Add files using upload-large-folder tool

4de8d861ed2563988d5f1871647ebc5fe70861b32d24a4b32f9363518653a328/crowd-code-faba6583-b2c9-4b94-9ba6-9f240428520a1750722089894-2025_06_23-22.50.32.930/source.csv

@@ -0,0 +1,182 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,2,"src/recording.ts",0,0,"import * as fs from 'node:fs'\nimport * as path from 'node:path'\nimport * as vscode from 'vscode'\nimport * as readline from 'node:readline'\nimport axios from 'axios'\nimport { hasConsent, showConsentChangeDialog } from './consent'\nimport {\n    getEditorFileName,\n    escapeString,\n    getEditorLanguage,\n    notificationWithProgress,\n    generateBaseFilePath,\n    formatDisplayTime,\n    getExportPath,\n    logToOutput,\n    formatSrtTime,\n    getConfig,\n    removeDoubleQuotes,\n    unescapeString,\n    addToGitignore,\n} from './utilities'\nimport { type File, ChangeType, type CSVRowBuilder, type Change, type Recording, type ConsentStatus } from './types'\nimport { extContext, statusBarItem, actionsProvider } from './extension'\n\nexport const commands = {\n    openSettings: 'crowd-code.openSettings',\n    startRecording: 'crowd-code.startRecording',\n    stopRecording: 'crowd-code.stopRecording',\n    panicButton: 'crowd-code.panicButton',\n}\n\nexport const recording: Recording = {\n    isRecording: false,\n    timer: 0,\n    startDateTime: null,\n    endDateTime: null,\n    sequence: 0,\n    customFolderName: '',\n    activatedFiles: new Set<string>(),\n}\n\nlet intervalId: NodeJS.Timeout\nconst fileQueue: File[] = []\nlet isAppending = false\n\nlet uploadIntervalId: NodeJS.Timeout;\nconst sessionUuid = vscode.env.sessionId;\n\nlet panicStatusBarItem: vscode.StatusBarItem | undefined;\nlet panicButtonPressCount = 0;\nlet panicButtonTimeoutId: NodeJS.Timeout | undefined;\nlet accumulatedRemovedContent: Array<{content: string, sequence: number}> = []; // Store content with sequence numbers\n\nconst CROWD_CODE_API_GATEWAY_URL = process.env.CROWD_CODE_API_GATEWAY_URL;\n\nconst PANIC_BUTTON_TIMEOUT = 3000; // 3 seconds timeout for successive presses\n\n/**\n * Builds a CSV row with the given parameters.\n *\n * @param {CSVRowBuilder} sequence - The sequence number of the change.\n * @param {CSVRowBuilder} rangeOffset - The offset of the changed range.\n * @param {CSVRowBuilder} rangeLength - The length of the changed range.\n * @param {CSVRowBuilder} text - The text of the change.\n * @param {string} type - The type of the change (optional, defaults to 'content').\n * @return {string} A CSV row string with the provided information.\n */\nexport function buildCsvRow({\n    sequence,\n    rangeOffset,\n    rangeLength,\n    text,\n    type = ChangeType.CONTENT,\n}: CSVRowBuilder): string | undefined {\n    if (!recording.startDateTime) {\n        return\n    }\n\n    const time = new Date().getTime() - recording.startDateTime.getTime()\n\n    if (type === ChangeType.HEADING) {\n        return 'Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type\n'\n    }\n\n    if (type === ChangeType.TERMINAL_FOCUS || type === ChangeType.TERMINAL_COMMAND || type === ChangeType.TERMINAL_OUTPUT) {\n        return `${sequence},${time},""TERMINAL"",${rangeOffset},${rangeLength},""${escapeString(text)}"",,${type}\n`\n    }\n\n    const editorFileName = getEditorFileName()\n    return `${sequence},${time},""${editorFileName}"",${rangeOffset},${rangeLength},""${escapeString(text)}"",${getEditorLanguage()},${type}\n`\n}\n\n/**\n * Checks if the current file being edited is within the configured export path.\n * This is used to determine if the current file should be recorded or not.\n *\n * @returns {boolean} `true` if the current file is within the export path, `false` otherwise.\n */\nexport function isCurrentFileExported(): boolean {\n    const editor = vscode.window.activeTextEditor\n    const filename = editor?.document.fileName.replaceAll('\\', '/')\n    const exportPath = getExportPath()\n    if (!editor || !filename || !exportPath) {\n        return false\n    }\n    return filename.startsWith(exportPath)\n}\n\nconst onChangeSubscription = vscode.workspace.onDidChangeTextDocument(event => {\n    if (!recording.isRecording) {\n        return\n    }\n\n    if (isCurrentFileExported()) {\n        return\n    }\n    const editor = vscode.window.activeTextEditor\n    if (editor && event.document === editor.document) {\n        for (const change of event.contentChanges) {\n            recording.sequence++\n            addToFileQueue(\n                buildCsvRow({\n                    sequence: recording.sequence,\n                    rangeOffset: change.rangeOffset,\n                    rangeLength: change.rangeLength,\n                    text: change.text,\n                })\n            )\n            appendToFile()\n        }\n    }\n})\n\n/**\n * Creates the recording folder if it doesn't exist.\n * @param folderPath - The path to the recording folder.\n */\nfunction createRecordingFolder(folderPath: string): void {\n    if (!fs.existsSync(folderPath)) {\n        fs.mkdirSync(folderPath, { recursive: true })\n    }\n}\n\n/**\n * Starts the recording process and initializes necessary variables.\n */\nexport async function startRecording(): Promise<void> {\n    if (recording.isRecording) {\n        notificationWithProgress('Already recording')\n        logToOutput('Already recording', 'info')\n        return\n    }\n    const exportPath = getExportPath()\n    if (!exportPath) {\n        return\n    }\n\n    // If the setting is enabled and the path is inside the workspace, add it to .gitignore\n    if (\n        getConfig().get<boolean>('export.addToGitignore') &&\n        getConfig().get<string>('export.exportPath')?.startsWith('${workspaceFolder}')\n    ) {\n        await addToGitignore()\n    }\n\n    recording.startDateTime = new Date()\n    recording.activatedFiles = new Set<string>()\n\n    // Ask for folder name if enabled in settings\n    let customFolderName: string | undefined\n    if (getConfig().get('recording.askFolderName')) {\n        customFolderName = await vscode.window.showInputBox({\n            prompt: 'Enter a name for the recording folder',\n            placeHolder: 'Enter recording folder name',\n        })\n        if (!customFolderName) {\n            stopRecording(true)\n            return\n        }\n        recording.customFolderName = customFolderName\n    }\n\n    const baseFilePath = generateBaseFilePath(recording.startDateTime, false, recording.customFolderName, sessionUuid)\n    if (!baseFilePath) {\n        stopRecording(true)\n        return\n    }\n\n    // Create the recording folder\n    const folderPath = path.dirname(path.join(exportPath, baseFilePath))\n    createRecordingFolder(folderPath)\n\n    recording.isRecording = true\n    recording.timer = 0\n    recording.endDateTime = null\n    recording.sequence = 0\n    panicButtonPressCount = 0 // Reset panic button counter for new recording\n    accumulatedRemovedContent = [] // Clear accumulated content for new recording\n    if (panicButtonTimeoutId) {\n        clearTimeout(panicButtonTimeoutId)\n        panicButtonTimeoutId = undefined\n    }\n    intervalId = setInterval(() => {\n        recording.timer++\n        updateStatusBarItem()\n    }, 1000)\n    notificationWithProgress('Recording started')\n    logToOutput('Recording started', 'info')\n\n    // Only log initial editor content if there's an active text editor\n    const editorText = vscode.window.activeTextEditor?.document.getText()\n    const activeEditorUri = vscode.window.activeTextEditor?.document.uri.toString()\n\n    if (editorText !== undefined && activeEditorUri) {\n        recording.sequence++\n        const csvRow = {\n            sequence: recording.sequence,\n            rangeOffset: 0,\n            rangeLength: 0,\n            text: editorText,\n            type: ChangeType.TAB,\n        }\n        addToFileQueue(buildCsvRow({ ...csvRow, type: ChangeType.HEADING }))\n        addToFileQueue(buildCsvRow(csvRow))\n        appendToFile()\n        recording.activatedFiles.add(activeEditorUri)\n        actionsProvider.setCurrentFile(vscode.window.activeTextEditor?.document.fileName || '')\n    } else {\n        // If no active editor, just add the header row\n        recording.sequence++\n        addToFileQueue(buildCsvRow({ \n            sequence: recording.sequence,\n            rangeOffset: 0,\n            rangeLength: 0,\n            text: '',\n            type: ChangeType.HEADING \n        }))\n        appendToFile()\n    }\n\n    extContext.subscriptions.push(onChangeSubscription)\n    updateStatusBarItem()\n    updatePanicButton()\n    actionsProvider.setRecordingState(true)\n\n    // Set up a timer to send data to the Lambda endpoint periodically\n    uploadIntervalId = setInterval(async () => {\n        if (!exportPath) {\n            return;\n        }\n        \n        if (typeof CROWD_CODE_API_GATEWAY_URL !== 'string' || !CROWD_CODE_API_GATEWAY_URL.trim()) {\n            logToOutput(""CROWD_CODE_API_GATEWAY_URL must be a non-empty string. Please check your build configuration."", 'error');\n            return;\n        }\n\n        // Only upload data if user has given consent\n        if (!hasConsent()) {\n            return;\n        }\n\n        const filePath = path.join(exportPath, `${baseFilePath}.csv`);\n        const extensionVersion = extContext.extension.packageJSON.version as string;\n        const userId = extContext.globalState.get<string>('userId');\n\n        try {\n            const fileContent = await fs.promises.readFile(filePath, 'utf-8');\n\n            if (fileContent) {\n                const payload = {\n                    fileName: `${baseFilePath}.csv`,\n                    content: fileContent,\n                    version: extensionVersion,\n                    userId: userId\n                };\n                await axios.post(CROWD_CODE_API_GATEWAY_URL, payload);\n                console.log(`Successfully sent ${payload.fileName} to Lambda endpoint.`);\n                logToOutput(`Successfully sent to Lambda endpoint.`, 'info');\n            }\n        } catch (error: any) {\n            if (error.code === 'ENOENT') {\n                console.warn(`File not found at ${filePath}. It might be created on first write.`);\n            } else {\n                console.error(`Error sending data to Lambda: ${error.message}`);\n                if (axios.isAxiosError(error) && error.response) {\n                    console.error(""Lambda response status:"", error.response.status);\n                    console.error(""Lambda response data:"", error.response.data);\n                }\n            }\n        }\n    }, 1 * 60 * 1000); // 5 minutes\n}\n\n/**\n * Stops the recording process and finalizes the recording data.\n * @param context - The extension context.\n */\nexport function stopRecording(force = false): Promise<void> | void {\n    if (!recording.isRecording) {\n        notificationWithProgress('Not recording')\n        return\n    }\n\n    recording.isRecording = false\n    clearInterval(intervalId)\n    clearInterval(uploadIntervalId); // Clear the upload timer\n    recording.timer = 0\n    recording.activatedFiles?.clear()\n    panicButtonPressCount = 0 // Reset panic button counter when recording stops\n    accumulatedRemovedContent = [] // Clear accumulated content when recording stops\n    if (panicButtonTimeoutId) {\n        clearTimeout(panicButtonTimeoutId)\n        panicButtonTimeoutId = undefined\n    }\n    const index = extContext.subscriptions.indexOf(onChangeSubscription)\n    if (index !== -1) {\n        extContext.subscriptions.splice(index, 1)\n    }\n    updateStatusBarItem()\n    updatePanicButton()\n    actionsProvider.setRecordingState(false)\n    if (force) {\n        notificationWithProgress('Recording cancelled')\n        logToOutput('Recording cancelled', 'info')\n        recording.customFolderName = undefined\n        return\n    }\n    notificationWithProgress('Recording finished')\n    logToOutput('Recording finished', 'info')\n    recording.endDateTime = new Date()\n    return processCsvFile().then(() => {\n        // Reset customFolderName after processing is complete\n        recording.customFolderName = undefined\n    }).catch(err => {\n        logToOutput(`Error processing CSV file during stop: ${String(err)}`, 'error')\n        recording.customFolderName = undefined\n    });\n}\n\n/**\n * Appends data from the file queue to the appropriate file in the workspace.\n */\nexport async function appendToFile(): Promise<void> {\n    if (isAppending) {\n        return\n    }\n    isAppending = true\n\n    const exportPath = getExportPath()\n    if (!exportPath) {\n        logToOutput('Export path not available in appendToFile, stopping recording.', 'error')\n        stopRecording(true)\n        isAppending = false\n        return\n    }\n\n    while (fileQueue.length > 0) {\n        const itemToAppend = fileQueue.shift()\n        if (!itemToAppend) {\n            continue\n        }\n\n        const filePath = path.join(exportPath, itemToAppend.name)\n\n        try {\n            const directory = path.dirname(filePath)\n            if (!fs.existsSync(directory)) {\n                fs.mkdirSync(directory, { recursive: true })\n            }\n            await fs.promises.appendFile(filePath, itemToAppend.content)\n        } catch (err) {\n            logToOutput(\n                `Failed to append to file ${filePath}: ${err}. Item dropped. Content: ${itemToAppend.content.substring(0, 100)}...`,\n                'error'\n            )\n        }\n    }\n    isAppending = false\n}\n\n/**\n * Appends an SRT line to the file queue for the previous change.\n *\n * This function is responsible for generating the SRT format line for the previous change and adding it to the file queue.\n * It checks if the SRT export format is enabled, and if so, it generates the SRT line for the previous change and adds it to the file queue.\n *\n * @param processedChanges - An array of processed changes.\n * @param i - The index of the current change in the processedChanges array.\n * @param exportInSrt - A boolean indicating whether the SRT export format is enabled.\n */\nfunction addToSRTFile(processedChanges: Change[], i: number, exportInSrt: boolean) {\n    if (!exportInSrt) {\n        return\n    }\n    if (i === 0) {\n        return\n    }\n    addToFileQueue(\n        addSrtLine(\n            processedChanges[i - 1].sequence,\n            processedChanges[i - 1].startTime,\n            processedChanges[i - 1].endTime,\n            JSON.stringify({\n                text: processedChanges[i - 1].text,\n                file: processedChanges[i - 1].file,\n                language: processedChanges[i - 1].language,\n            })\n        ),\n        'srt',\n        true\n    )\n}\n\n/**\n * Returns the new text content based on the change type and the previous change.\n * @param type - The type of the change.\n * @param text - The text of the change.\n * @param previousChange - The previous change.\n * @param rangeOffset - The offset of the range.\n * @param rangeLength - The length of the range.\n */\nfunction getNewTextContent(\n    type: string,\n    text: string,\n    previousChange: Change | null,\n    rangeOffset: number,\n    rangeLength: number\n): string {\n    if (type === ChangeType.TAB) {\n        return text\n    }\n    if (!previousChange) {\n        return ''\n    }\n    return getUpdatedText(previousChange.text, rangeOffset, rangeLength, text)\n}\n\n/**\n * Processes a single CSV line and returns the processed change\n */\nasync function processCSVLine(line: string, previousChange: Change | null): Promise<Change | null> {\n    const lineArr = line.split(/,(?=(?:[^""]*""[^""]*"")*[^""]*$)/)\n\n    if (Number.isNaN(Number.parseInt(lineArr[0]))) {\n        return null\n    }\n\n    const time = Number.parseInt(lineArr[1])\n    const file = removeDoubleQuotes(lineArr[2])\n    const rangeOffset = Number.parseInt(lineArr[3])\n    const rangeLength = Number.parseInt(lineArr[4])\n    const text = unescapeString(removeDoubleQuotes(lineArr[5]))\n    const language = lineArr[6]\n    const type = lineArr[7]\n\n    const newText = getNewTextContent(type, text, previousChange, rangeOffset, rangeLength)\n\n    /**\n     * Skip exporting changes with the same values to the previous change.\n     */\n    if (\n        previousChange &&\n        time === previousChange.startTime &&\n        file === previousChange.file &&\n        newText === previousChange.text &&\n        language === previousChange.language\n    ) {\n        return null\n    }\n\n    return {\n        sequence: previousChange ? previousChange.sequence + 1 : 1,\n        file,\n        startTime: time,\n        endTime: 0,\n        language,\n        text: newText,\n    }\n}\n\n/**\n * Returns the updated text content based on the previous text, range offset, range length, and new text.\n * @param previousText - The previous text.\n * @param rangeOffset - The offset of the range.\n * @param rangeLength - The length of the range.\n * @param newText - The new text.\n */\nfunction getUpdatedText(\n    previousText: string,\n    rangeOffset: number,\n    rangeLength: number,\n    newText: string\n): string {\n    const textArray = previousText.split('')\n    textArray.splice(rangeOffset, rangeLength, newText)\n    return textArray.join('')\n}\n\n/**\n * Processes the CSV file and generates the necessary output files.\n */\nasync function processCsvFile(): Promise<void> {\n    if (!validateRecordingState()) {\n        return\n    }\n\n    const exportFormats = getConfig().get<string[]>('export.exportFormats', [])\n    if (exportFormats.length === 0) {\n        logToOutput('No export formats specified', 'info')\n        vscode.window.showWarningMessage('No export formats specified')\n        return\n    }\n\n    const exportPath = getExportPath()\n    if (!exportPath) {\n        return\n    }\n\n    if (!recording.startDateTime) {\n        return\n    }\n\n    // Use the same custom folder name for reading the source file\n    const baseFilePathSource = generateBaseFilePath(\n        recording.startDateTime,\n        false,\n        recording.customFolderName,\n        sessionUuid\n    )\n    if (!baseFilePathSource) {\n        return\n    }\n\n    const filePath = path.join(exportPath, `${baseFilePathSource}.csv`)\n\n    try {\n        if (!fs.existsSync(filePath)) {\n            throw new Error(`Source file not found: ${filePath}`)\n        }\n\n        const processedChanges: Change[] = []\n\n        const rl = readline.createInterface({\n            input: fs.createReadStream(filePath),\n            crlfDelay: Number.POSITIVE_INFINITY,\n        })\n\n        for await (const line of rl) {\n            const previousChange = processedChanges[processedChanges.length - 1]\n            const change = await processCSVLine(line, previousChange)\n\n            if (change) {\n                if (previousChange) {\n                    previousChange.endTime = change.startTime\n                    if (exportFormats.includes('SRT')) {\n                        addToSRTFile(processedChanges, processedChanges.length, true)\n                    }\n                }\n                processedChanges.push(change)\n            }\n        }\n\n        rl.close();\n\n        return finalizeRecording(processedChanges, exportFormats);\n\n    } catch (err) {\n        vscode.window.showErrorMessage(`Error processing recording: ${err}`)\n        logToOutput('Error processing CSV file: ' + String(err), 'error')\n        return Promise.resolve(); // Resolve even on error after showing message\n    }\n}\n\nfunction validateRecordingState(): boolean {\n    if (!vscode.workspace.workspaceFolders) {\n        logToOutput(\n            'No workspace folder found. To process the recording is needed a workspace folder',\n            'error'\n        )\n        return false\n    }\n    if (!recording.endDateTime || !recording.startDateTime) {\n        logToOutput('Recording date time is not properly set', 'error')\n        return false\n    }\n    return true\n}\n\nfunction finalizeRecording(processedChanges: Change[], exportFormats: string[]): Promise<void> {\n    const lastChange = processedChanges[processedChanges.length - 1]\n    if (lastChange && recording.endDateTime && recording.startDateTime) {\n        lastChange.endTime = recording.endDateTime.getTime() - recording.startDateTime.getTime()\n        if (exportFormats.includes('SRT')) {\n            addToSRTFile(processedChanges, processedChanges.length, true)\n        }\n    }\n    if (exportFormats.includes('JSON')) {\n        addToFileQueue(JSON.stringify(processedChanges), 'json', true)\n    }\n    return appendToFile().then(() => {\n        // Refresh the recordFiles view after export is complete\n        vscode.commands.executeCommand('crowd-code.refreshRecordFiles')\n    })\n}\n\n/**\n * Adds a line to the SRT file format.