Diff
checker
文本
文本
圖像
文檔
Excel
文件夾
Legal
Enterprise
桌面版
定價
登入
下載 Diffchecker 桌面版
比較文本
尋找兩個文字檔案之間的差異
工具
歷史
即時編輯器
摺疊未變更行
關閉換行
檢視
拆分
統一
比對精度
智能
單詞
字符
語法突出顯示
選擇語法
忽略
文字轉換
前往第一個差異
編輯輸入
Diffchecker Desktop
執行Diffchecker最安全的方式。取得Diffchecker桌面應用程式:您的差異永遠不會離開您的電腦!
取得桌面版
Diff
建立於
3 年前
差異永不過期
清除
匯出
分享
解釋
22 刪除
行
總計
刪除
字符
總計
刪除
要繼續使用此功能,請升級到
Diff
checker
Pro
查看價格
96 行
全部複製
21 新增
行
總計
新增
字符
總計
新增
要繼續使用此功能,請升級到
Diff
checker
Pro
查看價格
96 行
全部複製
import { writeFileSync } from "fs";
import { writeFileSync } from "fs";
import { codePartToCompactString } from "./utils/codePartToCompactString";
import { codePartToCompactString } from "./utils/codePartToCompactString";
import { getFlydeFiles } from "./utils/fs-helpers";
import { getFlydeFiles } from "./utils/fs-helpers";
import { preprocessStdLibParts } from "./utils/preprocessStdLibParts";
import { preprocessStdLibParts } from "./utils/preprocessStdLibParts";
import { join } from "path";
import { join } from "path";
import { chunkArray } from "./utils";
import { chunkArray } from "./utils";
import { readVersionResult } from "./utils/generatePartVersions";
import { readVersionResult } from "./utils/generatePartVersions";
複製
已複製
複製
已複製
import { fullChatInstructions } from "./benchmark/chat-completion-instructions";
(async function () {
(async function () {
const files = getFlydeFiles();
const files = getFlydeFiles();
const parts = preprocessStdLibParts(files);
const parts = preprocessStdLibParts(files);
const partsWithVersions = parts.map((part) => {
const partsWithVersions = parts.map((part) => {
const versionData = readVersionResult(part);
const versionData = readVersionResult(part);
return { ...part, ...versionData };
return { ...part, ...versionData };
});
});
const partsPerNamespace = partsWithVersions.reduce<
const partsPerNamespace = partsWithVersions.reduce<
Record<string, typeof parts>
Record<string, typeof parts>
>((acc, part) => {
>((acc, part) => {
const namespace = part.original.namespace ?? "n/a";
const namespace = part.original.namespace ?? "n/a";
if (!acc[namespace]) {
if (!acc[namespace]) {
acc[namespace] = [];
acc[namespace] = [];
}
}
acc[namespace].push(part);
acc[namespace].push(part);
return acc;
return acc;
}, {});
}, {});
const trainingIds = new Set<string>();
const trainingIds = new Set<string>();
const validationIds = new Set<string>();
const validationIds = new Set<string>();
Object.entries(partsPerNamespace).forEach(([k, parts]) => {
Object.entries(partsPerNamespace).forEach(([k, parts]) => {
const chunks = chunkArray(parts, 10);
const chunks = chunkArray(parts, 10);
chunks.forEach((chunk) => {
chunks.forEach((chunk) => {
chunk.forEach((part, idx) => {
chunk.forEach((part, idx) => {
if (idx === 0 && chunk.length > 4) {
if (idx === 0 && chunk.length > 4) {
validationIds.add(part.original.id);
validationIds.add(part.original.id);
} else {
} else {
trainingIds.add(part.original.id);
trainingIds.add(part.original.id);
}
}
});
});
});
});
});
});
const validationDataset = partsWithVersions.flatMap((part) => {
const validationDataset = partsWithVersions.flatMap((part) => {
if (!validationIds.has(part.original.id)) {
if (!validationIds.has(part.original.id)) {
return [];
return [];
}
}
複製
已複製
複製
已複製
const compactParts =
[
part.original.runFnString,
part.alternativeFunction]
const compactParts =
codePartToCompactString({
.map((code) => ({ ...part.original, runFnString: code }))
...part.original,
.map(codePartToCompactString)
;
runFnString:
part.original.runFnString,
})
;
return part.prompts.map((desc, idx) => {
return part.prompts.map((desc, idx) => {
return {
return {
prompt: desc + "\n\n###\n\n",
prompt: desc + "\n\n###\n\n",
completion: " " + compactParts[idx % 2] + "###",
completion: " " + compactParts[idx % 2] + "###",
};
};
});
});
});
});
const trainingDataset = partsWithVersions.flatMap((part) => {
const trainingDataset = partsWithVersions.flatMap((part) => {
if (!trainingIds.has(part.original.id)) {
if (!trainingIds.has(part.original.id)) {
return [];
return [];
}
}
複製
已複製
複製
已複製
const compactPart
s = [part.original.runFnString, part.alternativeFunction]
const compactPart
=
codePartToCompactString
(
{
.map((code) => ({ ...part.original, runFnString: code }))
...part.original
,
.map(
codePartToCompactString
);
runFnString: part.original.runFnString,
return part.prompts.map((desc, idx) => {
return
{
prompt: desc + "\n\n###\n\n"
,
completion: " " + compactParts[idx % 2] + "###",
};
});
});
複製
已複製
複製
已複製
const prompt = part.prompts[0];
return {
messages: [
{ role: "system", content: fullChatInstructions },
{ role: "user", content: prompt },
{ role: "assistant", content: compactPart },
],
};
});
});
複製
已複製
複製
已複製
console.log(
const datasetFileLocation = join(__dirname, `../dataset
-cc
.json`);
partsWithVersions.length,
trainingDataset.length,
validationDataset.length
);
const datasetFileLocation = join(__dirname, `../dataset
.json`);
writeFileSync(
writeFileSync(
datasetFileLocation,
datasetFileLocation,
JSON.stringify([...trainingDataset, ...validationDataset], null, 2)
JSON.stringify([...trainingDataset, ...validationDataset], null, 2)
);
);
console.log(
console.log(
複製
已複製
複製
已複製
`
Dataset
