centroids.push(convertDataToHex(centroidInfo.centroid, this.web3, toFloat))
|
||||
if (Array.isArray(centroidInfo.centroid) && model.type !== 'sparse nearest centroid classifier') {
|
||||
centroids.push(convertData(centroidInfo.centroid, this.web3, toFloat))
|
||||
if (numDimensions === null) {
|
||||
numDimensions = centroidInfo.centroid.length
|
||||
} else {
|
||||
|
|
@ -130,9 +130,10 @@ export class ModelDeployer {
|
|||
}
|
||||
} else {
|
||||
const sparseCentroid: number[][] = []
|
||||
// `centroidInfo.centroid` could be an array or dict.
|
||||
for (let [featureIndexKey, value] of Object.entries(centroidInfo.centroid)) {
|
||||
const featureIndex = parseInt(featureIndexKey)
|
||||
sparseCentroid.push([this.web3.utils.toHex(featureIndex), convertToHex(value, this.web3, toFloat)])
|
||||
sparseCentroid.push([featureIndex, convertNum(value, this.web3, toFloat)])
|
||||
}
|
||||
centroids.push(sparseCentroid as any)
|
||||
}
|
||||
|
|
@ -179,14 +180,14 @@ export class ModelDeployer {
|
|||
// Extend each class.
|
||||
for (let classification = 0; classification < classifications.length; ++classification) {
|
||||
for (let j = initialChunkSize; j < centroids[classification].length; j += chunkSize) {
|
||||
const notification = notify(`Please accept the prompt to upload the values for dimensions [${j},${j + chunkSize}) for the "${classifications[classification]}" class`)
|
||||
const notification = notify(`Please accept the prompt to upload the values for dimensions [${j},${Math.min(j + chunkSize, centroids[classification].length)}) for the "${classifications[classification]}" class`)
|
||||
// Not parallel since order matters.
|
||||
await newContractInstance.methods.extendCentroid(
|
||||
centroids[classification].slice(j, j + chunkSize), classification).send().on('transactionHash', () => {
|
||||
dismissNotification(notification)
|
||||
}).on('error', (err: any) => {
|
||||
dismissNotification(notification)
|
||||
notify(`Error setting feature indices for [${j},${j + chunkSize}) for the "${classifications[classification]}" class`, { variant: 'error' })
|
||||
notify(`Error setting feature indices for [${j},${Math.min(j + chunkSize, centroids[classification].length)}) for the "${classifications[classification]}" class`, { variant: 'error' })
|
||||
throw err
|
||||
})
|
||||
}
|
||||
|
|
@ -214,16 +215,16 @@ export class ModelDeployer {
|
|||
if (typeof sparseModel.sparseWeights === 'object' && sparseModel.sparseWeights !== null) {
|
||||
for (let [featureIndexKey, weight] of Object.entries(sparseModel.sparseWeights)) {
|
||||
const featureIndex = parseInt(featureIndexKey, 10)
|
||||
sparseWeights.push([this.web3.utils.toHex(featureIndex), convertToHex(weight, this.web3, toFloat)])
|
||||
sparseWeights.push([featureIndex, convertNum(weight, this.web3, toFloat)])
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (model.weights !== undefined && model.weights !== null && Array.isArray(model.weights)) {
|
||||
weightsArray = convertDataToHex(model.weights, this.web3, toFloat)
|
||||
weightsArray = convertData(model.weights, this.web3, toFloat)
|
||||
}
|
||||
const intercept = convertToHex(model.intercept, this.web3, toFloat)
|
||||
const learningRate = convertToHex(model.learningRate || defaultLearningRate, this.web3, toFloat)
|
||||
const intercept = convertNum(model.intercept, this.web3, toFloat)
|
||||
const learningRate = convertNum(model.learningRate || defaultLearningRate, this.web3, toFloat)
|
||||
|
||||
if (featureIndices !== undefined && featureIndices.length !== weightsArray.length + sparseWeights.length) {
|
||||
return Promise.reject("The number of features must match the number of weights.")
|
||||
|
|
@ -262,7 +263,10 @@ export class ModelDeployer {
|
|||
// Subtract 1 from the count because the first chunk has already been uploaded.