\n * @param sequence - The sequence number of the change.\n * @param start - The start time of the change.\n * @param end - The end time of the change.\n * @param text - The text of the change.\n * @returns A string representing a line in the SRT file format.\n */\nfunction addSrtLine(sequence: number, start: number, end: number, text: string): string {\n    return `${sequence}\n${formatSrtTime(start)} --> ${formatSrtTime(end)}\n${text}\n\n`\n}\n\n/**\n * Adds content to the file queue.\n * @param content - The content to add.\n * @param fileExtension - The file extension (optional, defaults to 'csv').\n */\nexport function addToFileQueue(\n    content: string | undefined,\n    fileExtension = 'csv',\n    isExport = false\n): void {\n    if (!content) {\n        return\n    }\n    if (!recording.startDateTime) {\n        return\n    }\n    // Use the same custom name throughout the recording session\n    const baseFilePath = generateBaseFilePath(recording.startDateTime, isExport, recording.customFolderName, sessionUuid)\n    if (!baseFilePath) {\n        return\n    }\n    fileQueue.push({\n        name: `${baseFilePath}.${fileExtension}`,\n        content: content,\n    })\n}\n\n/**\n * Updates the status bar item with the current recording status and time.\n */\nexport function updateStatusBarItem(): void {\n    if (recording.isRecording) {\n        if (getConfig().get('appearance.showTimer') === false) {\n            statusBarItem.text = '$(debug-stop)'\n            statusBarItem.tooltip = 'Current time: ' + formatDisplayTime(recording.timer)\n        }\n        if (getConfig().get('appearance.showTimer') === true) {\n            statusBarItem.text = '$(debug-stop) ' + formatDisplayTime(recording.timer)\n            statusBarItem.tooltip = 'Stop Recording'\n        }\n        statusBarItem.command = commands.stopRecording\n        statusBarItem.show()\n    } else {\n        const editor = vscode.window.activeTextEditor\n        if (!editor) {\n            statusBarItem.hide()\n            return\n        }\n        if (getConfig().get('appearance.minimalMode') === true) {\n            statusBarItem.text = '$(circle-large-filled)'\n        } else {\n            statusBarItem.text = '$(circle-large-filled) Start Recording'\n        }\n        statusBarItem.tooltip = 'Start Recording'\n        statusBarItem.command = commands.startRecording\n        statusBarItem.show()\n    }\n}\n\n/**\n * Creates and updates the panic button status bar item.\n */\nexport function updatePanicButton(): void {\n    if (!recording.isRecording) {\n        if (panicStatusBarItem) {\n            panicStatusBarItem.hide()\n        }\n        return\n    }\n\n    // Create panic button if it doesn't exist\n    if (!panicStatusBarItem) {\n        panicStatusBarItem = vscode.window.createStatusBarItem(vscode.StatusBarAlignment.Right, 8999) // Position it to the left of the recording button\n        extContext.subscriptions.push(panicStatusBarItem)\n    }\n\n    const secondsToRemove = (panicButtonPressCount + 1) * 10 // Show what the next press will remove\n    panicStatusBarItem.text = '$(refresh)'\n    panicStatusBarItem.tooltip = `Remove last ${secondsToRemove} seconds of recording (click again within 3 seconds to remove more)`\n    panicStatusBarItem.command = commands.panicButton\n    panicStatusBarItem.show()\n}\n\n/**\n * Deletes the last N seconds of recording data from the CSV file.\n * This is a ""panic button"" feature that allows users to quickly remove recent sensitive data.\n * Each successive press within 3 seconds removes more time: 10s, 20s, 30s, etc.\n * After 3 seconds of inactivity, the next press will be treated as a fresh press (10s).\n */\nexport async function panicButton(): Promise<void> {\n    if (!recording.isRecording) {\n        vscode.window.showWarningMessage('No active recording to remove data from')\n        logToOutput('No active recording to remove data from', 'info')\n        return\n    }\n\n    if (!recording.startDateTime) {\n        vscode.window.showErrorMessage('Recording start time not available')\n        logToOutput('Recording start time not available', 'error')\n        return\n    }\n\n    const exportPath = getExportPath()\n    if (!exportPath) {\n        vscode.window.showErrorMessage('Export path not available')\n        logToOutput('Export path not available', 'error')\n        return\n    }\n\n    const baseFilePath = generateBaseFilePath(recording.startDateTime, false, recording.customFolderName, sessionUuid)\n    if (!baseFilePath) {\n        vscode.window.showErrorMessage('Could not generate file path')\n        logToOutput('Could not generate file path', 'error')\n        return\n    }\n\n    const filePath = path.join(exportPath, `${baseFilePath}.csv`)\n\n    try {\n        // Check if file exists\n        if (!fs.existsSync(filePath)) {\n            vscode.window.showWarningMessage('No recording file found to remove data from')\n            logToOutput('No recording file found to remove data from', 'info')\n            return\n        }\n\n        // Read the file\n        const content = fs.readFileSync(filePath, 'utf-8')\n        const lines = content.split('\n')\n        \n        if (lines.length <= 1) {\n            vscode.window.showWarningMessage('Recording file is empty, nothing to remove')\n            logToOutput('Recording file is empty, nothing to remove', 'info')\n            return\n        }\n\n        // Calculate how many lines to remove (10 seconds per press)\n        const linesToRemove = Math.min((panicButtonPressCount + 1) * 10, lines.length - 1)\n        const newLines = lines.slice(0, lines.length - linesToRemove)\n        \n        // Capture the lines that will be removed for display\n        const removedLines = lines.slice(lines.length - linesToRemove)\n\n        // Write back to file\n        fs.writeFileSync(filePath, newLines.join('\n'))\n\n        // Update panic button state\n        panicButtonPressCount++\n        \n        // Set up timeout to reset the counter after 3 seconds of inactivity\n        if (panicButtonTimeoutId) {\n            clearTimeout(panicButtonTimeoutId)\n        }\n        panicButtonTimeoutId = setTimeout(() => {\n            panicButtonPressCount = 0\n            accumulatedRemovedContent = [] // Clear accumulated content\n            updatePanicButton()\n        }, PANIC_BUTTON_TIMEOUT)\n        \n        updatePanicButton()\n\n        const secondsToRemove = panicButtonPressCount * 10\n        const actualLinesRemoved = lines.length - newLines.length\n        \n        // Accumulate removed content and show immediate popup\n        if (removedLines.length > 0) {\n            const nonEmptyLines = removedLines.filter(line => line.trim())\n            if (nonEmptyLines.length > 0) {\n                // Create a simple, readable summary of removed content\n                const contentSummary = nonEmptyLines.map(line => {\n                    // Extract just the text content from CSV for cleaner display\n                    const parts = line.split(',')\n                    if (parts.length >= 6) {\n                        const textContent = parts[5].replace(/^""|""$/g, '') // Remove quotes\n                        // Clean up common escape sequences\n                        const cleanText = textContent\n                            .replace(/\\n/g, '\n')\n                            .replace(/\\t/g, '\t')\n                            .replace(/\\r/g, '\r')\n                        return { content: cleanText, sequence: Number.parseInt(parts[0]) }\n                    }\n                    return { content: line, sequence: Number.parseInt(line.split(',')[0]) }\n                }).filter(item => item.content.trim().length > 0)\n                \n                // Add to accumulated content\n                accumulatedRemovedContent.push(...contentSummary)\n                \n                // Sort by sequence number to show in original file order\n                const sortedContent = accumulatedRemovedContent.sort((a, b) => a.sequence - b.sequence)\n                \n                // Show immediate popup with accumulated content\n                const totalContent = sortedContent.map(item => item.content).join(' ')\n                const summaryText = totalContent.length > 100 \n                    ? totalContent.substring(0, 100) + '...' \n                    : totalContent\n                \n                vscode.window.showInformationMessage(\n                    `Removed content: ""${summaryText}""`,\n                    'Dismiss'\n                )\n            }\n        }\n\n    } catch (error) {\n        const errorMessage = `Error during panic button operation: ${error}`\n        vscode.window.showErrorMessage(errorMessage)\n        logToOutput(errorMessage, 'error')\n    }\n}",typescript,tab
+2,78,"extension-output-pdoom-org.crowd-code-#5-crowd-code",0,0,"10:50:32 PM [info] Activating crowd-code\n10:50:32 PM [info] Recording started\n10:50:32 PM [info] Initializing git provider using file system watchers...\n10:50:32 PM [info] Git repository found\n10:50:32 PM [info] Git provider initialized successfully\n10:50:32 PM [info] Initial git state: [object Object]\n",Log,tab
+3,2192,"src/recording.ts",0,0,"",typescript,tab
+4,2193,"src/recording.ts",10135,0,"",typescript,selection_mouse
+5,256675,"src/recording.ts",9085,0,"",typescript,selection_command
+6,310049,"src/recording.ts",9119,0,"",typescript,selection_command
+7,310303,"src/recording.ts",9172,0,"",typescript,selection_command
+8,310329,"src/recording.ts",9214,0,"",typescript,selection_command
+9,310363,"src/recording.ts",9261,0,"",typescript,selection_command
+10,310399,"src/recording.ts",9291,0,"",typescript,selection_command
+11,310431,"src/recording.ts",9315,0,"",typescript,selection_command
+12,310681,"src/recording.ts",9386,0,"",typescript,selection_command
+13,311616,"src/recording.ts",9476,0,"",typescript,selection_command
+14,313315,"src/recording.ts",9454,78,"",typescript,content
+15,313342,"src/recording.ts",9466,0,"",typescript,selection_command
+16,314505,"src/recording.ts",9466,0,"    logToOutput(`Successfully sent to Lambda endpoint.`, 'info');\n            ",typescript,content
+17,314508,"src/recording.ts",9476,0,"",typescript,selection_command
+18,316423,"src/recording.ts",9454,78,"",typescript,content
+19,316439,"src/recording.ts",9466,0,"",typescript,selection_command
+20,317197,"src/recording.ts",9480,0,"",typescript,selection_command
+21,317443,"src/recording.ts",9511,0,"",typescript,selection_command
+22,317478,"src/recording.ts",9554,0,"",typescript,selection_command
+23,317508,"src/recording.ts",9654,0,"",typescript,selection_command
+24,317544,"src/recording.ts",9675,0,"",typescript,selection_command
+25,317578,"src/recording.ts",9756,0,"",typescript,selection_command
+26,317611,"src/recording.ts",9823,0,"",typescript,selection_command
+27,317645,"src/recording.ts",9908,0,"",typescript,selection_command
+28,317679,"src/recording.ts",9989,0,"",typescript,selection_command
+29,317794,"src/recording.ts",10007,0,"",typescript,selection_command
+30,318048,"src/recording.ts",10017,0,"",typescript,selection_command
+31,318077,"src/recording.ts",10031,0,"",typescript,selection_command
+32,318441,"src/recording.ts",10030,0,"",typescript,selection_command
+33,318590,"src/recording.ts",10028,0,"",typescript,selection_command
+34,318767,"src/recording.ts",10026,0,"",typescript,selection_command
+35,319213,"src/recording.ts",10026,1,"5",typescript,content
+36,320144,"src/recording.ts",10016,0,"",typescript,selection_command
+37,320398,"src/recording.ts",10002,0,"",typescript,selection_command
+38,320427,"src/recording.ts",9984,0,"",typescript,selection_command
+39,320465,"src/recording.ts",9903,0,"",typescript,selection_command
+40,320493,"src/recording.ts",9818,0,"",typescript,selection_command
+41,320528,"src/recording.ts",9751,0,"",typescript,selection_command
+42,320562,"src/recording.ts",9670,0,"",typescript,selection_command
+43,320594,"src/recording.ts",9649,0,"",typescript,selection_command
+44,320629,"src/recording.ts",9549,0,"",typescript,selection_command
+45,320662,"src/recording.ts",9506,0,"",typescript,selection_command
+46,320698,"src/recording.ts",9475,0,"",typescript,selection_command
+47,321493,"src/recording.ts",9461,0,"",typescript,selection_command
+48,321915,"src/recording.ts",9371,0,"",typescript,selection_command
+49,322434,"src/recording.ts",9300,0,"",typescript,selection_command
+50,329951,"src/recording.ts",9281,0,"",typescript,selection_command
+51,330204,"src/recording.ts",9246,0,"",typescript,selection_command
+52,330230,"src/recording.ts",9199,0,"",typescript,selection_command
+53,330375,"src/recording.ts",9157,0,"",typescript,selection_command
+54,330540,"src/recording.ts",9104,0,"",typescript,selection_command
+55,330712,"src/recording.ts",9157,0,"",typescript,selection_command
+56,330910,"src/recording.ts",9170,0,"",typescript,selection_command
+57,330953,"src/recording.ts",9117,0,"",typescript,selection_command
+58,331166,"src/recording.ts",9083,0,"",typescript,selection_command
+59,331302,"src/recording.ts",9085,0,"",typescript,selection_command
+60,331503,"src/recording.ts",9093,0,"",typescript,selection_command
+61,331610,"src/recording.ts",9127,0,"",typescript,selection_command
+62,331751,"src/recording.ts",9130,0,"",typescript,selection_command
+63,350255,"src/recording.ts",9183,0,"",typescript,selection_command
+64,350406,"src/recording.ts",9225,0,"",typescript,selection_command
+65,350598,"src/recording.ts",9272,0,"",typescript,selection_command
+66,352808,"src/recording.ts",9225,0,"",typescript,selection_command
+67,354872,"src/recording.ts",9183,0,"",typescript,selection_command
+68,359461,"src/recording.ts",9130,0,"",typescript,selection_command
+69,359617,"src/recording.ts",9095,0,"",typescript,selection_command
+70,359673,"src/recording.ts",9117,0,"",typescript,selection_command
+71,359794,"src/recording.ts",9083,0,"",typescript,selection_command
+72,359976,"src/recording.ts",9052,0,"",typescript,selection_command
+73,360191,"src/recording.ts",9031,0,"",typescript,selection_command
+74,360447,"src/recording.ts",8972,0,"",typescript,selection_command
+75,360582,"src/recording.ts",8982,0,"",typescript,selection_command
+76,360831,"src/recording.ts",8984,0,"",typescript,selection_command
+77,360860,"src/recording.ts",8990,0,"",typescript,selection_command
+78,360894,"src/recording.ts",8992,0,"",typescript,selection_command
+79,360925,"src/recording.ts",8993,0,"",typescript,selection_command
+80,360961,"src/recording.ts",9001,0,"",typescript,selection_command
+81,360991,"src/recording.ts",9002,0,"",typescript,selection_command
+82,361027,"src/recording.ts",9010,0,"",typescript,selection_command
+83,361058,"src/recording.ts",9011,0,"",typescript,selection_command
+84,361092,"src/recording.ts",9019,0,"",typescript,selection_command
+85,361127,"src/recording.ts",9021,0,"",typescript,selection_command
+86,361386,"src/recording.ts",9019,0,"",typescript,selection_command
+87,361536,"src/recording.ts",9011,0,"",typescript,selection_command
+88,371148,"src/recording.ts",9031,0,"",typescript,selection_command
+89,371398,"src/recording.ts",9061,0,"",typescript,selection_command
+90,371425,"src/recording.ts",9095,0,"",typescript,selection_command
+91,371594,"src/recording.ts",9148,0,"",typescript,selection_command
+92,372005,"src/recording.ts",9147,0,"",typescript,selection_command
+93,372143,"src/recording.ts",9144,0,"",typescript,selection_command
+94,372278,"src/recording.ts",9142,0,"",typescript,selection_command
+95,372403,"src/recording.ts",9130,0,"",typescript,selection_command
+96,373188,"src/recording.ts",8762,0,"",typescript,selection_command
+97,373475,"src/recording.ts",6202,0,"",typescript,selection_command
+98,401363,"src/recording.ts",6203,0,"",typescript,selection_mouse
+99,471528,"src/recording.ts",6144,72,"    const folderPath = path.dirname(path.join(exportPath, baseFilePath))",typescript,selection_command
+100,472180,"src/recording.ts",6203,0,"",typescript,selection_command
+101,474540,"src/recording.ts",6202,0,"",typescript,selection_command
+102,474720,"src/recording.ts",6200,0,"",typescript,selection_command
+103,474892,"src/recording.ts",6190,0,"",typescript,selection_command
+104,475017,"src/recording.ts",6189,0,"",typescript,selection_command
+105,475195,"src/recording.ts",6185,0,"",typescript,selection_command
+106,475350,"src/recording.ts",6184,0,"",typescript,selection_command
+107,475507,"src/recording.ts",6180,0,"",typescript,selection_command
+108,475678,"src/recording.ts",6179,0,"",typescript,selection_command
+109,475981,"src/recording.ts",6172,0,"",typescript,selection_command
+110,476576,"src/recording.ts",6215,0,"",typescript,selection_command
+111,563556,"src/recording.ts",6148,0,"",typescript,selection_command
+112,563913,"src/recording.ts",6154,0,"",typescript,selection_command
+113,565484,"src/recording.ts",6165,0,"",typescript,selection_command
+114,565741,"src/recording.ts",6167,0,"",typescript,selection_command
+115,565766,"src/recording.ts",6171,0,"",typescript,selection_command295,2976491,"src/recording.ts",28645,0,"\n",typescript,content
+296,2976613,"src/recording.ts",28646,0,"s",typescript,content
+297,2976614,"src/recording.ts",28647,0,"",typescript,selection_keyboard
+298,2976738,"src/recording.ts",28647,0,"u",typescript,content
+299,2976741,"src/recording.ts",28648,0,"",typescript,selection_keyboard
+300,2976798,"src/recording.ts",28648,0,"p",typescript,content
+301,2976801,"src/recording.ts",28649,0,"",typescript,selection_keyboard
+302,2976934,"src/recording.ts",28649,0,"e",typescript,content
+303,2976957,"src/recording.ts",28650,0,"",typescript,selection_keyboard
+304,2976992,"src/recording.ts",28650,0,"r",typescript,content
+305,2976997,"src/recording.ts",28651,0,"",typescript,selection_keyboard
+306,2977200,"src/recording.ts",28651,0,"s",typescript,content
+307,2977202,"src/recording.ts",28652,0,"",typescript,selection_keyboard
+308,2977525,"src/recording.ts",28652,0,"e",typescript,content
+309,2977530,"src/recording.ts",28653,0,"",typescript,selection_keyboard
+310,2977634,"src/recording.ts",28653,0,"c",typescript,content
+311,2977635,"src/recording.ts",28654,0,"",typescript,selection_keyboard
+312,2977825,"src/recording.ts",28654,0,"r",typescript,content
+313,2977826,"src/recording.ts",28655,0,"",typescript,selection_keyboard
+314,2977896,"src/recording.ts",28655,0,"e",typescript,content324,2997670,"src/recording.ts",28659,0,"",typescript,selection_command
+325,2998126,"src/recording.ts",28646,14,"",typescript,content
+326,3005814,"src/recording.ts",28646,0,"s",typescript,content
+327,3005815,"src/recording.ts",28647,0,"",typescript,selection_keyboard
+328,3005883,"src/recording.ts",28647,0,"e",typescript,content
+329,3005888,"src/recording.ts",28648,0,"",typescript,selection_keyboard
+330,3006331,"src/recording.ts",28648,0,"c",typescript,content
+331,3006335,"src/recording.ts",28649,0,"",typescript,selection_keyboard
+332,3006535,"src/recording.ts",28649,0,"r",typescript,content
+333,3006538,"src/recording.ts",28650,0,"",typescript,selection_keyboard
+334,3006577,"src/recording.ts",28650,0,"e",typescript,content
+335,3006581,"src/recording.ts",28651,0,"",typescript,selection_keyboard
+336,3006735,"src/recording.ts",28651,0,"t",typescript,content
+337,3006738,"src/recording.ts",28652,0,"",typescript,selection_keyboard
+338,3006976,"src/recording.ts",28652,0,"p",typescript,content
+339,3006979,"src/recording.ts",28653,0,"",typescript,selection_keyboard
+340,3007037,"src/recording.ts",28653,0,"a",typescript,content380,3137765,"src/recording.ts",28814,0,"",typescript,selection_command
+381,3138176,"src/recording.ts",28647,168,"",typescript,content
+382,3140498,"src/recording.ts",28647,0,"s",typescript,content
+383,3140500,"src/recording.ts",28648,0,"",typescript,selection_keyboard
+384,3140565,"src/recording.ts",28648,0,"e",typescript,content
+385,3140566,"src/recording.ts",28649,0,"",typescript,selection_keyboard
+386,3140680,"src/recording.ts",28649,0,"c",typescript,content
+387,3140684,"src/recording.ts",28650,0,"",typescript,selection_keyboard
+388,3140885,"src/recording.ts",28650,0,"r",typescript,content
+389,3140889,"src/recording.ts",28651,0,"",typescript,selection_keyboard
+390,3140958,"src/recording.ts",28651,0,"e",typescript,content
+391,3140962,"src/recording.ts",28652,0,"",typescript,selection_keyboard
+392,3141094,"src/recording.ts",28652,0,"t",typescript,content
+393,3141097,"src/recording.ts",28653,0,"",typescript,selection_keyboard
+394,3141268,"src/recording.ts",28653,0,"p",typescript,content
+395,3141270,"src/recording.ts",28654,0,"",typescript,selection_keyboard
+396,3141340,"src/recording.ts",28654,0,"a",typescript,content
+397,3141341,"src/recording.ts",28655,0,"",typescript,selection_keyboard
+398,3141375,"src/recording.ts",28655,0,"s",typescript,content
+399,3141376,"src/recording.ts",28656,0,"",typescript,selection_keyboard