written to: ${datasetFileLocation}. Remember, the last ${validationDataset.length} entries are validation data.`
`
${trainingDataset.length} examples
written to: ${datasetFileLocation}. Remember, the last ${validationDataset.length} entries are validation data.`
);
);
})();
})();
已保存差異
原始文本
開啟檔案
import { writeFileSync } from "fs"; import { codePartToCompactString } from "./utils/codePartToCompactString"; import { getFlydeFiles } from "./utils/fs-helpers"; import { preprocessStdLibParts } from "./utils/preprocessStdLibParts"; import { join } from "path"; import { chunkArray } from "./utils"; import { readVersionResult } from "./utils/generatePartVersions"; (async function () { const files = getFlydeFiles(); const parts = preprocessStdLibParts(files); const partsWithVersions = parts.map((part) => { const versionData = readVersionResult(part); return { ...part, ...versionData }; }); const partsPerNamespace = partsWithVersions.reduce< Record<string, typeof parts> >((acc, part) => { const namespace = part.original.namespace ?? "n/a"; if (!acc[namespace]) { acc[namespace] = []; } acc[namespace].push(part); return acc; }, {}); const trainingIds = new Set<string>(); const validationIds = new Set<string>(); Object.entries(partsPerNamespace).forEach(([k, parts]) => { const chunks = chunkArray(parts, 10); chunks.forEach((chunk) => { chunk.forEach((part, idx) => { if (idx === 0 && chunk.length > 4) { validationIds.add(part.original.id); } else { trainingIds.add(part.original.id); } }); }); }); const validationDataset = partsWithVersions.flatMap((part) => { if (!validationIds.has(part.original.id)) { return []; } const compactParts = [part.original.runFnString, part.alternativeFunction] .map((code) => ({ ...part.original, runFnString: code })) .map(codePartToCompactString); return part.prompts.map((desc, idx) => { return { prompt: desc + "\n\n###\n\n", completion: " " + compactParts[idx % 2] + "###", }; }); }); const trainingDataset = partsWithVersions.flatMap((part) => { if (!trainingIds.has(part.original.id)) { return []; } const compactParts = [part.original.runFnString, part.alternativeFunction] .map((code) => ({ ...part.original, runFnString: code })) .map(codePartToCompactString); return part.prompts.map((desc, idx) => { return { prompt: desc + "\n\n###\n\n", completion: " " + compactParts[idx % 2] + "###", }; }); }); console.log( partsWithVersions.length, trainingDataset.length, validationDataset.length ); const datasetFileLocation = join(__dirname, `../dataset.json`); writeFileSync( datasetFileLocation, JSON.stringify([...trainingDataset, ...validationDataset], null, 2) ); console.log( `Dataset written to: ${datasetFileLocation}. Remember, the last ${validationDataset.length} entries are validation data.` ); })();
更改後文本
開啟檔案
import { writeFileSync } from "fs"; import { codePartToCompactString } from "./utils/codePartToCompactString"; import { getFlydeFiles } from "./utils/fs-helpers"; import { preprocessStdLibParts } from "./utils/preprocessStdLibParts"; import { join } from "path"; import { chunkArray } from "./utils"; import { readVersionResult } from "./utils/generatePartVersions"; import { fullChatInstructions } from "./benchmark/chat-completion-instructions"; (async function () { const files = getFlydeFiles(); const parts = preprocessStdLibParts(files); const partsWithVersions = parts.map((part) => { const versionData = readVersionResult(part); return { ...part, ...versionData }; }); const partsPerNamespace = partsWithVersions.reduce< Record<string, typeof parts> >((acc, part) => { const namespace = part.original.namespace ?? "n/a"; if (!acc[namespace]) { acc[namespace] = []; } acc[namespace].push(part); return acc; }, {}); const trainingIds = new Set<string>(); const validationIds = new Set<string>(); Object.entries(partsPerNamespace).forEach(([k, parts]) => { const chunks = chunkArray(parts, 10); chunks.forEach((chunk) => { chunk.forEach((part, idx) => { if (idx === 0 && chunk.length > 4) { validationIds.add(part.original.id); } else { trainingIds.add(part.original.id); } }); }); }); const validationDataset = partsWithVersions.flatMap((part) => { if (!validationIds.has(part.original.id)) { return []; } const compactParts = codePartToCompactString({ ...part.original, runFnString: part.original.runFnString, }); return part.prompts.map((desc, idx) => { return { prompt: desc + "\n\n###\n\n", completion: " " + compactParts[idx % 2] + "###", }; }); }); const trainingDataset = partsWithVersions.flatMap((part) => { if (!trainingIds.has(part.original.id)) { return []; } const compactPart = codePartToCompactString({ ...part.original, runFnString: part.original.runFnString, }); const prompt = part.prompts[0]; return { messages: [ { role: "system", content: fullChatInstructions }, { role: "user", content: prompt }, { role: "assistant", content: compactPart }, ], }; }); const datasetFileLocation = join(__dirname, `../dataset-cc.json`); writeFileSync( datasetFileLocation, JSON.stringify([...trainingDataset, ...validationDataset], null, 2) ); console.log( `${trainingDataset.length} examples written to: ${datasetFileLocation}. Remember, the last ${validationDataset.length} entries are validation data.` ); })();
尋找差異