|
||||
const notification = notify(`Please accept the prompt to upload classifier
|
||||
weights [${i},${Math.min(i + weightChunkSize, weightsArray.length)}) (${i / weightChunkSize}/${Math.ceil(weightsArray.length / weightChunkSize) - 1})`)
|
||||
await transaction.send().on('transactionHash', () => {
|
||||
await transaction.send({
|
||||
from: account,
|
||||
gas: this.gasLimit,
|
||||
}).on('transactionHash', () => {
|
||||
dismissNotification(notification)
|
||||
}).on('error', (err: any) => {
|
||||
dismissNotification(notification)
|
||||
|
|
@ -274,27 +278,32 @@ export class ModelDeployer {
|
|||
// Add feature indices to use.
|
||||
for (let i = 0; i < featureIndices.length; i += weightChunkSize) {
|
||||
const notification = notify(`Please accept the prompt to upload the feature indices [${i},${Math.min(i + weightChunkSize, featureIndices.length)})`)
|
||||
await newContractInstance.methods.addFeatureIndices(featureIndices.slice(i, i + weightChunkSize)).send()
|
||||
.on('transactionHash', () => {
|
||||
dismissNotification(notification)
|
||||
}).on('error', (err: any) => {
|
||||
dismissNotification(notification)
|
||||
notify(`Error setting feature indices for [${i},${Math.min(i + weightChunkSize, featureIndices.length)})`, { variant: 'error' })
|
||||
console.error(err)
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
for (let i = 0; i < sparseWeights.length; i += Math.round(weightChunkSize / 2)) {
|
||||
const notification = notify(`Please accept the prompt to upload sparse classifier weights [${i},${i + Math.round(weightChunkSize / 2)}) out of ${sparseWeights.length}`)
|
||||
await newContractInstance.methods.initializeSparseWeights(
|
||||
sparseWeights.slice(i, i + Math.round(weightChunkSize / 2))).send().on('transactionHash', () => {
|
||||
await newContractInstance.methods.addFeatureIndices(featureIndices.slice(i, i + weightChunkSize)).send({
|
||||
from: account,
|
||||
gas: this.gasLimit,
|
||||
}).on('transactionHash', () => {
|
||||
dismissNotification(notification)
|
||||
}).on('error', (err: any) => {
|
||||
dismissNotification(notification)
|
||||
notify(`Error setting sparse classifier weights [${i},${i + Math.round(weightChunkSize / 2)}) out of ${sparseWeights.length}`, { variant: 'error' })
|
||||
throw err
|
||||
notify(`Error setting feature indices for [${i},${Math.min(i + weightChunkSize, featureIndices.length)})`, { variant: 'error' })
|
||||
console.error(err)
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
const sparseWeightsChunkSize = Math.round(weightChunkSize / 2)
|
||||
for (let i = 0; i < sparseWeights.length; i += sparseWeightsChunkSize) {
|
||||
const notification = notify(`Please accept the prompt to upload sparse classifier weights [${i},${Math.min(i + sparseWeightsChunkSize, sparseWeights.length)}) out of ${sparseWeights.length}`)
|
||||
await newContractInstance.methods.initializeSparseWeights(sparseWeights.slice(i, i + sparseWeightsChunkSize)).send({
|
||||
from: account,
|
||||
gas: this.gasLimit,
|
||||
}).on('transactionHash', () => {
|
||||
dismissNotification(notification)
|
||||
}).on('error', (err: any) => {
|
||||
dismissNotification(notification)
|
||||
notify(`Error setting sparse classifier weights [${i},${Math.min(i + sparseWeightsChunkSize, sparseWeights.length)}) out of ${sparseWeights.length}`, { variant: 'error' })
|
||||
throw err
|
||||
})
|
||||
}
|
||||
|
||||
notify(`The model contract has been deployed to ${newContractInstance.options.address}`, { variant: 'success' })
|
||||
|
|
|
|||
Различия файлов скрыты, потому что одна или несколько строк слишком длинны
|
|
@ -143,7 +143,7 @@ export class SparsePerceptronModel extends Model {
|
|||
type: 'sparse perceptron',
|
||||
public classifications: string[],
|
||||
public weights: number[],
|