+400,3141534,"src/recording.ts",28656,0,"s",typescript,content
+401,3141534,"src/recording.ts",28657,0,"",typescript,selection_keyboard
+402,3141721,"src/recording.ts",28657,0,"w",typescript,content
+403,3141721,"src/recording.ts",28658,0,"",typescript,selection_keyboard
+404,3141921,"src/recording.ts",28658,0,"o",typescript,content
+405,3141921,"src/recording.ts",28659,0,"",typescript,selection_keyboard
+406,3141976,"src/recording.ts",28659,0,"r",typescript,content
+407,3141976,"src/recording.ts",28660,0,"",typescript,selection_keyboard
+408,3142099,"src/recording.ts",28660,0,"d",typescript,content418,3167843,"src/recording.ts",28660,0,"",typescript,selection_command
+419,3168521,"src/recording.ts",28646,15,"",typescript,content
+420,3169730,"src/recording.ts",28645,1,"",typescript,content
+421,3169747,"src/recording.ts",28644,0,"",typescript,selection_command

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-0556481e-9711-4a16-8295-53ec72ff527b1757423165949-2025_09_09-15.06.24.820/source.csv

@@ -0,0 +1,303 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,4,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\nsource .venv/bin/activate\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\ntags=""coinrun_og dynanmics 10m_dataset repro_mihir""\n\nnpy_records_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_coinrun/coinrun_episodes_10m""\n\n# TODO mihir: update the tokenizer and lam checkpoints\ntokenizer_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/tokenizer/train_tokenizer_coinrun_og_reproduction/3466286/tokenizer_1757013407_280000""\nlam_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/lam/train_lam_coinrun_og_reproduction/3466287/lam_1757013407_200000""\n\nCHECKPOINT_DIR=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/dynamics/${job_name}/${slurm_job_id}""\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\nsrun python train_dynamics.py \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --tokenizer_checkpoint=""${tokenizer_ckpt_dir}"" \\n    --patch_size=16 \\n    --lam_checkpoint=""${lam_ckpt_dir}"" \\n    --log_image_interval=1000 \\n    --log \\n    --name=""${job_name}_${slurm_job_id}"" \\n    --tags ${tags} \\n    --entity instant-uv \\n    --project jafar \\n    --data_dir $npy_records_dir \\n    --wandb_id $slurm_job_id\n",shellscript,tab
+2,1161,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"3:06:24 PM [info] Activating crowd-code\n3:06:24 PM [info] Recording started\n3:06:24 PM [info] Initializing git provider using file system watchers...\n3:06:25 PM [info] Git repository found\n3:06:25 PM [info] Git provider initialized successfully\n3:06:25 PM [info] Initial git state: [object Object]\n",Log,tab
+3,26377,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",0,0,"",shellscript,tab
+4,31392,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",1487,0,"",shellscript,selection_mouse
+5,31433,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",1486,0,"",shellscript,selection_command
+6,31987,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",1744,0,"",shellscript,selection_mouse
+7,32002,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",1743,0,"",shellscript,selection_command
+8,32439,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",1855,0,"",shellscript,selection_mouse
+9,32453,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",1854,0,"",shellscript,selection_command
+10,45064,"sample.py",0,0,"from dataclasses import dataclass\nimport time\n\nimport dm_pix as pix\nimport einops\nimport jax\nimport jax.numpy as jnp\nimport numpy as np\nfrom orbax.checkpoint import PyTreeCheckpointer\nfrom PIL import Image, ImageDraw\nimport tyro\n\nfrom genie import Genie\nfrom utils.dataloader import get_dataloader\n\n\n@dataclass\nclass Args:\n    # Experiment\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_resolution: int = 64\n    data_dir: str = ""data/coinrun_episodes""\n    checkpoint: str = """"\n    # Sampling\n    batch_size: int = 1\n    maskgit_steps: int = 25\n    temperature: float = 1.0\n    sample_argmax: bool = True\n    start_frame: int = 0\n    # Tokenizer checkpoint\n    tokenizer_dim: int = 512\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 4\n    tokenizer_num_blocks: int = 8\n    tokenizer_num_heads: int = 8\n    # LAM checkpoint\n    lam_dim: int = 512\n    latent_action_dim: int = 32\n    num_latent_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 8\n    lam_num_heads: int = 8\n    # Dynamics checkpoint\n    dyna_dim: int = 512\n    dyna_num_blocks: int = 12\n    dyna_num_heads: int = 8\n\n\nargs = tyro.cli(Args)\nrng = jax.random.PRNGKey(args.seed)\n\n# --- Load Genie checkpoint ---\ngenie = Genie(\n    # Tokenizer\n    in_dim=args.image_channels,\n    tokenizer_dim=args.tokenizer_dim,\n    latent_patch_dim=args.latent_patch_dim,\n    num_patch_latents=args.num_patch_latents,\n    patch_size=args.patch_size,\n    tokenizer_num_blocks=args.tokenizer_num_blocks,\n    tokenizer_num_heads=args.tokenizer_num_heads,\n    # LAM\n    lam_dim=args.lam_dim,\n    latent_action_dim=args.latent_action_dim,\n    num_latent_actions=args.num_latent_actions,\n    lam_patch_size=args.lam_patch_size,\n    lam_num_blocks=args.lam_num_blocks,\n    lam_num_heads=args.lam_num_heads,\n    # Dynamics\n    dyna_dim=args.dyna_dim,\n    dyna_num_blocks=args.dyna_num_blocks,\n    dyna_num_heads=args.dyna_num_heads,\n)\nrng, _rng = jax.random.split(rng)\nimage_shape = (args.image_resolution, args.image_resolution, args.image_channels)\ndummy_inputs = dict(\n    videos=jnp.zeros((args.batch_size, args.seq_len, *image_shape), dtype=jnp.float32),\n    mask_rng=_rng,\n)\nrng, _rng = jax.random.split(rng)\nparams = genie.init(_rng, dummy_inputs)\nckpt = PyTreeCheckpointer().restore(args.checkpoint)[""model""][""params""][""params""]\nparams[""params""].update(ckpt)\n\n# --- Define autoregressive sampling loop ---\ndef _autoreg_sample(rng, video_batch, action_batch):\n    vid = video_batch[:, : args.start_frame + 1]\n    for frame_idx in range(args.start_frame + 1, args.seq_len):\n        # --- Sample next frame ---\n        print(""Frame"", frame_idx)\n        rng, _rng = jax.random.split(rng)\n        batch = dict(videos=vid, latent_actions=action_batch[:, :frame_idx], rng=_rng)\n        new_frame = genie.apply(\n            params,\n            batch,\n            args.maskgit_steps,\n            args.temperature,\n            args.sample_argmax,\n            method=Genie.sample,\n        )\n        vid = jnp.concatenate([vid, new_frame], axis=1)\n    return vid\n\n\n# --- Get video + latent actions ---\ndataloader = get_dataloader(args.data_dir, args.seq_len, args.batch_size)\nvideo_batch = next(iter(dataloader))\n# Get latent actions from first video only\nfirst_video = video_batch[:1]\nbatch = dict(videos=first_video)\naction_batch = genie.apply(params, batch, False, method=Genie.vq_encode)\naction_batch = action_batch.reshape(1, args.seq_len - 1, 1)\n# Use actions from first video for all videos\naction_batch = jnp.repeat(action_batch, video_batch.shape[0], axis=0)\n\n# --- Sample + evaluate video ---\nvid = _autoreg_sample(rng, video_batch, action_batch)\ngt = video_batch[:, : vid.shape[1]].clip(0, 1).reshape(-1, *video_batch.shape[2:])\nrecon = vid.clip(0, 1).reshape(-1, *vid.shape[2:])\nssim = pix.ssim(gt[:, args.start_frame + 1 :], recon[:, args.start_frame + 1 :]).mean()\nprint(f""SSIM: {ssim}"")\n\n# --- Construct video ---\nfirst_true = (video_batch[0:1] * 255).astype(np.uint8)\nfirst_pred = (vid[0:1] * 255).astype(np.uint8)\nfirst_video_comparison = np.zeros((2, *vid.shape[1:5]), dtype=np.uint8)\nfirst_video_comparison[0] = first_true[:, : vid.shape[1]]\nfirst_video_comparison[1] = first_pred\n# For other videos, only show generated video\nother_preds = (vid[1:] * 255).astype(np.uint8)\nall_frames = np.concatenate([first_video_comparison, other_preds], axis=0)\nflat_vid = einops.rearrange(all_frames, ""n t h w c -> t h (n w) c"")\n\n# --- Save video ---\nimgs = [Image.fromarray(img) for img in flat_vid]\n# Write actions on each frame\nfor img, action in zip(imgs[1:], action_batch[0, :, 0]):\n    d = ImageDraw.Draw(img)\n    d.text((2, 2), f""{action}"", fill=255)\nimgs[0].save(\n    f""generation_{time.time()}.gif"",\n    save_all=True,\n    append_images=imgs[1:],\n    duration=250,\n    loop=0,\n)\n",python,tab
+11,114104,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction.sbatch",0,0,"",shellscript,tab
+12,121068,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-h100.sbatch",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\nsource .venv/bin/activate\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\ntags=""coinrun_og dynanmics 10m_dataset repro_mihir""\n\nnpy_records_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_coinrun/coinrun_episodes_10m""\n\n# TODO mihir: update the tokenizer and lam checkpoints\ntokenizer_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/tokenizer/train_tokenizer_coinrun_og_reproduction/3466286""\nlam_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/lam/train_lam_coinrun_og_reproduction/3466287""\n\nCHECKPOINT_DIR=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/dynamics/${job_name}/${slurm_job_id}""\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\nsrun python train_dynamics.py \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --tokenizer_checkpoint=""${tokenizer_ckpt_dir}"" \\n    --patch_size=16 \\n    --lam_checkpoint=""${lam_ckpt_dir}"" \\n    --log_image_interval=1000 \\n    --log \\n    --name=""${job_name}_${slurm_job_id}"" \\n    --tags ${tags} \\n    --entity instant-uv \\n    --project jafar \\n    --data_dir $npy_records_dir \\n    --wandb_id $slurm_job_id\n",shellscript,tab
+13,121836,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-h100 copy.sbatch",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\nsource .venv/bin/activate\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\ntags=""coinrun_og dynanmics 10m_dataset repro_mihir""\n\nnpy_records_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_coinrun/coinrun_episodes_10m""\n\n# TODO mihir: update the tokenizer and lam checkpoints\ntokenizer_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/tokenizer/train_tokenizer_coinrun_og_reproduction/3466286""\nlam_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/lam/train_lam_coinrun_og_reproduction/3466287""\n\nCHECKPOINT_DIR=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/dynamics/${job_name}/${slurm_job_id}""\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\nsrun python train_dynamics.py \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --tokenizer_checkpoint=""${tokenizer_ckpt_dir}"" \\n    --patch_size=16 \\n    --lam_checkpoint=""${lam_ckpt_dir}"" \\n    --log_image_interval=1000 \\n    --log \\n    --name=""${job_name}_${slurm_job_id}"" \\n    --tags ${tags} \\n    --entity instant-uv \\n    --project jafar \\n    --data_dir $npy_records_dir \\n    --wandb_id $slurm_job_id\n",shellscript,tab
+14,132472,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\nsource .venv/bin/activate\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\ntags=""coinrun_og dynanmics 10m_dataset repro_mihir""\n\nnpy_records_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_coinrun/coinrun_episodes_10m""\n\n# TODO mihir: update the tokenizer and lam checkpoints\ntokenizer_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/tokenizer/train_tokenizer_coinrun_og_reproduction/3466286""\nlam_ckpt_dir=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/lam/train_lam_coinrun_og_reproduction/3466287""\n\nCHECKPOINT_DIR=""/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/dynamics/${job_name}/${slurm_job_id}""\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\nsrun python train_dynamics.py \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --tokenizer_checkpoint=""${tokenizer_ckpt_dir}"" \\n    --patch_size=16 \\n    --lam_checkpoint=""${lam_ckpt_dir}"" \\n    --log_image_interval=1000 \\n    --log \\n    --name=""${job_name}_${slurm_job_id}"" \\n    --tags ${tags} \\n    --entity instant-uv \\n    --project jafar \\n    --data_dir $npy_records_dir \\n    --wandb_id $slurm_job_id\n",shellscript,tab
+15,134029,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",526,0,"",shellscript,selection_mouse
+16,134182,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",525,1,"\n",shellscript,selection_mouse
+17,134183,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",494,32,"\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+18,134184,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",177,349,"#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+19,134185,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",525,1,"\n",shellscript,selection_command
+20,134263,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",140,386,"#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+21,134266,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",119,407,"#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+22,134328,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",93,433,"#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+23,134367,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",67,459,"#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+24,134418,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",39,487,"#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+25,134441,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",21,505,"#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+26,134482,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",20,506,"\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+27,134564,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",0,526,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=2-00:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --partition=accelerated-h100\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/jafar_og_reproduction/dynamics/%x_%j.log\n#SBATCH --job-name=train_dynamics_coinrun_og_reproduction\n\n# Log the sbatch script\ncat $0\n",shellscript,selection_mouse
+28,135908,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",0,527,"",shellscript,content
+29,136537,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",30,0,"",shellscript,selection_command
+30,136723,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",60,0,"",shellscript,selection_command
+31,136866,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",86,0,"",shellscript,selection_command
+32,138675,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",86,26,"",shellscript,content
+33,139297,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",86,27,"",shellscript,content
+34,139687,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",86,1,"",shellscript,content
+35,140062,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",86,52,"",shellscript,content
+36,140378,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",87,0,"",shellscript,selection_command
+37,140861,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",195,0,"",shellscript,selection_command
+38,141070,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",196,0,"",shellscript,selection_command
+39,141206,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",251,0,"",shellscript,selection_command
+40,141391,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",426,0,"",shellscript,selection_command
+41,141527,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",583,0,"",shellscript,selection_command
+42,141750,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",584,0,"",shellscript,selection_command
+43,141946,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",583,0,"",shellscript,selection_command
+44,142105,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",426,0,"",shellscript,selection_command
+45,142298,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",251,0,"",shellscript,selection_command
+46,142912,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",196,0,"",shellscript,selection_command
+47,144098,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",251,0,"",shellscript,selection_command
+48,144303,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",426,0,"",shellscript,selection_command
+49,144472,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",583,0,"",shellscript,selection_command
+50,144650,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",584,0,"",shellscript,selection_command
+51,144982,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",734,0,"",shellscript,selection_command
+52,147289,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",584,0,"",shellscript,selection_command
+53,147485,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",583,0,"",shellscript,selection_command
+54,147608,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",426,0,"",shellscript,selection_command
+55,148296,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",582,0,"\n",shellscript,content
+56,150043,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",583,0,"\n",shellscript,content
+57,150288,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",584,0,"d",shellscript,content
+58,150289,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",585,0,"",shellscript,selection_keyboard
+59,150526,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",585,0,"y",shellscript,content
+60,150526,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",586,0,"",shellscript,selection_keyboard
+61,150592,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",586,0,"n",shellscript,content
+62,150593,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",587,0,"",shellscript,selection_keyboard
+63,150939,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",587,0,"_",shellscript,content
+64,150940,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",588,0,"",shellscript,selection_keyboard
+65,151201,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",588,0,"c",shellscript,content
+66,151202,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",589,0,"",shellscript,selection_keyboard
+67,151306,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",589,0,"k",shellscript,content
+68,151307,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",590,0,"",shellscript,selection_keyboard
+69,151515,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",590,0,"p",shellscript,content
+70,151516,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",591,0,"",shellscript,selection_keyboard
+71,151648,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",591,0,"t",shellscript,content
+72,151649,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",592,0,"",shellscript,selection_keyboard
+73,151860,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",592,0,"_",shellscript,content
+74,151861,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",593,0,"",shellscript,selection_keyboard
+75,152047,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",593,0,"d",shellscript,content
+76,152048,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",594,0,"",shellscript,selection_keyboard
+77,152275,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",594,0,"r",shellscript,content
+78,152276,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",595,0,"",shellscript,selection_keyboard
+79,152595,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",594,1,"",shellscript,content
+80,152727,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",594,0,"i",shellscript,content
+81,152728,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",595,0,"",shellscript,selection_keyboard
+82,152856,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",595,0,"r",shellscript,content
+83,152857,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",596,0,"",shellscript,selection_keyboard
+84,153300,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",596,0,"=",shellscript,content
+85,153301,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",597,0,"",shellscript,selection_keyboard
+86,153591,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",597,0,"""",shellscript,content
+87,153592,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",598,0,"",shellscript,selection_keyboard
+88,154498,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",598,0,"""",shellscript,content
+89,154499,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",599,0,"",shellscript,selection_keyboard
+90,155082,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",598,0,"",shellscript,selection_command
+91,155269,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",600,0,"",shellscript,selection_command
+92,155678,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",615,0,"",shellscript,selection_command
+93,158939,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",765,0,"",shellscript,selection_command
+94,159460,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",776,0,"",shellscript,selection_command
+95,159508,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",791,0,"",shellscript,selection_command
+96,159541,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",794,0,"",shellscript,selection_command
+97,159576,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",809,0,"",shellscript,selection_command
+98,159604,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",841,0,"",shellscript,selection_command
+99,159642,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",874,0,"",shellscript,selection_command
+100,159675,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",927,0,"",shellscript,selection_command
+101,159714,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",949,0,"",shellscript,selection_command
+102,159715,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",990,0,"",shellscript,selection_command
+103,159736,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1018,0,"",shellscript,selection_command
+104,159759,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1034,0,"",shellscript,selection_command
+105,159818,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1077,0,"",shellscript,selection_command
+106,159819,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1098,0,"",shellscript,selection_command
+107,159858,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1124,0,"",shellscript,selection_command
+108,160048,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1146,0,"",shellscript,selection_command
+109,160519,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1124,0,"",shellscript,selection_command
+110,161053,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1098,0,"",shellscript,selection_command
+111,161086,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1077,0,"",shellscript,selection_command
+112,161100,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1034,0,"",shellscript,selection_command
+113,161145,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1018,0,"",shellscript,selection_command
+114,161183,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",990,0,"",shellscript,selection_command
+115,161207,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",949,0,"",shellscript,selection_command
+116,161253,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",927,0,"",shellscript,selection_command
+117,161279,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",874,0,"",shellscript,selection_command
+118,161312,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",841,0,"",shellscript,selection_command
+119,161345,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",809,0,"",shellscript,selection_command
+120,161346,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",794,0,"",shellscript,selection_command
+121,161894,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",809,0,"",shellscript,selection_command
+122,162235,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",808,0,"",shellscript,selection_command
+123,162461,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",807,0,"",shellscript,selection_command
+124,162986,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",807,14,"",shellscript,content
+125,163491,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",807,0,"s",shellscript,content
+126,163494,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",808,0,"",shellscript,selection_keyboard
+127,163681,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",808,0,"a",shellscript,content
+128,163682,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",809,0,"",shellscript,selection_keyboard
+129,163729,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",809,0,"m",shellscript,content
+130,163730,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",810,0,"",shellscript,selection_keyboard
+131,163921,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",810,0,"p",shellscript,content
+132,163922,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",811,0,"",shellscript,selection_keyboard
+133,163986,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",811,0,"l",shellscript,content
+134,163987,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",812,0,"",shellscript,selection_keyboard
+135,164067,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",812,0,"e",shellscript,content
+136,164068,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",813,0,"",shellscript,selection_keyboard
+137,164613,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",813,0,"h",shellscript,content
+138,164614,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",814,0,"",shellscript,selection_keyboard
+139,165128,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",813,1,"",shellscript,content
+140,165267,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",812,0,"",shellscript,selection_command
+141,165432,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",811,0,"",shellscript,selection_command
+142,165844,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",810,0,"",shellscript,selection_command
+143,165918,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",809,0,"",shellscript,selection_command
+144,165923,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",808,0,"",shellscript,selection_command
+145,165944,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",807,0,"",shellscript,selection_command
+146,166026,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",806,0,"",shellscript,selection_command
+147,166027,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",805,0,"",shellscript,selection_command
+148,166065,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",804,0,"",shellscript,selection_command
+149,166100,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",803,0,"",shellscript,selection_command
+150,166121,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",802,0,"",shellscript,selection_command
+151,166143,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",801,0,"",shellscript,selection_command
+152,166179,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",800,0,"",shellscript,selection_command
+153,166191,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",799,0,"",shellscript,selection_command
+154,166229,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",798,0,"",shellscript,selection_command
+155,166341,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",797,0,"",shellscript,selection_command
+156,166466,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",796,0,"",shellscript,selection_command
+157,166647,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",795,0,"",shellscript,selection_command
+158,166888,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",795,5,"",shellscript,content
+159,167113,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",795,1,"",shellscript,content
+160,168517,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",795,0,"p",shellscript,content
+161,168518,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",796,0,"",shellscript,selection_keyboard
+162,168632,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",795,0,"",shellscript,selection_command
+163,169118,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",814,0,"",shellscript,selection_command
+164,169254,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",847,0,"",shellscript,selection_command
+165,169655,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",814,0,"",shellscript,selection_command
+166,172983,"sample.py",0,0,"",python,tab
+167,175230,"sample.py",0,0,"",python,tab
+168,180098,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",0,0,"",shellscript,tab
+169,180099,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1006,0,"",shellscript,selection_mouse
+170,180184,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",1005,0,"",shellscript,selection_command
+171,180623,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",832,0,"",shellscript,selection_mouse
+172,182730,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",831,0,"",shellscript,selection_command
+173,183050,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",830,0,"",shellscript,selection_command
+174,189074,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",829,0,"",shellscript,selection_command
+175,189301,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",828,0,"",shellscript,selection_command
+176,189498,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",827,0,"",shellscript,selection_command
+177,189639,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",826,0,"",shellscript,selection_command
+178,189779,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",825,0,"",shellscript,selection_command
+179,189928,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",824,0,"",shellscript,selection_command
+180,190112,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",823,0,"",shellscript,selection_command
+181,190242,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",822,0,"",shellscript,selection_command
+182,190401,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",821,0,"",shellscript,selection_command
+183,190622,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",820,0,"",shellscript,selection_command
+184,190833,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",820,9,"",shellscript,content
+185,191214,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",820,0,"c",shellscript,content
+186,191215,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",821,0,"",shellscript,selection_keyboard
+187,191337,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",821,0,"h",shellscript,content
+188,191338,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",822,0,"",shellscript,selection_keyboard
+189,191593,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",822,0,"e",shellscript,content
+190,191594,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",823,0,"",shellscript,selection_keyboard
+191,191704,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",823,0,"k",shellscript,content
+192,191705,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",824,0,"",shellscript,selection_keyboard
+193,192297,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",823,1,"",shellscript,content
+194,192388,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",823,0,"c",shellscript,content
+195,192389,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",824,0,"",shellscript,selection_keyboard
+196,192490,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",824,0,"k",shellscript,content
+197,192491,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",825,0,"",shellscript,selection_keyboard
+198,192723,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",825,0,"p",shellscript,content
+199,192724,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",826,0,"",shellscript,selection_keyboard
+200,192919,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",826,0,"o",shellscript,content
+201,192919,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",827,0,"",shellscript,selection_keyboard
+202,193118,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",827,0,"i",shellscript,content
+203,193119,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",828,0,"",shellscript,selection_keyboard
+204,193189,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",828,0,"n",shellscript,content
+205,193190,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",829,0,"",shellscript,selection_keyboard
+206,193272,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",829,0,"t",shellscript,content
+207,193273,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",830,0,"",shellscript,selection_keyboard
+208,193611,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",830,0," ",shellscript,content
+209,193611,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",831,0,"",shellscript,selection_keyboard
+210,196125,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",831,1,"",shellscript,content
+211,196409,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",831,14,"",shellscript,content
+212,197941,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",831,0,"$",shellscript,content
+213,197942,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",832,0,"",shellscript,selection_keyboard
+214,198423,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",832,0,"d",shellscript,content
+215,198424,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",833,0,"",shellscript,selection_keyboard
+216,198658,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",833,0,"y",shellscript,content
+217,198659,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",834,0,"",shellscript,selection_keyboard
+218,198731,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",834,0,"n",shellscript,content
+219,198732,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",835,0,"",shellscript,selection_keyboard
+220,199216,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",832,3,"dyn_ckpt_dir",shellscript,content
+221,199987,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",843,0,"",shellscript,selection_command
+222,200118,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",812,0,"",shellscript,selection_command
+223,200307,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",794,0,"",shellscript,selection_command
+224,200464,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",792,0,"",shellscript,selection_command
+225,200583,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",776,0,"",shellscript,selection_command
+226,200723,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",774,0,"",shellscript,selection_command
+227,201141,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",630,0,"",shellscript,selection_command
+228,201401,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",601,175,"",shellscript,content
+229,202345,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",601,1,"",shellscript,content
+230,203851,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",601,17,"",shellscript,content
+231,204314,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",601,1,"",shellscript,content
+232,204656,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",620,0,"",shellscript,selection_command
+233,204835,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",653,0,"",shellscript,selection_command
+234,206551,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",653,53,"",shellscript,content
+235,206599,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",657,0,"",shellscript,selection_command
+236,209296,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+237,210151,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,41,"",shellscript,content
+238,210200,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+239,210689,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,32,"",shellscript,content
+240,210745,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+241,211152,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,12,"",shellscript,content
+242,211204,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+243,212244,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,43,"",shellscript,content
+244,212293,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+245,213392,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,21,"",shellscript,content
+246,213453,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+247,214097,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,26,"",shellscript,content
+248,214188,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+249,214702,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,22,"",shellscript,content
+250,214731,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",679,0,"",shellscript,selection_command
+251,215869,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",713,0,"",shellscript,selection_command
+252,216376,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",709,29,"",shellscript,content
+253,216926,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,0,"",shellscript,selection_command
+254,217100,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",653,0,"",shellscript,selection_command
+255,217411,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",620,0,"",shellscript,selection_command
+256,217568,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",601,0,"",shellscript,selection_command
+257,217800,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",620,0,"",shellscript,selection_command
+258,217930,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",653,0,"",shellscript,selection_command
+259,218077,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",675,0,"",shellscript,selection_command
+260,219507,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",708,0,"",shellscript,selection_command
+261,219598,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",707,1,"",shellscript,content
+262,220000,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",706,1,"",shellscript,content
+263,220121,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",705,0,"",shellscript,selection_command
+264,297473,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-h100.sbatch",0,0,"",shellscript,tab
+265,308728,"TERMINAL",0,0,"cd $ws_dir",,terminal_command
+266,315700,"TERMINAL",0,0,"cd checkpoints/mihir/jafar_og_reproduction/dynamics/",,terminal_command
+267,315988,"TERMINAL",0,0,"ls",,terminal_command
+268,317773,"TERMINAL",0,0,"cd train_dynamics_coinrun_og_reproduction/",,terminal_command
+269,318295,"TERMINAL",0,0,"l",,terminal_command
+270,319555,"TERMINAL",0,0,"ls",,terminal_command
+271,335100,"TERMINAL",0,0,"CD 3472617",,terminal_command
+272,338412,"TERMINAL",0,0,"cd 3472617",,terminal_command
+273,339785,"TERMINAL",0,0,"ls",,terminal_command
+274,339826,"TERMINAL",0,0,"]633;C",,terminal_output
+275,339964,"TERMINAL",0,0,"genie_1757238544_100000  genie_1757238544_125000  genie_1757238544_150000  genie_1757238544_175000  genie_1757238544_200000  genie_1757238544_25000  genie_1757238544_50000  genie_1757238544_75000\r\n]0;tum_cte0515@hkn1990:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/dynamics/train_dynamics_coinrun_og_reproduction/3472617",,terminal_output
+276,346764,"TERMINAL",0,0,"cd genie_1757238544_200000/",,terminal_command
+277,347634,"TERMINAL",0,0,"pwd",,terminal_command
+278,352599,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",0,0,"",shellscript,tab
+279,355300,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",598,0,"",shellscript,selection_mouse
+280,356455,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",598,0,"/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/mihir/jafar_og_reproduction/dynamics/train_dynamics_coinrun_og_reproduction/3472617/genie_1757238544_200000",shellscript,content
+281,844805,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",0,0,"",shellscript,tab
+282,969620,"slurm/jobs/mihir/horeka/jafar_og_reproduction/og_coinrun_dynamics_reproduction-sample.sh",882,0,"",shellscript,selection_mouse