||||
public sparseWeights: { [featureIndex: string]: number },
|
||||
public sparseWeights: { [featureIndex: string]: number } | null,
|
||||
public intercept: number,
|
||||
public learningRate?: number,
|
||||
public featureIndices?: number[],
|
||||
|
|
|
|||
|
|
@ -107,7 +107,7 @@ contract('NaiveBayesClassifier', function (accounts) {
|
|||
const updateResponse = await classifier.update(data, classification);
|
||||
// To help with optimizing gas usage:
|
||||
// console.log(` update gasUsed: ${updateResponse.receipt.gasUsed}`);
|
||||
assert.isBelow(updateResponse.receipt.gasUsed, 106424 + 1, "Too much gas used.");
|
||||
assert.isBelow(updateResponse.receipt.gasUsed, 113704 + 1, "Too much gas used.");
|
||||
|
||||
for (let i in prevFeatureCounts) {
|
||||
const featureIndex = data[i];
|
||||
|
|
|
|||
|
|
@ -4,6 +4,7 @@ const Classifier = artifacts.require("./classification/SparsePerceptron");
|
|||
const Stakeable64 = artifacts.require("./incentive/Stakeable64");
|
||||
|
||||
const { convertData, convertNum } = require('../../../src/float-utils-node');
|
||||
const { deployModel } = require('../../../src/ml-models/deploy-model-node')
|
||||
|
||||
contract('CollaborativeTrainer with Perceptron', function (accounts) {
|
||||
const toFloat = 1E9;
|
||||
|
|
@ -262,4 +263,25 @@ contract('CollaborativeTrainer with Perceptron', function (accounts) {
|
|||
});
|
||||
});
|
||||
});
|
||||
|
||||
it("...should initializeSparseWeights", async function () {
|
||||
const model = {
|
||||
type: 'sparse perceptron',
|
||||
classifications,
|
||||
weights: [0, 5, -1, 2.1, -1.4],
|
||||
sparseWeights: { '8': 9.1, '12': -7.3 },
|
||||
intercept: 2,
|
||||
learningRate: 0.5,
|
||||
}
|
||||
const { classifierContract } = await deployModel(model, web3, toFloat)
|
||||
assert.equal(await classifierContract.intercept().then(parseFloatBN), model.intercept)
|
||||
assert.equal(await classifierContract.learningRate().then(parseFloatBN), model.learningRate)
|
||||
for (let i = 0; i < model.weights; ++i) {
|
||||
assert.equal(await classifierContract.weights(i).then(parseFloatBN), model.weights[i])
|
||||
}
|
||||
|
||||
for (const [featureIndex, weight] of Object.entries(model.sparseWeights)) {
|
||||
assert.equal(await classifierContract.weights(parseInt(featureIndex, 10)).then(parseFloatBN), weight)
|
||||
}
|
||||
});
|
||||
});
|
||||
|
|
|
|||
Разница между файлами не показана из-за своего большого размера
Загрузить разницу
|
|
@ -1,10 +1,10 @@
|
|||
{
|
||||
"name": "decai-demo",
|
||||
"version": "1.3.0",
|
||||
"version": "1.4.0",
|
||||
"license": "MIT",
|
||||
"private": true,
|
||||
"scripts": {
|
||||
"blockchain": "mkdir --parents blockchain_db && ganache-cli --db blockchain_db --networkId 5782 --gasLimit 9000000 --gasPrice 1 --host 0.0.0.0 --port ${PORT:-7545} --mnemonic \"net pistol fun vibrant sausage vintage general silk weekend street brave rubber\"",
|
||||
"blockchain": "mkdir --parents blockchain_db && ganache-cli --db blockchain_db --networkId 5782 --chainId 0xDeCA10B --gasLimit 9000000 --gasPrice 1 --host 0.0.0.0 --port ${PORT:-7545} --mnemonic \"net pistol fun vibrant sausage vintage general silk weekend street brave rubber\"",
|
||||
"clean": "rm -rf blockchain_db db.sqlite client/src/contracts/compiled/*.json",
|
||||
"client": "cd client && yarn start",
|
||||
"prod-client": "cd client && yarn start-prod",
|
||||
|
|
@ -16,7 +16,7 @@
|
|||
"dependencies": {
|
||||
"body-parser": "^1.18.3",
|
||||
"express": "^4.16.4",
|
||||
"ganache-cli": "6.5.1",
|
||||
"ganache-cli": "^6.12.1",
|
||||
"sql.js": "^1.1.0"
|
||||
},
|
||||
"devDependencies": {
|
||||
|
|
|
|||
1028
demo/yarn.lock
1028
demo/yarn.lock
Разница между файлами не показана из-за своего большого размера
Загрузить разницу
Загрузка…
Ссылка в новой задаче