+283,1631269,"train_dynamics.py",0,0,"from dataclasses import dataclass, field\nimport os\nimport time\n\nimport einops\nfrom flax.training import orbax_utils\nfrom flax.training.train_state import TrainState\nimport optax\nimport orbax\nimport numpy as np\nimport jax\nimport jax.numpy as jnp\nimport tyro\nimport wandb\n\nfrom genie import Genie, restore_genie_components\nfrom utils.dataloader import get_dataloader\n\nts = int(time.time())\n\n\n@dataclass\nclass Args:\n    # Experiment\n    num_steps: int = 200_000\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_resolution: int = 64\n    data_dir: str = ""data/coinrun_episodes""\n    # Optimization\n    batch_size: int = 36\n    min_lr: float = 3e-6\n    max_lr: float = 3e-5\n    warmup_steps: int = 5000\n    # Tokenizer\n    tokenizer_dim: int = 512\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 4\n    tokenizer_num_blocks: int = 8\n    tokenizer_num_heads: int = 8\n    tokenizer_checkpoint: str = """"\n    # LAM\n    lam_dim: int = 512\n    latent_action_dim: int = 32\n    num_latent_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 8\n    lam_num_heads: int = 8\n    lam_checkpoint: str = """"\n    # Dynamics\n    dyna_dim: int = 512\n    dyna_num_blocks: int = 12\n    dyna_num_heads: int = 8\n    dropout: float = 0.0\n    mask_limit: float = 0.5\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 25000\n    log_gradients: bool = False\n    name: str = """"\n    tags: list[str] = field(default_factory=lambda: [""dynamics""])\n    wandb_id: str = """"\n\nargs = tyro.cli(Args)\n\n\ndef dynamics_loss_fn(params, state, inputs):\n    """"""Compute masked dynamics loss""""""\n    outputs = state.apply_fn(\n        params, inputs, training=True, rngs={""dropout"": inputs[""dropout_rng""]}\n    )\n    mask = outputs[""mask""]\n    ce_loss = optax.softmax_cross_entropy_with_integer_labels(\n        outputs[""token_logits""], outputs[""video_tokens""]\n    )\n    ce_loss = (mask * ce_loss).sum() / mask.sum()\n    acc = outputs[""token_logits""].argmax(-1) == outputs[""video_tokens""]\n    acc = (mask * acc).sum() / mask.sum()\n    select_probs = jax.nn.softmax(outputs[""token_logits""])\n    metrics = dict(\n        cross_entropy_loss=ce_loss,\n        masked_token_accuracy=acc,\n        select_logit=outputs[""token_logits""].max(-1).mean(),\n        select_p=select_probs.max(-1).mean(),\n        entropy=jax.scipy.special.entr(select_probs).sum(-1).mean(),\n    )\n    return ce_loss, (outputs[""recon""], metrics)\n\n\n@jax.jit\ndef train_step(state, inputs):\n    """"""Update state and compute metrics""""""\n    grad_fn = jax.value_and_grad(dynamics_loss_fn, has_aux=True, allow_int=True)\n    (loss, (recon, metrics)), grads = grad_fn(state.params, state, inputs)\n    state = state.apply_gradients(grads=grads)\n    if args.log_gradients:\n        metrics[""gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""dynamics""]\n        )\n    return state, loss, recon, metrics\n\n\nif __name__ == ""__main__"":\n    rng = jax.random.PRNGKey(args.seed)\n    if args.log:\n        wandb_init_kwargs = {\n            ""entity"": args.entity,\n            ""project"": args.project,\n            ""name"": args.name,\n            ""tags"": args.tags,\n            ""group"": ""debug"",\n            ""config"": args,\n        }\n        if args.wandb_id:\n            wandb_init_kwargs.update(\n                {\n                    ""id"": args.wandb_id,\n                    ""resume"": ""allow"",\n                }\n            )\n        wandb.init(**wandb_init_kwargs)\n\n    # --- Initialize model ---\n    genie = Genie(\n        # Tokenizer\n        in_dim=args.image_channels,\n        tokenizer_dim=args.tokenizer_dim,\n        latent_patch_dim=args.latent_patch_dim,\n        num_patch_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        tokenizer_num_blocks=args.tokenizer_num_blocks,\n        tokenizer_num_heads=args.tokenizer_num_heads,\n        # LAM\n        lam_dim=args.lam_dim,\n        latent_action_dim=args.latent_action_dim,\n        num_latent_actions=args.num_latent_actions,\n        lam_patch_size=args.lam_patch_size,\n        lam_num_blocks=args.lam_num_blocks,\n        lam_num_heads=args.lam_num_heads,\n        # Dynamics\n        dyna_dim=args.dyna_dim,\n        dyna_num_blocks=args.dyna_num_blocks,\n        dyna_num_heads=args.dyna_num_heads,\n        dropout=args.dropout,\n        mask_limit=args.mask_limit,\n    )\n    rng, _rng = jax.random.split(rng)\n    image_shape = (args.image_resolution, args.image_resolution, args.image_channels)\n    dummy_inputs = dict(\n        videos=jnp.zeros(\n            (args.batch_size, args.seq_len, *image_shape), dtype=jnp.float32\n        ),\n        mask_rng=_rng,\n    )\n    rng, _rng = jax.random.split(rng)\n    init_params = genie.init(_rng, dummy_inputs)\n    init_params = restore_genie_components(\n        init_params, args.tokenizer_checkpoint, args.lam_checkpoint\n    )\n\n    # --- Initialize optimizer ---\n    lr_schedule = optax.warmup_cosine_decay_schedule(\n        args.min_lr, args.max_lr, args.warmup_steps, args.num_steps\n    )\n    tx = optax.adamw(learning_rate=lr_schedule, b1=0.9, b2=0.9, weight_decay=1e-4)\n    train_state = TrainState.create(apply_fn=genie.apply, params=init_params, tx=tx)\n\n    # --- TRAIN LOOP ---\n    dataloader = get_dataloader(args.data_dir, args.seq_len, args.batch_size)\n    step = 0\n    while step < args.num_steps:\n        for videos in dataloader:\n            # --- Train step ---\n            rng, _rng, _mask_rng = jax.random.split(rng, 3)\n            inputs = dict(\n                videos=videos,\n                action=jnp.zeros((args.batch_size, args.seq_len), dtype=jnp.float32),\n                dropout_rng=_rng,\n                mask_rng=_mask_rng,\n            )\n            train_state, loss, recon, metrics = train_step(train_state, inputs)\n            print(f""Step {step}, loss: {loss}"")\n            step += 1\n\n            # --- Logging ---\n            if args.log:\n                if step % args.log_interval == 0:\n                    wandb.log({""loss"": loss, ""step"": step, **metrics})\n                if step % args.log_image_interval == 0:\n                    gt_seq = inputs[""videos""][0]\n                    recon_seq = recon[0].clip(0, 1)\n                    comparison_seq = jnp.concatenate((gt_seq, recon_seq), axis=1)\n                    comparison_seq = einops.rearrange(\n                        comparison_seq * 255, ""t h w c -> h (t w) c""\n                    )\n                    log_images = dict(\n                        image=wandb.Image(np.asarray(gt_seq[15])),\n                        recon=wandb.Image(np.asarray(recon_seq[15])),\n                        true_vs_recon=wandb.Image(\n                            np.asarray(comparison_seq.astype(np.uint8))\n                        ),\n                    )\n                    wandb.log(log_images)\n                if step % args.log_checkpoint_interval == 0:\n                    ckpt = {""model"": train_state}\n                    orbax_checkpointer = orbax.checkpoint.PyTreeCheckpointer()\n                    save_args = orbax_utils.save_args_from_target(ckpt)\n                    orbax_checkpointer.save(\n                        os.path.join(os.getcwd(), args.ckpt_dir, f""genie_{ts}_{step}""),\n                        ckpt,\n                        save_args=save_args,\n                    )\n            if step >= args.num_steps:\n                break\n",python,tab
+284,1635449,"train_tokenizer.py",0,0,"from dataclasses import dataclass, field\n\nimport os\nimport time\n\nimport einops\nfrom flax.training import orbax_utils\nfrom flax.training.train_state import TrainState\nimport optax\nimport orbax\nfrom orbax.checkpoint import PyTreeCheckpointer\nimport numpy as np\nimport dm_pix as pix\nimport jax\nimport jax.numpy as jnp\nimport tyro\nimport wandb\n\nfrom models.tokenizer import TokenizerVQVAE\nfrom utils.dataloader import get_dataloader\n\nts = int(time.time())\n\n\n@dataclass\nclass Args:\n    # Experiment\n    num_steps: int = 300_000\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_resolution: int = 64\n    data_dir: str = ""data/coinrun_episodes""\n    checkpoint: str = """"\n    # Optimization\n    vq_beta: float = 0.25\n    batch_size: int = 48\n    min_lr: float = 3e-4\n    max_lr: float = 3e-4\n    warmup_steps: int = 10000\n    # Tokenizer\n    model_dim: int = 512\n    latent_dim: int = 32\n    num_latents: int = 1024\n    patch_size: int = 4\n    num_blocks: int = 8\n    num_heads: int = 8\n    dropout: float = 0.0\n    codebook_dropout: float = 0.01\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 10000\n    log_gradients: bool = False\n    name: str = """"\n    tags: list[str] = field(default_factory=lambda: [""tokenizer""])\n    wandb_id: str = """"\n\n\nargs = tyro.cli(Args)\n\n\ndef tokenizer_loss_fn(params, state, inputs):\n    # --- Compute loss ---\n    outputs = state.apply_fn(\n        params, inputs, training=True, rngs={""dropout"": inputs[""rng""]}\n    )\n    mse = jnp.square(inputs[""videos""] - outputs[""recon""]).mean()\n    q_loss = jnp.square(jax.lax.stop_gradient(outputs[""emb""]) - outputs[""z""]).mean()\n    commitment_loss = jnp.square(\n        outputs[""emb""] - jax.lax.stop_gradient(outputs[""z""])\n    ).mean()\n    loss = mse + q_loss + args.vq_beta * commitment_loss\n\n    # --- Compute validation metrics ---\n    gt = inputs[""videos""].clip(0, 1).reshape(-1, *inputs[""videos""].shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = pix.psnr(gt, recon).mean()\n    ssim = pix.ssim(gt, recon).mean()\n    _, index_counts = jnp.unique_counts(\n        jnp.ravel(outputs[""indices""]), size=args.num_latents, fill_value=0\n    )\n    codebook_usage = (index_counts != 0).mean()\n    metrics = dict(\n        loss=loss,\n        mse=mse,\n        q_loss=q_loss,\n        commitment_loss=commitment_loss,\n        psnr=psnr,\n        ssim=ssim,\n        codebook_usage=codebook_usage,\n    )\n    return loss, (outputs[""recon""], metrics)\n\n\n@jax.jit\ndef train_step(state, inputs):\n    grad_fn = jax.value_and_grad(tokenizer_loss_fn, has_aux=True, allow_int=True)\n    (loss, (recon, metrics)), grads = grad_fn(state.params, state, inputs)\n    state = state.apply_gradients(grads=grads)\n    if args.log_gradients:\n        metrics[""encoder_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""encoder""]\n        )\n        metrics[""vq_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""vq""]\n        )\n        metrics[""decoder_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""decoder""]\n        )\n    return state, loss, recon, metrics\n\n\nif __name__ == ""__main__"":\n    rng = jax.random.PRNGKey(args.seed)\n    if args.log:\n        wandb_init_kwargs = {\n            ""entity"": args.entity,\n            ""project"": args.project,\n            ""name"": args.name,\n            ""tags"": args.tags,\n            ""group"": ""debug"",\n            ""config"": args,\n        }\n        if args.wandb_id:\n            wandb_init_kwargs.update(\n                {\n                    ""id"": args.wandb_id,\n                    ""resume"": ""allow"",\n                }\n            )\n        wandb.init(**wandb_init_kwargs)\n\n    # --- Initialize model ---\n    tokenizer = TokenizerVQVAE(\n        in_dim=args.image_channels,\n        model_dim=args.model_dim,\n        latent_dim=args.latent_dim,\n        num_latents=args.num_latents,\n        patch_size=args.patch_size,\n        num_blocks=args.num_blocks,\n        num_heads=args.num_heads,\n        dropout=args.dropout,\n        codebook_dropout=args.codebook_dropout,\n    )\n    rng, _rng = jax.random.split(rng)\n    image_shape = (args.image_resolution, args.image_resolution, args.image_channels)\n    inputs = dict(\n        videos=jnp.zeros(\n            (args.batch_size, args.seq_len, *image_shape), dtype=jnp.float32\n        ),\n    )\n    init_params = tokenizer.init(_rng, inputs)\n\n    # --- Load checkpoint ---\n    step = 0\n    if args.checkpoint:\n        init_params[""params""].update(\n            PyTreeCheckpointer().restore(args.checkpoint)[""model""][""params""][""params""]\n        )\n        # Assume checkpoint is of the form tokenizer_<timestamp>_<step>\n        step += int(args.checkpoint.split(""_"")[-1])\n\n    # --- Initialize optimizer ---\n    lr_schedule = optax.warmup_cosine_decay_schedule(\n        args.min_lr, args.max_lr, args.warmup_steps, args.num_steps\n    )\n    tx = optax.adamw(learning_rate=lr_schedule, b1=0.9, b2=0.9, weight_decay=1e-4)\n    train_state = TrainState.create(apply_fn=tokenizer.apply, params=init_params, tx=tx)\n\n    # --- TRAIN LOOP ---\n    dataloader = get_dataloader(args.data_dir, args.seq_len, args.batch_size)\n    while step < args.num_steps:\n        for videos in dataloader:\n            # --- Train step ---\n            rng, _rng = jax.random.split(rng)\n            inputs = dict(videos=videos, rng=_rng)\n            train_state, loss, recon, metrics = train_step(train_state, inputs)\n            print(f""Step {step}, loss: {loss}"")\n            step += 1\n\n            # --- Logging ---\n            if args.log:\n                if step % args.log_interval == 0:\n                    wandb.log({""loss"": loss, ""step"": step, **metrics})\n                if step % args.log_image_interval == 0:\n                    gt_seq = inputs[""videos""][0]\n                    recon_seq = recon[0].clip(0, 1)\n                    comparison_seq = jnp.concatenate((gt_seq, recon_seq), axis=1)\n                    comparison_seq = einops.rearrange(\n                        comparison_seq * 255, ""t h w c -> h (t w) c""\n                    )\n                    log_images = dict(\n                        image=wandb.Image(np.asarray(gt_seq[0])),\n                        recon=wandb.Image(np.asarray(recon_seq[0])),\n                        true_vs_recon=wandb.Image(\n                            np.asarray(comparison_seq.astype(np.uint8))\n                        ),\n                    )\n                    wandb.log(log_images)\n                if step % args.log_checkpoint_interval == 0:\n                    ckpt = {""model"": train_state}\n                    orbax_checkpointer = orbax.checkpoint.PyTreeCheckpointer()\n                    save_args = orbax_utils.save_args_from_target(ckpt)\n                    orbax_checkpointer.save(\n                        os.path.join(\n                            os.getcwd(), args.ckpt_dir, f""tokenizer_{ts}_{step}""\n                        ),\n                        ckpt,\n                        save_args=save_args,\n                    )\n            if step >= args.num_steps:\n                break\n",python,tab
+285,1645976,"train_tokenizer.py",970,0,"",python,selection_mouse
+286,1646085,"train_tokenizer.py",963,10,"num_blocks",python,selection_mouse
+287,1659041,"train_tokenizer.py",4566,0,"",python,selection_mouse
+288,1659164,"train_tokenizer.py",4566,0,"num_blocks",python,content
+289,1659166,"train_tokenizer.py",4576,0,"",python,selection_keyboard
+290,1660752,"train_tokenizer.py",4566,10,"",python,content
+291,1662024,"train_tokenizer.py",3944,0,"",python,selection_mouse
+292,1662117,"train_tokenizer.py",3944,0,"num_blocks",python,content
+293,1662119,"train_tokenizer.py",3954,0,"",python,selection_keyboard
+294,1664011,"train_tokenizer.py",3944,10,"",python,content
+295,1670702,"train_tokenizer.py",3915,0,"",python,selection_mouse
+296,1671454,"models/tokenizer.py",0,0,"from typing import Dict, Any, Tuple\n\nimport flax.linen as nn\n\nfrom utils.preprocess import patchify, unpatchify\nfrom utils.nn import STTransformer, VectorQuantizer\n\n\nclass TokenizerVQVAE(nn.Module):\n    """"""ST-ViVit VQ-VAE""""""\n\n    in_dim: int\n    model_dim: int\n    latent_dim: int\n    num_latents: int\n    patch_size: int\n    num_blocks: int\n    num_heads: int\n    dropout: float\n    codebook_dropout: float\n\n    def setup(self):\n        self.encoder = STTransformer(\n            self.model_dim,\n            self.latent_dim,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n        )\n        self.vq = VectorQuantizer(\n            self.latent_dim,\n            self.num_latents,\n            self.codebook_dropout,\n        )\n        self.out_dim = self.in_dim * self.patch_size**2\n        self.decoder = STTransformer(\n            self.model_dim,\n            self.out_dim,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n        )\n\n    def __call__(self, batch: Dict[str, Any], training: bool = True) -> Dict[str, Any]:\n        H, W = batch[""videos""].shape[2:4]\n        outputs = self.vq_encode(batch[""videos""], training)\n        recon = self.decoder(outputs[""z_q""])  # (B, T, H_down * W_down, C)\n        recon = nn.sigmoid(recon)\n        outputs[""recon""] = unpatchify(recon, self.patch_size, H, W)\n        return outputs\n\n    def vq_encode(self, videos: Any, training: bool = True) -> Dict[str, Any]:\n        # --- Preprocess + encode ---\n        B, T = videos.shape[:2]\n        x = patchify(videos, self.patch_size)\n        N = x.shape[2]\n        x = self.encoder(x)  # (B, T, N, E)\n\n        # --- Vector quantize ---\n        x = x.reshape(B * T * N, self.latent_dim)\n        z_q, z, emb, indices = self.vq(x, training)\n        z_q = z_q.reshape(B, T, N, self.latent_dim)\n        indices = indices.reshape(B, T, N)\n        return dict(z_q=z_q, z=z, emb=emb, indices=indices)\n\n    def decode(self, indices: Any, video_hw: Tuple[int]):\n        z = self.vq.codebook[indices]\n        recon = self.decoder(z)\n        recon = nn.sigmoid(recon)\n        return unpatchify(recon, self.patch_size, *video_hw)\n",python,tab
+297,1671455,"models/tokenizer.py",172,0,"",python,selection_command
+298,1672870,"models/tokenizer.py",330,0,"",python,selection_mouse
+299,1673052,"models/tokenizer.py",326,10,"num_blocks",python,selection_mouse
+300,1712759,"train_tokenizer.py",0,0,"",python,tab
+301,1719310,"models/dynamics.py",0,0,"from typing import Dict, Any\n\nimport jax\nimport jax.numpy as jnp\nimport flax.linen as nn\n\nfrom utils.nn import STTransformer\n\n\nclass DynamicsMaskGIT(nn.Module):\n    """"""MaskGIT dynamics model""""""\n\n    model_dim: int\n    num_latents: int\n    num_blocks: int\n    num_heads: int\n    dropout: float\n    mask_limit: float\n\n    def setup(self):\n        self.dynamics = STTransformer(\n            self.model_dim,\n            self.num_latents,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n        )\n        self.patch_embed = nn.Embed(self.num_latents, self.model_dim)\n        self.mask_token = self.param(\n            ""mask_token"",\n            nn.initializers.lecun_uniform(),\n            (1, 1, 1, self.model_dim),\n        )\n        self.action_up = nn.Dense(self.model_dim)\n\n    def __call__(self, batch: Dict[str, Any], training: bool = True) -> Dict[str, Any]:\n        # --- Mask videos ---\n        vid_embed = self.patch_embed(batch[""video_tokens""])\n        if training:\n            rng1, rng2 = jax.random.split(batch[""mask_rng""])\n            mask_prob = jax.random.uniform(rng1, minval=self.mask_limit)\n            mask = jax.random.bernoulli(rng2, mask_prob, vid_embed.shape[:-1])\n            mask = mask.at[:, 0].set(False)\n            vid_embed = jnp.where(jnp.expand_dims(mask, -1), self.mask_token, vid_embed)\n        else:\n            mask = None\n\n        # --- Predict transition ---\n        act_embed = self.action_up(batch[""latent_actions""])\n        vid_embed += jnp.pad(act_embed, ((0, 0), (1, 0), (0, 0), (0, 0)))\n        logits = self.dynamics(vid_embed)\n        return dict(token_logits=logits, mask=mask)\n",python,tab
+302,1730133,"train_dynamics.py",0,0,"",python,tab

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-12b522dd-8518-4c62-b207-ca1ed4ce90571752782954186-2025_07_17-22.10.14.626/source.csv

@@ -0,0 +1,60 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,4,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",0,0,"from dataclasses import dataclass, field\nimport os\n\nimport einops\nfrom flax.training.train_state import TrainState\nfrom jax.sharding import Mesh, PartitionSpec, NamedSharding\nfrom jax.experimental.mesh_utils import create_device_mesh\nimport optax\nimport orbax.checkpoint as ocp\nimport numpy as np\nimport dm_pix as pix\nimport jax\nimport jax.numpy as jnp\nimport tyro\nimport wandb\nimport grain\n\nfrom genie import Genie, restore_genie_components\nfrom utils.dataloader import get_dataloader\nfrom utils.lr_utils import get_lr_schedule\nfrom utils.parameter_utils import count_parameters_by_component\n\n@dataclass\nclass Args:\n    # Experiment\n    num_steps: int = 200_000\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_height: int = 90\n    image_width: int = 160\n    data_dir: str = """"\n    save_ckpt: bool = False\n    restore_ckpt: bool = False\n    # Optimization\n    batch_size: int = 36\n    init_lr: float = 0.0\n    max_lr: float = 3e-5\n    decay_end: float = 0.0\n    wsd_decay_steps: int = 10000 # NOTE: wsd_decay_steps will only be used when using a wsd-schedule\n    warmup_steps: int = 5000\n    lr_schedule : str = ""wsd"" # supported options: wsd, cos\n    # Tokenizer\n    tokenizer_dim: int = 512\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 4\n    tokenizer_num_blocks: int = 8\n    tokenizer_num_heads: int = 8\n    tokenizer_checkpoint: str = """"\n    # LAM\n    lam_dim: int = 512\n    latent_action_dim: int = 32\n    num_latent_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 8\n    lam_num_heads: int = 8\n    lam_checkpoint: str = """"\n    # Dynamics\n    dyna_dim: int = 512\n    dyna_num_blocks: int = 12\n    dyna_num_heads: int = 8\n    dropout: float = 0.0\n    mask_limit: float = 0.5\n    use_maskgit: bool = False\n    param_dtype: jnp.dtype = jnp.float32\n    dtype: jnp.dtype = jnp.bfloat16\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    name: str = ""train_dynamics""\n    tags: list[str] = field(default_factory=lambda: [""dynamics""])\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 25000\n    log_checkpoint_keep_period: int = 20000\n    log_gradients: bool = False\n    wandb_id: str = """"\n\n\nargs = tyro.cli(Args)\n\n\ndef dynamics_loss_fn(params, state, inputs):\n    """"""Compute masked dynamics loss""""""\n    inputs[""videos""] = inputs[""videos""].astype(args.dtype) / 255.0\n    outputs = state.apply_fn(\n        params,\n        inputs,\n        training=True,\n        rngs={""params"": inputs[""rng""], ""dropout"": inputs[""dropout_rng""]},\n    )\n    mask = outputs[""mask""]\n    outputs[""token_logits""] = outputs[""token_logits""].astype(jnp.float32)\n    logits = outputs[""token_logits""]\n    targets = outputs[""video_tokens""]\n\n    # if not args.use_maskgit:\n        # logits = outputs[""token_logits""][:, :, :-1]\n        # targets = outputs[""video_tokens""][:, :, 1:]\n        # mask = outputs[""mask""][:, :, 1:] \n    ce_loss = optax.softmax_cross_entropy_with_integer_labels(\n        logits, targets\n    )\n    ce_loss = (mask * ce_loss).sum() / mask.sum()\n    acc = logits.argmax(-1) == targets\n    acc = (mask * acc).sum() / mask.sum()\n    select_probs = jax.nn.softmax(logits)\n    gt = inputs[""videos""].clip(0, 1).reshape(-1, *inputs[""videos""].shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = pix.psnr(gt, recon).mean() # type: ignore\n    ssim = pix.ssim(gt, recon).mean() # type: ignore\n    _, index_counts_lam = jnp.unique_counts(\n        jnp.ravel(outputs[""lam_indices""]), size=args.num_latent_actions, fill_value=0\n    )\n    _, index_counts_tokenizer = jnp.unique_counts(\n        jnp.ravel(outputs[""video_tokens""]), size=args.num_patch_latents, fill_value=0\n    )\n    codebook_usage_lam = (index_counts_lam != 0).mean()\n    codebook_usage_tokenizer = (index_counts_tokenizer != 0).mean()\n    metrics = dict(\n        cross_entropy_loss=ce_loss,\n        masked_token_accuracy=acc,\n        select_logit=logits.max(-1).mean(),\n        select_p=select_probs.max(-1).mean(),\n        entropy=jax.scipy.special.entr(select_probs).sum(-1).mean(),\n        psnr=psnr,\n        ssim=ssim,\n        codebook_usage_lam=codebook_usage_lam,\n        codebook_usage_tokenizer=codebook_usage_tokenizer,\n    )\n    return ce_loss, (outputs[""recon""], metrics)\n\n\n@jax.jit\ndef train_step(state, inputs):\n    """"""Update state and compute metrics""""""\n    grad_fn = jax.value_and_grad(dynamics_loss_fn, has_aux=True, allow_int=True)\n    (loss, (recon, metrics)), grads = grad_fn(state.params, state, inputs)\n    state = state.apply_gradients(grads=grads)\n    if args.log_gradients:\n        metrics[""gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""dynamics""]\n        )\n    return state, loss, recon, metrics\n\n\nif __name__ == ""__main__"":\n    jax.distributed.initialize()\n    num_devices = jax.device_count()\n    if num_devices == 0:\n        raise ValueError(""No JAX devices found."")\n    print(f""Running on {num_devices} devices."")\n\n    if args.batch_size % num_devices != 0:\n        raise ValueError(\n            f""Global batch size {args.batch_size} must be divisible by ""\n            f""number of devices {num_devices}.""\n        )\n\n    per_device_batch_size_for_init = args.batch_size // num_devices\n\n    rng = jax.random.PRNGKey(args.seed)\n\n    # --- Initialize model ---\n    genie = Genie(\n        # Tokenizer\n        in_dim=args.image_channels,\n        tokenizer_dim=args.tokenizer_dim,\n        latent_patch_dim=args.latent_patch_dim,\n        num_patch_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        tokenizer_num_blocks=args.tokenizer_num_blocks,\n        tokenizer_num_heads=args.tokenizer_num_heads,\n        # LAM\n        lam_dim=args.lam_dim,\n        latent_action_dim=args.latent_action_dim,\n        num_latent_actions=args.num_latent_actions,\n        lam_patch_size=args.lam_patch_size,\n        lam_num_blocks=args.lam_num_blocks,\n        lam_num_heads=args.lam_num_heads,\n        lam_co_train=not args.lam_checkpoint,\n        # Dynamics\n        dyna_dim=args.dyna_dim,\n        dyna_num_blocks=args.dyna_num_blocks,\n        dyna_num_heads=args.dyna_num_heads,\n        dropout=args.dropout,\n        mask_limit=args.mask_limit,\n        use_maskgit=args.use_maskgit,\n        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n    )\n    rng, _rng = jax.random.split(rng)\n    image_shape = (args.image_height, args.image_width, args.image_channels)\n    dummy_inputs = dict(\n        videos=jnp.zeros(\n            (per_device_batch_size_for_init, args.seq_len, *image_shape),\n            dtype=args.dtype,\n        ),\n        action=jnp.zeros(\n            (per_device_batch_size_for_init, args.seq_len), dtype=args.dtype\n        ),\n        mask_rng=_rng,\n    )\n    rng, _rng = jax.random.split(rng)\n    init_params = genie.init(_rng, dummy_inputs)\n\n    param_counts = count_parameters_by_component(init_params)\n\n    if args.log and jax.process_index() == 0:\n        wandb_init_kwargs = {\n            ""entity"": args.entity,\n            ""project"": args.project,\n            ""name"": args.name,\n            ""tags"": args.tags,\n            ""group"": ""debug"",\n            ""config"": args,\n        }\n\n        if args.wandb_id:\n            wandb_init_kwargs.update(\n                {\n                    ""id"": args.wandb_id,\n                    ""resume"": ""allow"",\n                }\n            )\n        wandb.init(**wandb_init_kwargs)\n\n        wandb.config.update({""model_param_count"": param_counts})\n\n    print(""Parameter counts:"")\n    print(param_counts)\n\n    # --- Initialize optimizer ---\n    lr_schedule = get_lr_schedule(args.lr_schedule, \n                                  args.init_lr, \n                                  args.max_lr, \n                                  args.decay_end, \n                                  args.num_steps, \n                                  args.warmup_steps, \n                                  args.wsd_decay_steps)\n    tx = optax.adamw(learning_rate=lr_schedule, b1=0.9, b2=0.9, weight_decay=1e-4, mu_dtype=args.dtype)\n    train_state = TrainState.create(apply_fn=genie.apply, params=init_params, tx=tx)\n\n    device_mesh_arr = create_device_mesh((num_devices,))\n    mesh = Mesh(devices=device_mesh_arr, axis_names=(""data"",))\n\n    replicated_sharding = NamedSharding(mesh, PartitionSpec())\n    videos_sharding = NamedSharding(mesh, PartitionSpec(""data"", None, None, None, None))\n    train_state = jax.device_put(train_state, replicated_sharding)\n\n    # --- Initialize checkpoint manager ---\n    step = 0\n    handler_registry = ocp.handlers.DefaultCheckpointHandlerRegistry()\n    handler_registry.add(\n        ""model_state"", ocp.args.StandardSave, ocp.handlers.StandardCheckpointHandler\n    )\n    handler_registry.add(\n        ""model_state"", ocp.args.StandardRestore, ocp.handlers.StandardCheckpointHandler\n    )\n    handler_registry.add(""dataloader_state"", grain.checkpoint.CheckpointSave, grain.checkpoint.CheckpointHandler)  # type: ignore\n    handler_registry.add(""dataloader_state"", grain.checkpoint.CheckpointRestore, grain.checkpoint.CheckpointHandler)  # type: ignore\n\n    checkpoint_options = ocp.CheckpointManagerOptions(\n        save_interval_steps=args.log_checkpoint_interval,\n        max_to_keep=3,\n        keep_period=args.log_checkpoint_keep_period,\n        step_format_fixed_length=6,\n        cleanup_tmp_directories=True,\n    )\n\n    checkpoint_manager = ocp.CheckpointManager(\n        args.ckpt_dir,\n        options=checkpoint_options,\n        handler_registry=handler_registry,\n    )\n\n    # --- Create DataLoaderIterator from dataloader ---\n    array_record_files = [\n        os.path.join(args.data_dir, x)\n        for x in os.listdir(args.data_dir)\n        if x.endswith("".array_record"")\n    ]\n    grain_dataloader = get_dataloader(\n        array_record_files,\n        args.seq_len,\n        # NOTE: We deliberately pass the global batch size\n        # The dataloader shards the dataset across all processes\n        args.batch_size,\n        *image_shape,\n        num_workers=8,\n        prefetch_buffer_size=1,\n        seed=args.seed,\n    )\n    initial_state = grain_dataloader._create_initial_state()\n    grain_iterator = grain.DataLoaderIterator(grain_dataloader, initial_state)\n\n    # --- Restore checkpoint ---\n    if args.restore_ckpt:\n        # Restore full dynamics model\n        abstract_train_state = jax.tree_util.tree_map(\n            ocp.utils.to_shape_dtype_struct, train_state\n        )\n        restored = checkpoint_manager.restore(\n            checkpoint_manager.latest_step(),\n            args=ocp.args.Composite(\n                model_state=ocp.args.StandardRestore(abstract_train_state),\n                dataloader_state=grain.checkpoint.CheckpointRestore(grain_iterator),\n            ),\n        )\n        train_state = restored[""model_state""]\n        grain_iterator = restored[""dataloader_state""]\n        step = checkpoint_manager.latest_step() or 0\n        print(f""Restored dataloader and model state from step {step}"")\n    else:\n        # Restore from pre-trained tokenizer (and LAM)\n        train_state = restore_genie_components(\n            train_state, replicated_sharding, grain_iterator, dummy_inputs, rng, args\n        )\n\n    # --- TRAIN LOOP ---\n    dataloader = (jax.make_array_from_process_local_data(videos_sharding, elem) for elem in grain_iterator)  # type: ignore\n    while step < args.num_steps:\n        # for videos in dataloader:\n        videos = np.load(""overfit_dir/corner_8repl.npy"")\n        videos = jax.make_array_from_process_local_data(videos_sharding, videos)\n        while True:\n            # --- Train step ---\n            rng, _rng, _rng_dropout, _rng_mask = jax.random.split(rng, 4)\n\n            inputs = dict(\n                videos=videos,\n                rng=_rng,\n                dropout_rng=_rng_dropout,\n                mask_rng=_rng_mask,\n            )\n            train_state, loss, recon, metrics = train_step(train_state, inputs)\n            metrics[""lr""] = lr_schedule(step)\n            print(f""Step {step}, loss: {loss}"")\n            step += 1\n\n            # --- Logging ---\n            if args.log:\n                if step % args.log_interval == 0 and jax.process_index() == 0:\n                    wandb.log(\n                        {\n                            ""loss"": loss,\n                            ""step"": step,\n                            **metrics,\n                        }\n                    )\n                if step % args.log_image_interval == 0:\n                    gt_seq = inputs[""videos""][0].astype(jnp.float32) #/ 255.0\n                    recon_seq = recon[0].clip(0, 1)\n                    comparison_seq = jnp.concatenate((gt_seq, recon_seq), axis=1)\n                    comparison_seq = einops.rearrange(\n                        comparison_seq * 255, ""t h w c -> h (t w) c""\n                    )\n                    if jax.process_index() == 0:\n                        log_images = dict(\n                            image=wandb.Image(np.asarray(gt_seq[args.seq_len - 1])),\n                            recon=wandb.Image(np.asarray(recon_seq[args.seq_len - 1])),\n                            true_vs_recon=wandb.Image(\n                                np.asarray(comparison_seq.astype(np.uint8))\n                            ),\n                        )\n                        wandb.log(log_images)\n            # --- Checkpointing ---\n            if args.save_ckpt and step % args.log_checkpoint_interval == 0:\n                checkpoint_manager.save(\n                    step,\n                    args=ocp.args.Composite(\n                        model_state=ocp.args.StandardSave(train_state),\n                        dataloader_state=grain.checkpoint.CheckpointSave(\n                            grain_iterator\n                        ),\n                    ),\n                )\n                print(f""Saved checkpoint at step {step}"")\n            if step >= args.num_steps:\n                break\n\n    checkpoint_manager.close()\n",python,tab
+2,590,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"10:10:14 PM [info] Activating crowd-code\n10:10:14 PM [info] Recording started\n10:10:14 PM [info] Initializing git provider using file system watchers...\n10:10:14 PM [info] Git repository found\n10:10:14 PM [info] Git provider initialized successfully\n10:10:14 PM [info] Initial git state: [object Object]\n",Log,tab
+3,3709,"TERMINAL",0,0,"/bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt",,terminal_command
+4,3743,"TERMINAL",0,0,"]633;E;2025-07-17 22:10:18 /bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt;541e3580-6a0f-4f0b-bb4c-198841a406f7]633;C",,terminal_output
+5,3850,"TERMINAL",0,0,"]0;tum_cte0515@hkn1991:/hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash]633;D;0",,terminal_output
+6,21978,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",0,0,"",python,tab
+7,21982,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2750,0,"",python,selection_mouse
+8,25341,"TERMINAL",0,0,"queue",,terminal_command
+9,25354,"TERMINAL",0,0,"]633;E;2025-07-17 22:10:39 queue;8a8d6fe6-a600-49ab-ad7b-e3418768748b]633;C",,terminal_output
+10,25432,"TERMINAL",0,0,"[?1049h(B[?7hEvery 1.0s: squeue --mehkn1991.localdomain: Thu Jul 17 22:10:39 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3353884 accelerat interact tum_cte0  R    7:01:56\t  2 hkn[0508,0517]",,terminal_output
+11,26402,"TERMINAL",0,0,"407",,terminal_output
+12,27450,"TERMINAL",0,0,"18",,terminal_output
+13,28557,"TERMINAL",0,0,"29",,terminal_output
+14,29633,"TERMINAL",0,0,"32:01",,terminal_output
+15,29809,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+16,34077,"TERMINAL",0,0,"scancel 3353884",,terminal_command
+17,34117,"TERMINAL",0,0,"]633;E;2025-07-17 22:10:48 scancel 3353884;8a8d6fe6-a600-49ab-ad7b-e3418768748b]633;C]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+18,100445,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2966,0,"",python,selection_mouse
+19,102391,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2912,0,"",python,selection_command
+20,102556,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2858,0,"",python,selection_command
+21,102702,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2827,0,"",python,selection_command
+22,102897,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2858,0,"",python,selection_command
+23,103038,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2912,0,"",python,selection_command
+24,103183,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2966,0,"",python,selection_command
+25,103313,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2912,0,"",python,selection_command
+26,103461,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2858,0,"",python,selection_command
+27,103646,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2827,0,"",python,selection_command
+28,103749,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2858,0,"",python,selection_command
+29,103907,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2912,0,"",python,selection_command
+30,104034,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2966,0,"",python,selection_command
+31,104150,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",3010,0,"",python,selection_command
+32,104180,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2966,0,"",python,selection_command
+33,104347,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2912,0,"",python,selection_command
+34,104508,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2858,0,"",python,selection_command
+35,104634,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2827,0,"",python,selection_command
+36,104768,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2800,0,"",python,selection_command
+37,105312,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2801,0,"",python,selection_command
+38,105809,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2802,0,"",python,selection_command
+39,106322,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2803,0,"",python,selection_command
+40,106363,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2804,0,"",python,selection_command
+41,106416,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2805,0,"",python,selection_command
+42,106454,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2806,0,"",python,selection_command
+43,106498,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2807,0,"",python,selection_command
+44,106499,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2808,0,"",python,selection_command
+45,106500,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2809,0,"",python,selection_command
+46,106539,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2810,0,"",python,selection_command
+47,106557,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2811,0,"",python,selection_command
+48,106584,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2812,0,"",python,selection_command
+49,106634,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2843,0,"",python,selection_command
+50,107209,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2897,0,"",python,selection_command
+51,107343,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jafar/train_dynamics.py",2951,0,"",python,selection_command
+52,109636,"slurm/jobs/mihir/horeka/yolo-runs/sampling.sh",0,0,"\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\n# source .venv/bin/activate\n\narray_records_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_new/open_ai_minecraft_arrayrecords_chunked\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\nCHECKPOINT_DIR=$ws_dir/checkpoints/$job_name/$slurm_job_id\nmkdir -p $CHECKPOINT_DIR\n\ntokenizer_ckpt_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/big-runs/tokenizer-lr-scaling/train_tokenizer_lr_sweep_1e-4\ndynamics_ckpt_dir=$1\necho $dynamics_ckpt_dir\n\nenv | grep SLURM\n\npython sample.py \\n    --checkpoint $dynamics_ckpt_dir \\n    --dyna_dim=128 \\n    --dyna_num_blocks=2 \\n    --dyna_num_heads=4 \\n    --seq_len=2 \\n    --data_dir $array_records_dir\n\n",shellscript,tab
+53,114762,"TERMINAL",0,0,"",,terminal_focus
+54,121002,"TERMINAL",0,0,"salloc --time=10:00:00 --partition=accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5",,terminal_command
+55,121089,"TERMINAL",0,0,"]633;E;2025-07-17 22:12:15 salloc --time=10:00:00 --partition=accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5;2fac4da8-d4f0-4d83-a6ce-f6776ed5ed51]633;Csalloc: Granted job allocation 3355871\r\n",,terminal_output
+56,121223,"TERMINAL",0,0,"salloc: Waiting for resource configuration\r\n",,terminal_output
+57,148351,"TERMINAL",0,0,"salloc: Nodes hkn0508 are ready for job\r\n",,terminal_output
+58,149414,"TERMINAL",0,0,"]0;tum_cte0515@hkn0508:~/Projects/jafar[?2004h[tum_cte0515@hkn0508 jafar]$ ",,terminal_output
+59,165129,"TERMINAL",0,0,"\r[tum_cte0515@hkn0508 jafar]$ ",,terminal_output

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-25569aaa-6e77-4ce2-b9b6-8ae8c33420051753180192494-2025_07_22-12.30.11.399/source.csv

@@ -0,0 +1,16 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,5,"slurm/jobs/mihir/horeka/yolo-runs/sampling.sh",0,0,"\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\n# source .venv/bin/activate\n\narray_records_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_new/open_ai_minecraft_arrayrecords_chunked\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\nCHECKPOINT_DIR=$ws_dir/checkpoints/$job_name/$slurm_job_id\nmkdir -p $CHECKPOINT_DIR\n\ntokenizer_ckpt_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/big-runs/tokenizer-lr-scaling/train_tokenizer_lr_sweep_1e-4\ndynamics_ckpt_dir=$1\necho $dynamics_ckpt_dir\n\nenv | grep SLURM\n\nsrun python sample.py \\n    --checkpoint $dynamics_ckpt_dir \\n    --dyna_dim=128 \\n    --dyna_num_blocks=2 \\n    --dyna_num_heads=4 \\n    --seq_len=16 \\n    --start_frame=10 \\n    --data_dir $array_records_dir\n\n# srun python sample.py \\n    # --checkpoint $dynamics_ckpt_dir \\n    # --start_frame=0 \\n    # --batch_size=12 \\n    # --seq_len=2 \\n    # --data_dir $array_records_dir\n",shellscript,tab
+2,446,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"12:30:11 PM [info] Activating crowd-code\n12:30:11 PM [info] Recording started\n12:30:11 PM [info] Initializing git provider using file system watchers...\n12:30:11 PM [info] Git repository found\n12:30:11 PM [info] Git provider initialized successfully\n12:30:11 PM [info] Initial git state: [object Object]\n",Log,tab
+3,1898,"slurm/jobs/mihir/horeka/yolo-runs/sampling.sh",0,0,"",shellscript,tab
+4,8030,"TERMINAL",0,0,"cd checkpoints/big-runs/",,terminal_command
+5,8072,"TERMINAL",0,0,"]633;E;2025-07-22 12:30:19 cd checkpoints/big-runs/;5fcdc89e-3b1c-4d05-a6d2-6f1ce0ba6ffb]633;C]0;tum_cte0515@hkn1993:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/big-runs]633;D;0",,terminal_output
+6,10557,"TERMINAL",0,0,"cd tokenizer-lr-scaling/",,terminal_command
+7,12047,"TERMINAL",0,0,"ls",,terminal_command
+8,12130,"TERMINAL",0,0,"]633;E;2025-07-22 12:30:23 ls;5fcdc89e-3b1c-4d05-a6d2-6f1ce0ba6ffb]633;Ctrain_tokenizer_lr_sweep_1e-4         train_tokenizer_lr_sweep_5e-5\r\ntrain_tokenizer_lr_sweep_1e-4_8nodes  train_tokenizer_lr_sweep_5e-5_8nodes\r\n]0;tum_cte0515@hkn1993:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/big-runs/tokenizer-lr-scaling]633;D;0",,terminal_output
+9,16772,"TERMINAL",0,0,"cd train_tokenizer_lr_sweep_1e-4",,terminal_command
+10,16797,"TERMINAL",0,0,"]633;E;2025-07-22 12:30:28 cd train_tokenizer_lr_sweep_1e-4;5fcdc89e-3b1c-4d05-a6d2-6f1ce0ba6ffb]633;C]0;tum_cte0515@hkn1993:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/big-runs/tokenizer-lr-scaling/train_tokenizer_lr_sweep_1e-4]633;D;0",,terminal_output
+11,17848,"TERMINAL",0,0,"ls",,terminal_command
+12,17849,"TERMINAL",0,0,"]633;E;2025-07-22 12:30:28 ls;5fcdc89e-3b1c-4d05-a6d2-6f1ce0ba6ffb]633;C",,terminal_output
+13,17862,"TERMINAL",0,0,"020000  060000  100000  140000  145000  146000.zip\r\n040000  080000  120000  144000  146000\r\n]0;tum_cte0515@hkn1993:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/big-runs/tokenizer-lr-scaling/train_tokenizer_lr_sweep_1e-4]633;D;0",,terminal_output
+14,21855,"TERMINAL",0,0,"cursor .",,terminal_command
+15,21897,"TERMINAL",0,0,"]633;E;2025-07-22 12:30:33 cursor .;5fcdc89e-3b1c-4d05-a6d2-6f1ce0ba6ffb]633;C",,terminal_output

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-640d2ea2-6d4b-4f60-ac22-96274589d9ad1759267592825-2025_09_30-23.27.51.17/source.csv

@@ -0,0 +1,227 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,8,"slurm/jobs/franz/berlin/coinrun/mila_submission/ablations/coinrun_dynamics_no_flash_attn.sh",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=24:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --output=/fast/project/HFMI_SynergyUnit/jafar_ws/logs/franz/coinrun/dynamics/%x_%j.log\n#SBATCH --error=/fast/project/HFMI_SynergyUnit/jafar_ws/logs/franz/coinrun/dynamics/%x_%j.log\n#SBATCH --job-name=dynamics_coinrun_mila_submission_no_flash_attn\n#SBATCH --requeue\n#SBATCH --signal=b:usr1@300 # 5 min before timeout\n\n# --- signal trap to requeue job before timeout ---\nrequeue_job() {\n    echo ""[$(date)] caught sigusr1 (timeout warning), requeueing slurm job $SLURM_JOB_ID...""\n    # optional: trigger checkpoint saving here\n    # e.g., touch $checkpoint_dir/requeue_trigger\n    scontrol requeue $SLURM_JOB_ID\n    exit 0\n}\n\ntrap requeue_job sigusr1\n\n# set checkpoint flag based on restart count\nrestart_count=$(scontrol show job $SLURM_JOB_ID | grep -o 'Restarts=[0-9]*' | cut -d'=' -f2)\n\nif [ $restart_count -eq 0 ]; then\n    restore_ckpt_flag=""--no-restore-ckpt""\nelse\n    restore_ckpt_flag=""--restore-ckpt""\nfi\n\n\n\n# Log the sbatch script\ncat $0\n\nsource .venv/bin/activate\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\ntags=""coinrun dynamics 500m_dataset mila_submission ablation no-flash-attn""\n\narray_records_dir=""/fast/project/HFMI_SynergyUnit/jafar_ws/data/coinrun/array_records_500m_seed_w_increment""\ntokenizer_ckpt_dir=""/fast/project/HFMI_SynergyUnit/jafar_ws/checkpoints/coinrun/tokenizer/tokenizer_coinrun_mila_submission_29736/""\nCHECKPOINT_DIR=""/fast/project/HFMI_SynergyUnit/jafar_ws/checkpoints/coinrun/dynamics/${job_name}/${slurm_job_id}""\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\n\nsrun python jasmine/train_dynamics.py \\n    --no-use-flash-attention \\n    --save_ckpt \\n    $restore_ckpt_flag \\n    --wandb_id $SLURM_JOB_ID \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --name=""${job_name}_${slurm_job_id}"" \\n    --tags ${tags} \\n    --entity instant-uv \\n    --project jafar \\n    --tokenizer_checkpoint=""${tokenizer_ckpt_dir}"" \\n    --val_data_dir=""${array_records_dir}/val"" \\n    --data_dir=""${array_records_dir}/train"" &\n\nchild_pid=$!\n\nwait $child_pid\n\n",shellscript,tab
+2,5021,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"11:27:50 PM [info] Activating crowd-code\n11:27:51 PM [info] Recording started\n11:27:51 PM [info] Initializing git provider using file system watchers...\n11:27:53 PM [info] Retrying git provider initialization...\n",Log,tab
+3,5680,"extension-output-pdoom-org.crowd-code-#1-crowd-code",212,0,"11:27:54 PM [info] Git repository found\n11:27:54 PM [info] Git provider initialized successfully\n11:27:54 PM [info] Initial git state: [object Object]\n11:27:56 PM [info] Git repository found\n11:27:56 PM [info] Git provider initialized successfully\n11:27:56 PM [info] Initial git state: [object Object]\n",Log,content
+4,397553,"slurm/jobs/franz/berlin/coinrun/mila_submission/ablations/coinrun_dynamics_no_flash_attn.sh",0,0,"",shellscript,tab
+5,416841,"jasmine/train_dynamics.py",0,0,"import os\n\n\nos.environ.setdefault(""XLA_PYTHON_CLIENT_MEM_FRACTION"", ""0.98"")\n\nfrom dataclasses import dataclass, field\nimport itertools\nfrom typing import cast, Optional\n\nimport einops\nfrom jax.sharding import Mesh, PartitionSpec, NamedSharding\nfrom jax.experimental.mesh_utils import create_device_mesh\nimport optax\nimport orbax.checkpoint as ocp\nimport numpy as np\nimport dm_pix as pix\nimport jax\nimport jax.numpy as jnp\nimport tyro\nimport wandb\nimport grain\nimport flax.nnx as nnx\n\nfrom genie import Genie, restore_genie_components\nfrom utils.dataloader import get_dataloader\nfrom utils.train_utils import (\n    get_lr_schedule,\n    count_parameters_by_component,\n    print_mem_stats,\n    print_compiled_memory_stats,\n    print_compiled_cost_analysis,\n)\n\n\n@dataclass\nclass Args:\n    # Experiment\n    num_steps: int = 200_000\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_height: int = 64\n    image_width: int = 64\n    data_dir: str = """"\n    save_ckpt: bool = False\n    restore_ckpt: bool = False\n    # Optimization\n    batch_size: int = 36\n    init_lr: float = 0.0\n    max_lr: float = 3e-5\n    decay_end: float = 0.0\n    wsd_decay_steps: int = (\n        20_000  # NOTE: wsd_decay_steps will only be used when using a wsd-schedule\n    )\n    warmup_steps: int = 5000\n    lr_schedule: str = ""wsd""  # supported options: wsd, cos\n    # Tokenizer\n    tokenizer_dim: int = 512\n    tokenizer_ffn_dim: int = 2048\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 16\n    tokenizer_num_blocks: int = 4\n    tokenizer_num_heads: int = 8\n    tokenizer_checkpoint: str = """"\n    # LAM\n    lam_dim: int = 512\n    lam_ffn_dim: int = 2048\n    latent_action_dim: int = 32\n    num_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 4\n    lam_num_heads: int = 8\n    lam_checkpoint: str = """"\n    # Dynamics\n    dyna_type: str = ""maskgit""  # supported options: maskgit, causal\n    dyna_dim: int = 512\n    dyna_ffn_dim: int = 2048\n    dyna_num_blocks: int = 6\n    dyna_num_heads: int = 8\n    dropout: float = 0.0\n    mask_limit: float = 0.5\n    z_loss_weight: float = 0.0\n    param_dtype = jnp.float32\n    dtype = jnp.bfloat16\n    use_flash_attention: bool = True\n    use_gt_actions: bool = False\n    # Logging\n    log: bool = True\n    entity: str = """"\n    project: str = """"\n    name: str = ""train_dynamics""\n    tags: list[str] = field(default_factory=lambda: [""dynamics""])\n    log_interval: int = 50\n    log_image_interval: int = 1000\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 5000\n    log_checkpoint_keep_period: int = 20_000\n    log_gradients: bool = False\n    val_data_dir: str = """"\n    val_interval: int = 20_000\n    val_steps: int = 50\n    eval_full_frame: bool = True\n    val_maskgit_steps: int = 25\n    val_temperature: float = 1\n    val_sample_argmax: bool = False\n    wandb_id: str = """"\n\n\ndef build_model(args: Args, rng: jax.Array) -> tuple[Genie, jax.Array]:\n    rng, _rng = jax.random.split(rng)\n    rngs = nnx.Rngs(_rng)\n    genie = Genie(\n        # Tokenizer\n        in_dim=args.image_channels,\n        tokenizer_dim=args.tokenizer_dim,\n        tokenizer_ffn_dim=args.tokenizer_ffn_dim,\n        latent_patch_dim=args.latent_patch_dim,\n        num_patch_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        tokenizer_num_blocks=args.tokenizer_num_blocks,\n        tokenizer_num_heads=args.tokenizer_num_heads,\n        # LAM\n        lam_dim=args.lam_dim,\n        lam_ffn_dim=args.lam_ffn_dim,\n        latent_action_dim=args.latent_action_dim,\n        num_actions=args.num_actions,\n        lam_patch_size=args.lam_patch_size,\n        lam_num_blocks=args.lam_num_blocks,\n        lam_num_heads=args.lam_num_heads,\n        lam_co_train=not args.lam_checkpoint,\n        use_gt_actions=args.use_gt_actions,\n        # Dynamics\n        dyna_type=args.dyna_type,\n        dyna_dim=args.dyna_dim,\n        dyna_ffn_dim=args.dyna_ffn_dim,\n        dyna_num_blocks=args.dyna_num_blocks,\n        dyna_num_heads=args.dyna_num_heads,\n        dropout=args.dropout,\n        mask_limit=args.mask_limit,\n        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n        use_flash_attention=args.use_flash_attention,\n        decode=False,\n        rngs=rngs,\n    )\n    if args.use_gt_actions:\n        assert (\n            not args.lam_checkpoint\n        ), ""Cannot use LAM when using ground-truth actions.""\n    else:\n        assert genie.lam is not None\n        del genie.lam.decoder\n    return genie, rng\n\n\ndef build_optimizer(genie: Genie, args: Args) -> nnx.ModelAndOptimizer:\n    lr_schedule = get_lr_schedule(\n        args.lr_schedule,\n        args.init_lr,\n        args.max_lr,\n        args.decay_end,\n        args.num_steps,\n        args.warmup_steps,\n        args.wsd_decay_steps,\n    )\n    tx = optax.adamw(\n        learning_rate=lr_schedule,\n        b1=0.9,\n        b2=0.9,\n        weight_decay=1e-4,\n        mu_dtype=args.param_dtype,  # moments in full precision\n    )\n    optimizer = nnx.ModelAndOptimizer(genie, tx)\n    return optimizer\n\n\ndef build_mesh_and_sharding(\n    num_devices: int,\n) -> tuple[Mesh, NamedSharding, NamedSharding, NamedSharding]:\n    device_mesh_arr = create_device_mesh((num_devices,))\n    mesh = Mesh(devices=device_mesh_arr, axis_names=(""data"",))\n    replicated_sharding = NamedSharding(mesh, PartitionSpec())\n    videos_sharding = NamedSharding(mesh, PartitionSpec(""data"", None, None, None, None))\n    actions_sharding = NamedSharding(mesh, PartitionSpec(""data"", None))\n    return mesh, replicated_sharding, videos_sharding, actions_sharding\n\n\ndef shard_optimizer_states(\n    optimizer: nnx.ModelAndOptimizer, replicated_sharding: NamedSharding\n) -> None:\n    model_state = nnx.state(optimizer.model)\n    model_sharded_state = jax.lax.with_sharding_constraint(\n        model_state, replicated_sharding\n    )\n    nnx.update(optimizer.model, model_sharded_state)\n    optimizer_state = nnx.state(optimizer, nnx.optimizer.OptState)\n    optimizer_sharded_state = jax.lax.with_sharding_constraint(\n        optimizer_state, replicated_sharding\n    )\n    nnx.update(optimizer, optimizer_sharded_state)\n\n\ndef build_dataloader(args: Args, data_dir: str) -> grain.DataLoaderIterator:\n    image_shape = (args.image_height, args.image_width, args.image_channels)\n    array_record_files = [\n        os.path.join(data_dir, x)\n        for x in os.listdir(data_dir)\n        if x.endswith("".array_record"")\n    ]\n    grain_dataloader = get_dataloader(\n        array_record_files,\n        args.seq_len,\n        # NOTE: We deliberately pass the global batch size\n        # The dataloader shards the dataset across all processes\n        args.batch_size,\n        *image_shape,\n        num_workers=8,\n        prefetch_buffer_size=1,\n        seed=args.seed,\n    )\n    initial_state = grain_dataloader._create_initial_state()\n    grain_iterator = grain.DataLoaderIterator(grain_dataloader, initial_state)\n    return grain_iterator\n\n\ndef build_checkpoint_manager(args: Args) -> Optional[ocp.CheckpointManager]:\n    if args.restore_ckpt or args.save_ckpt:\n        handler_registry = ocp.handlers.DefaultCheckpointHandlerRegistry()\n        handler_registry.add(\n            ""model_state"", ocp.args.PyTreeSave, ocp.handlers.PyTreeCheckpointHandler\n        )\n        handler_registry.add(\n            ""model_state"", ocp.args.PyTreeRestore, ocp.handlers.PyTreeCheckpointHandler\n        )\n        handler_registry.add(\n            ""train_dataloader_state"",\n            grain.checkpoint.CheckpointSave,\n            cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n        )\n        handler_registry.add(\n            ""train_dataloader_state"",\n            grain.checkpoint.CheckpointRestore,\n            cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n        )\n        if args.val_data_dir:\n            handler_registry.add(\n                ""val_dataloader_state"",\n                grain.checkpoint.CheckpointSave,\n                cast(\n                    ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler\n                ),\n            )\n            handler_registry.add(\n                ""val_dataloader_state"",\n                grain.checkpoint.CheckpointRestore,\n                cast(\n                    ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler\n                ),\n            )\n        checkpoint_options = ocp.CheckpointManagerOptions(\n            save_interval_steps=args.log_checkpoint_interval,\n            max_to_keep=3,\n            keep_period=args.log_checkpoint_keep_period,\n            step_format_fixed_length=6,\n            cleanup_tmp_directories=True,\n        )\n        checkpoint_manager = ocp.CheckpointManager(\n            args.ckpt_dir,\n            options=checkpoint_options,\n            handler_registry=handler_registry,\n        )\n        return checkpoint_manager\n    else:\n        return None\n\n\ndef restore_or_initialize_components(\n    args: Args,\n    checkpoint_manager: Optional[ocp.CheckpointManager],\n    optimizer: nnx.ModelAndOptimizer,\n    train_iterator: grain.DataLoaderIterator,\n    rng: jax.Array,\n    replicated_sharding: NamedSharding,\n    val_iterator: Optional[grain.DataLoaderIterator],\n    restore_step: Optional[int] = None,\n) -> tuple[\n    int,\n    nnx.ModelAndOptimizer,\n    grain.DataLoaderIterator,\n    grain.DataLoaderIterator,\n    jax.Array,\n]:\n    step = 0\n    if checkpoint_manager and restore_step is None:\n        restore_step = checkpoint_manager.latest_step()\n    if args.restore_ckpt:\n        assert checkpoint_manager is not None\n        abstract_optimizer = nnx.eval_shape(lambda: optimizer)\n        abstract_optimizer_state = nnx.state(abstract_optimizer)\n        if val_iterator:\n            restore_args = ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                train_dataloader_state=grain.checkpoint.CheckpointRestore(train_iterator),  # type: ignore\n                val_dataloader_state=grain.checkpoint.CheckpointRestore(val_iterator),  # type: ignore\n            )\n        else:\n            restore_args = ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                train_dataloader_state=grain.checkpoint.CheckpointRestore(train_iterator),  # type: ignore\n            )\n        restored = checkpoint_manager.restore(\n            checkpoint_manager.latest_step(), args=restore_args\n        )\n        restored_optimizer_state = restored[""model_state""]\n        nnx.update(optimizer, restored_optimizer_state)\n        train_iterator = restored[""train_dataloader_state""]\n        if val_iterator:\n            val_iterator = restored[""val_dataloader_state""]\n        step = checkpoint_manager.latest_step() or 0\n        print(f""Restored dataloader and model state from step {step}"")\n    else:\n        # Restore from pre-trained tokenizer (and LAM)\n        rng, _rng = jax.random.split(rng)\n        optimizer = restore_genie_components(optimizer, replicated_sharding, _rng, args)\n    return step, optimizer, train_iterator, val_iterator, rng\n\n\ndef _calculate_top_k_accuracy(\n    token_logits_BTNV: jax.Array,\n    video_tokens_BTN: jax.Array,\n    mask_BTN: jax.Array,\n    k: int,\n) -> jax.Array:\n    _, topk_indices_BTNK = jax.lax.top_k(token_logits_BTNV, k)\n    topk_correct = jnp.any(\n        topk_indices_BTNK == video_tokens_BTN[..., jnp.newaxis], axis=-1\n    )\n    topk_acc = (mask_BTN * topk_correct).sum() / mask_BTN.sum()\n    return topk_acc\n\n\ndef _calculate_step_metrics(\n    outputs: dict[str, jax.Array],\n    gt: jax.Array,\n    num_actions: int,\n    num_patch_latents: int,\n) -> tuple[jax.Array, dict]:\n    mask_BTN = outputs[""mask""]\n    outputs[""token_logits""] = outputs[""token_logits""].astype(jnp.float32)\n    ce_loss = optax.softmax_cross_entropy_with_integer_labels(\n        outputs[""token_logits""], outputs[""video_tokens""]\n    )\n    ce_loss = (mask_BTN * ce_loss).sum() / mask_BTN.sum()\n    z_val = jax.nn.logsumexp(outputs[""token_logits""], axis=-1)\n    z_loss_metric = (mask_BTN * (z_val**2)).sum() / mask_BTN.sum()\n\n    masked_token_top_1_acc = _calculate_top_k_accuracy(\n        outputs[""token_logits""], outputs[""video_tokens""], mask_BTN, 1\n    )\n    masked_token_top_2_acc = _calculate_top_k_accuracy(\n        outputs[""token_logits""], outputs[""video_tokens""], mask_BTN, 2\n    )\n    masked_token_top_5_acc = _calculate_top_k_accuracy(\n        outputs[""token_logits""], outputs[""video_tokens""], mask_BTN, 5\n    )\n    masked_token_top_16_acc = _calculate_top_k_accuracy(\n        outputs[""token_logits""], outputs[""video_tokens""], mask_BTN, 16\n    )\n\n    select_probs = jax.nn.softmax(outputs[""token_logits""])\n    gt_val = gt.clip(0, 1).reshape(-1, *gt.shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = jnp.asarray(pix.psnr(gt_val, recon)).mean()\n    ssim = jnp.asarray(pix.ssim(gt_val, recon)).mean()\n    _, index_counts_tokenizer = jnp.unique_counts(\n        jnp.ravel(outputs[""video_tokens""]),\n        size=num_patch_latents,\n        fill_value=0,\n    )\n    codebook_usage_tokenizer = (index_counts_tokenizer != 0).mean()\n    metrics = dict(\n        cross_entropy_loss=ce_loss,\n        masked_token_top1_accuracy=masked_token_top_1_acc,\n        masked_token_top2_accuracy=masked_token_top_2_acc,\n        masked_token_top5_accuracy=masked_token_top_5_acc,\n        masked_token_top16_accuracy=masked_token_top_16_acc,\n        select_logit=outputs[""token_logits""].max(-1).mean(),\n        select_p=select_probs.max(-1).mean(),\n        entropy=jax.scipy.special.entr(select_probs).sum(-1).mean(),\n        z_loss=z_loss_metric,\n        psnr=psnr,\n        ssim=ssim,\n        codebook_usage_tokenizer=codebook_usage_tokenizer,\n    )\n    if ""lam_indices"" in outputs.keys():\n        _, index_counts_lam = jnp.unique_counts(\n            jnp.ravel(outputs[""lam_indices""]),\n            size=num_actions,\n            fill_value=0,\n        )\n        codebook_usage_lam = (index_counts_lam != 0).mean()\n        metrics[""codebook_usage_lam""] = codebook_usage_lam\n    return ce_loss, metrics\n\n\ndef main(args: Args) -> None:\n    jax.distributed.initialize()\n    num_devices = jax.device_count()\n    if num_devices == 0:\n        raise ValueError(""No JAX devices found."")\n    print(f""Running on {num_devices} devices."")\n\n    if args.batch_size % num_devices != 0:\n        raise ValueError(\n            f""Global batch size {args.batch_size} must be divisible by ""\n            f""number of devices {num_devices}.""\n        )\n\n    rng = jax.random.key(args.seed)\n\n    # --- Initialize model ---\n    genie, rng = build_model(args, rng)\n    _, params, _ = nnx.split(genie, nnx.Param, ...)\n    param_counts = count_parameters_by_component(params)\n\n    if args.log and jax.process_index() == 0:\n        wandb_init_kwargs = {\n            ""entity"": args.entity,\n            ""project"": args.project,\n            ""name"": args.name,\n            ""tags"": args.tags,\n            ""group"": ""debug"",\n            ""config"": args,\n        }\n\n        if args.wandb_id:\n            wandb_init_kwargs.update(\n                {\n                    ""id"": args.wandb_id,\n                    ""resume"": ""allow"",\n                }\n            )\n        wandb.init(**wandb_init_kwargs)\n\n        wandb.config.update({""model_param_count"": param_counts})\n\n    print(""Parameter counts:"")\n    print(param_counts)\n\n    # --- Initialize optimizer ---\n    optimizer = build_optimizer(genie, args)\n    del genie\n\n    # FIXME: switch to create_hybrid_device_mesh for runs spanning multiple nodes\n    _, replicated_sharding, videos_sharding, actions_sharding = build_mesh_and_sharding(\n        num_devices\n    )\n\n    shard_optimizer_states(optimizer, replicated_sharding)\n\n    # --- Initialize checkpoint manager ---\n    checkpoint_manager = build_checkpoint_manager(args)\n\n    # --- Create DataLoaderIterator from dataloader ---\n    train_iterator = build_dataloader(args, args.data_dir)\n    val_iterator = None\n    if args.val_data_dir:\n        val_iterator = build_dataloader(args, args.val_data_dir)\n\n    # --- Restore checkpoint ---\n    step, optimizer, train_iterator, val_iterator, rng = (\n        restore_or_initialize_components(\n            args,\n            checkpoint_manager,\n            optimizer,\n            train_iterator,\n            rng,\n            replicated_sharding,\n            val_iterator,\n        )\n    )\n\n    # --- Define loss and train step (close over args) ---\n    def dynamics_loss_fn(\n        model: Genie,\n        inputs: dict,\n    ) -> tuple[jax.Array, tuple[jax.Array, dict]]:\n        gt = jnp.asarray(inputs[""videos""], dtype=jnp.float32) / 255.0\n        inputs[""videos""] = gt.astype(args.dtype)\n        outputs = model(inputs)\n        ce_loss, metrics = _calculate_step_metrics(\n            outputs, gt, args.num_actions, args.num_patch_latents\n        )\n        z_loss = metrics[""z_loss""]\n        total_loss = ce_loss + args.z_loss_weight * z_loss\n        metrics[""total_loss""] = total_loss\n        return total_loss, (outputs[""recon""], metrics)\n\n    @nnx.jit(donate_argnums=0)\n    def train_step(\n        optimizer: nnx.ModelAndOptimizer, inputs: dict\n    ) -> tuple[jax.Array, jax.Array, dict]:\n        def loss_fn(model: Genie) -> tuple[jax.Array, tuple[jax.Array, dict]]:\n            model.train()\n            return dynamics_loss_fn(model, inputs)\n\n        (loss, (recon, metrics)), grads = nnx.value_and_grad(loss_fn, has_aux=True)(\n            optimizer.model\n        )\n        optimizer.update(grads)\n        if args.log_gradients:\n            metrics[""gradients_std/""] = jax.tree.map(\n                lambda x: x.std(), grads[""params""][""dynamics""]\n            )\n        return loss, recon, metrics\n\n    @nnx.jit\n    def val_step(genie: Genie, inputs: dict) -> dict:\n        """"""Evaluate model and compute metrics""""""\n        genie.eval()\n        gt = jnp.asarray(inputs[""videos""], dtype=jnp.float32) / 255.0\n        (loss, (recon, metrics)) = dynamics_loss_fn(genie, inputs)\n        val_output = {""loss"": loss, ""recon"": recon, ""metrics"": metrics}\n\n        # --- Evaluate full frame prediction (sampling) ---\n        if args.eval_full_frame:\n            inputs[""videos""] = gt.astype(args.dtype)\n            tokenizer_outputs = genie.tokenizer.vq_encode(\n                inputs[""videos""], training=False\n            )\n            tokens_full_frame = tokenizer_outputs[""indices""]\n            lam_indices_E = None\n            if not args.use_gt_actions:\n                lam_indices_E = genie.vq_encode(inputs, training=False)\n                inputs[""latent_actions""] = lam_indices_E\n            inputs[""videos""] = inputs[""videos""][\n                :, :-1\n            ]  # remove last frame for generation\n            recon_full_frame, logits_full_frame = genie.sample(\n                inputs,\n                args.seq_len,\n                args.val_temperature,\n                args.val_sample_argmax,\n                args.val_maskgit_steps,\n            )\n            # Calculate metrics for the last frame only\n            step_outputs = {\n                ""recon"": recon_full_frame[:, -1],\n                ""token_logits"": logits_full_frame[:, -1],\n                ""video_tokens"": tokens_full_frame[:, -1],\n                ""mask"": jnp.ones_like(tokens_full_frame[:, -1]),\n            }\n            if lam_indices_E is not None:\n                lam_indices_B = lam_indices_E.reshape((-1, args.seq_len - 1))[:, -1]\n                step_outputs[""lam_indices""] = lam_indices_B\n\n            loss_full_frame, metrics_full_frame = _calculate_step_metrics(\n                step_outputs, gt[:, -1], args.num_actions, args.num_patch_latents\n            )\n            val_output.update(\n                {\n                    ""loss_full_frame"": loss_full_frame,\n                    ""recon_full_frame"": recon_full_frame,\n                    ""metrics_full_frame"": metrics_full_frame,\n                }\n            )\n        return val_output\n\n    def calculate_validation_metrics(val_dataloader, genie, rng):\n        step = 0\n        loss_per_step = []\n        metrics_per_step = []\n        loss_full_frame_per_step = []\n        metrics_full_frame_per_step = []\n        batch = None\n        recon = None\n        recon_full_frame = None\n        for batch in val_dataloader:\n            rng, _rng_mask = jax.random.split(rng, 2)\n            batch[""rng""] = _rng_mask\n            val_outputs = val_step(genie, batch)\n            loss_per_step.append(val_outputs[""loss""])\n            metrics_per_step.append(val_outputs[""metrics""])\n            recon = val_outputs[""recon""]\n            if args.eval_full_frame:\n                loss_full_frame_per_step.append(val_outputs[""loss_full_frame""])\n                metrics_full_frame_per_step.append(val_outputs[""metrics_full_frame""])\n                recon_full_frame = val_outputs[""recon_full_frame""]\n            step += 1\n            if step > args.val_steps:\n                break\n\n        if step < args.val_steps:\n            print(\n                f""Warning: Your validation dataset is too small to make val_steps many steps. Made {step} steps, expected {args.val_steps}""\n            )\n\n        val_metrics = {\n            f""val_{key}"": np.mean([float(m[key]) for m in metrics_per_step])\n            for key in metrics_per_step[0].keys()\n        }\n        val_metrics[""val_loss""] = np.mean(loss_per_step)\n        if args.eval_full_frame:\n            val_metrics_full_frame = {\n                f""val_full_frame_{key}"": np.mean(\n                    [float(m[key]) for m in metrics_full_frame_per_step]\n                )\n                for key in metrics_full_frame_per_step[0].keys()\n            }\n            val_metrics.update(val_metrics_full_frame)\n            val_metrics[""val_full_frame_loss""] = np.mean(loss_full_frame_per_step)\n        return val_metrics, batch, recon, recon_full_frame\n\n    # --- TRAIN LOOP ---\n    dataloader_train = (\n        {\n            ""videos"": jax.make_array_from_process_local_data(\n                videos_sharding, local_data=elem[""videos""]\n            ),\n            ""actions"": (\n                jax.make_array_from_process_local_data(\n                    actions_sharding, elem[""actions""]\n                )\n                if args.use_gt_actions\n                else None\n            ),\n        }\n        for elem in train_iterator\n    )\n    dataloader_val = None\n    if val_iterator:\n        dataloader_val = (\n            {\n                ""videos"": jax.make_array_from_process_local_data(\n                    videos_sharding, elem[""videos""]\n                ),\n                ""actions"": (\n                    jax.make_array_from_process_local_data(\n                        actions_sharding, elem[""actions""]\n                    )\n                    if args.use_gt_actions\n                    else None\n                ),\n            }\n            for elem in val_iterator\n        )\n    if jax.process_index() == 0:\n        first_batch = next(dataloader_train)\n        first_batch[""rng""] = rng  # type: ignore\n        compiled = train_step.lower(optimizer, first_batch).compile()\n        print_compiled_memory_stats(compiled.memory_analysis())\n        print_compiled_cost_analysis(compiled.cost_analysis())\n        # Do not skip the first batch during training\n        dataloader_train = itertools.chain([first_batch], dataloader_train)\n    print(f""Starting training from step {step}..."")\n    first_step = step\n    while step < args.num_steps:\n        for batch in dataloader_train:\n            # --- Train step ---\n            rng, _rng_mask = jax.random.split(rng, 2)\n            batch[""rng""] = _rng_mask\n            loss, recon, metrics = train_step(optimizer, batch)\n            if step == first_step:\n                print_mem_stats(""After params initialized"")\n            step += 1\n\n            # --- Validation loss ---\n            val_results = {}\n            if dataloader_val and step % args.val_interval == 0:\n                rng, _rng_mask_val = jax.random.split(rng, 2)\n                print(""Calculating validation metrics..."")\n                val_metrics, val_gt_batch, val_recon, val_recon_full_frame = (\n                    calculate_validation_metrics(\n                        dataloader_val, optimizer.model, _rng_mask_val\n                    )\n                )\n                print(f""Step {step}, validation loss: {val_metrics['val_loss']}"")\n                val_results = {\n                    ""metrics"": val_metrics,\n                    ""gt_batch"": val_gt_batch,\n                    ""recon"": val_recon,\n                    ""full_frame"": val_recon_full_frame,\n                }\n\n            # --- Logging ---\n            if args.log:\n                if step % args.log_interval == 0 and jax.process_index() == 0:\n                    log_dict = {""loss"": loss, ""step"": step, **metrics}\n                    if val_results:\n                        log_dict.update(val_results[""metrics""])\n                    wandb.log(log_dict)\n                if step % args.log_image_interval == 0:\n                    gt_seq = batch[""videos""][0].astype(jnp.float32) / 255.0\n                    recon_seq = recon[0].clip(0, 1)\n                    comparison_seq = jnp.concatenate((gt_seq, recon_seq), axis=1)\n                    comparison_seq = einops.rearrange(\n                        comparison_seq * 255, ""t h w c -> h (t w) c""\n                    )\n                    if val_results:\n                        val_results[""gt_seq_val""] = (\n                            val_results[""gt_batch""][""videos""][0].astype(jnp.float32)\n                            / 255.0\n                        )\n                        val_results[""recon_seq_val""] = val_results[""recon""][0].clip(\n                            0, 1\n                        )\n                        val_comparison_seq = jnp.concatenate(\n                            (val_results[""gt_seq_val""], val_results[""recon_seq_val""]),\n                            axis=1,\n                        )\n                        val_results[""val_comparison_seq""] = einops.rearrange(\n                            val_comparison_seq * 255, ""t h w c -> h (t w) c""\n                        )\n                        if args.eval_full_frame:\n                            val_results[""full_frame_seq_val""] = val_results[\n                                ""full_frame""\n                            ][0].clip(0, 1)\n                            val_results[""val_full_frame_comparison_seq""] = (\n                                jnp.concatenate(\n                                    (\n                                        val_results[""gt_seq_val""],\n                                        val_results[""full_frame_seq_val""],\n                                    ),\n                                    axis=1,\n                                )\n                            )\n                            val_results[""val_full_frame_comparison_seq""] = (\n                                einops.rearrange(\n                                    val_results[""val_full_frame_comparison_seq""] * 255,\n                                    ""t h w c -> h (t w) c"",\n                                )\n                            )\n                    # NOTE: Process-dependent control flow deliberately happens\n                    # after indexing operation since it must not contain code\n                    # sections that lead to cross-accelerator communication.\n                    if jax.process_index() == 0:\n                        log_images = dict(\n                            image=wandb.Image(np.asarray(gt_seq[args.seq_len - 1])),\n                            recon=wandb.Image(np.asarray(recon_seq[args.seq_len - 1])),\n                            true_vs_recon=wandb.Image(\n                                np.asarray(comparison_seq.astype(np.uint8))\n                            ),\n                        )\n                        if val_results:\n                            log_images.update(\n                                dict(\n                                    val_image=wandb.Image(\n                                        np.asarray(\n                                            val_results[""gt_seq_val""][args.seq_len - 1]\n                                        )\n                                    ),\n                                    val_recon=wandb.Image(\n                                        np.asarray(\n                                            val_results[""recon_seq_val""][\n                                                args.seq_len - 1\n                                            ]\n                                        )\n                                    ),\n                                    val_true_vs_recon=wandb.Image(\n                                        np.asarray(\n                                            val_results[""val_comparison_seq""].astype(\n                                                np.uint8\n                                            )\n                                        )\n                                    ),\n                                )\n                            )\n                            if args.eval_full_frame:\n                                log_images.update(\n                                    dict(\n                                        val_full_frame=wandb.Image(\n                                            np.asarray(\n                                                val_results[""full_frame_seq_val""][\n                                                    args.seq_len - 1\n                                                ]\n                                            )\n                                        ),\n                                        val_true_vs_full_frame=wandb.Image(\n                                            np.asarray(\n                                                val_results[\n                                                    ""val_full_frame_comparison_seq""\n                                                ].astype(np.uint8)\n                                            )\n                                        ),\n                                    )\n                                )\n                        wandb.log(log_images)\n            # --- Checkpointing ---\n            if args.save_ckpt and step % args.log_checkpoint_interval == 0:\n                assert checkpoint_manager is not None\n                optimizer_state = nnx.state(optimizer)\n                if val_iterator:\n                    ckpt_manager_args = ocp.args.Composite(\n                        model_state=ocp.args.PyTreeSave(optimizer_state),  # type: ignore\n                        train_dataloader_state=grain.checkpoint.CheckpointSave(  # type: ignore\n                            train_iterator  # type: ignore\n                        ),\n                        val_dataloader_state=grain.checkpoint.CheckpointSave(  # type: ignore\n                            val_iterator  # type: ignore\n                        ),\n                    )\n                else:\n                    ckpt_manager_args = ocp.args.Composite(\n                        model_state=ocp.args.PyTreeSave(optimizer_state),  # type: ignore\n                        train_dataloader_state=grain.checkpoint.CheckpointSave(  # type: ignore\n                            train_iterator  # type: ignore\n                        ),\n                    )\n                checkpoint_manager.save(step, args=ckpt_manager_args)\n                print(f""Saved checkpoint at step {step}"")\n            if step >= args.num_steps:\n                break\n\n    if checkpoint_manager:\n        checkpoint_manager.close()\n\n\nif __name__ == ""__main__"":\n    args = tyro.cli(Args)\n    main(args)\n",python,tab
+6,418953,"jasmine/train_dynamics.py",1403,0,"",python,selection_mouse
+7,418974,"jasmine/train_dynamics.py",1402,0,"",python,selection_command
+8,421044,"jasmine/train_dynamics.py",6773,0,"",python,selection_mouse
+9,421047,"jasmine/train_dynamics.py",6772,0,"",python,selection_command
+10,425324,"jasmine/train_dynamics.py",8477,0,"",python,selection_command
+11,426023,"jasmine/train_dynamics.py",8558,0,"",python,selection_command
+12,426272,"jasmine/train_dynamics.py",11188,0,"",python,selection_command
+13,427474,"jasmine/train_dynamics.py",11345,0,"",python,selection_command
+14,427867,"jasmine/train_dynamics.py",11766,0,"",python,selection_command
+15,431824,"jasmine/train_dynamics.py",11755,0,"",python,selection_mouse
+16,431991,"jasmine/train_dynamics.py",11751,7,"outputs",python,selection_mouse
+17,438594,"jasmine/train_dynamics.py",11832,0,"",python,selection_mouse
+18,438754,"jasmine/train_dynamics.py",11825,7,"ce_loss",python,selection_mouse
+19,512377,"TERMINAL",0,0,"queue",,terminal_command
+20,512442,"TERMINAL",0,0,"]633;C[?1049h(B[?7hEvery 1.0s: squeue --mehkn1993.localdomain: Tue Sep 30 23:36:23 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3533426 accelerat train_dy tum_cte0  R    8:32:16\t  1 hkn04233532468 accelerat train_dy tum_cte0  R   11:46:53\t  1 hkn08073532465 accelerat train_dy tum_cte0  R   11:47:36\t  1 hkn05073532466 accelerat train_dy tum_cte0  R   11:47:36\t  1 hkn05073532383 accelerat train_to tum_cte0  R   11:54:36\t  1 hkn08073529698 accelerat train_to tum_cte0  R 1-03:00:30\t  1 hkn04293533705     large generate tum_cte0  R    7:37:34\t  1 hkn1901",,terminal_output
+21,513489,"TERMINAL",0,0,"47477715",,terminal_output
+22,514552,"TERMINAL",0,0,"58588826",,terminal_output
+23,515726,"TERMINAL",0,0,"69699937",,terminal_output
+24,517006,"TERMINAL",0,0,"720740404048",,terminal_output
+25,517743,"TERMINAL",0,0,"81811159",,terminal_output
+26,519188,"TERMINAL",0,0,"929222640",,terminal_output
+27,519999,"TERMINAL",0,0,"3037:0033371",,terminal_output
+28,520805,"TERMINAL",0,0,"14144482",,terminal_output
+29,521872,"TERMINAL",0,0,"25255593",,terminal_output
+30,522917,"TERMINAL",0,0,"363666404",,terminal_output
+31,523962,"TERMINAL",0,0,"47477715",,terminal_output
+32,524982,"TERMINAL",0,0,"58588826",,terminal_output
+33,525971,"TERMINAL",0,0,"69699937",,terminal_output
+34,526920,"TERMINAL",0,0,"730750505048",,terminal_output
+35,527965,"TERMINAL",0,0,"81811159",,terminal_output
+36,529012,"TERMINAL",0,0,"929222650",,terminal_output
+37,530117,"TERMINAL",0,0,"4031033371",,terminal_output
+38,531140,"TERMINAL",0,0,"15255593",,terminal_output
+39,532142,"TERMINAL",0,0,"363666504",,terminal_output
+40,533190,"TERMINAL",0,0,"47477715",,terminal_output
+41,534303,"TERMINAL",0,0,"58588826",,terminal_output
+42,535440,"TERMINAL",0,0,"69699937",,terminal_output
+43,536464,"TERMINAL",0,0,"74078:008:005:0048",,terminal_output
+44,537395,"TERMINAL",0,0,"81811159",,terminal_output
+45,538560,"TERMINAL",0,0,"92922268:00",,terminal_output
+46,539677,"TERMINAL",0,0,"5032033371",,terminal_output
+47,540701,"TERMINAL",0,0,"14144482",,terminal_output
+48,541763,"TERMINAL",0,0,"25255593",,terminal_output
+49,542806,"TERMINAL",0,0,"3636661:004",,terminal_output
+50,543751,"TERMINAL",0,0,"47477715",,terminal_output
+51,544086,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+52,550677,"TERMINAL",0,0,"git branch",,terminal_command
+53,550747,"TERMINAL",0,0,"]633;C[?1h=\r",,terminal_output
+54,550895,"TERMINAL",0,0,"  ablation/use-pytorch-dataloader\r\n  action-mapper\r\n  add-noise-to-combat-exposure-bias\r\n  add-wandb-name-and-tags\r\n  before-nnx\r\n  causal-mem-reduce\r\n  causal-spatiotemporal-kv-cache\r\n  causal-st-transformer\r\n  causal-transformer-dynamics-model\r\n  causal-transformer-nnx-no-kv-cache\r\n  change-default-parameters\r\n* change-default-to-wsd\r\n  coinrun-gt-actions\r\n:",,terminal_output
+55,554416,"TERMINAL",0,0,"\r\r:",,terminal_output
+56,555067,"TERMINAL",0,0,"\r\r:",,terminal_output
+57,555161,"TERMINAL",0,0,"\r\r:\r\r:\r\r:\r\r:",,terminal_output
+58,557447,"TERMINAL",0,0,"\r/",,terminal_output
+59,565140,"TERMINAL",0,0,"cc",,terminal_output
+60,565209,"TERMINAL",0,0,"oo",,terminal_output
+61,565506,"TERMINAL",0,0,"ii",,terminal_output
+62,565613,"TERMINAL",0,0,"nn",,terminal_output
+63,566542,"TERMINAL",0,0,"rruu",,terminal_output
+64,566607,"TERMINAL",0,0,"nn",,terminal_output
+65,567107,"TERMINAL",0,0,"\r  ablation/use-pytorch-dataloader\r\n  action-mapper\r\n  add-noise-to-combat-exposure-bias\r\n  add-wandb-name-and-tags\r\n  before-nnx\r\n  causal-mem-reduce\r\n  causal-spatiotemporal-kv-cache\r\n  causal-st-transformer\r\n  causal-transformer-dynamics-model\r\n  causal-transformer-nnx-no-kv-cache\r\n  change-default-parameters\r\n* change-default-to-wsd\r\n  coinrun-gt-actions\r\n  ablation/use-pytorch-dataloader\r\n  action-mapper\r\n  add-noise-to-combat-exposure-bias\r\n  add-wandb-name-and-tags\r\n  before-nnx\r\n  causal-mem-reduce\r\n  causal-spatiotemporal-kv-cache\r\n  causal-st-transformer\r\n  causal-transformer-dynamics-model\r\n  causal-transformer-nnx-no-kv-cache\r\n  change-default-parameters\r\n* change-default-to-wsd\r\n  coinrun-gt-actions\r\n  convert-to-jax-array-in-iter\r\n  correct-batched-sampling\r\n  dev\r\n  dont-let-tf-see-gpu\r\n  feat/darkness-filter\r\n  feat/explicit-image-dims\r\n  fix-action-padding-lam-future-information-access\r\n  fix-sampling\r\n  fix-transformer-forwardpass\r\n  fix/dyn-restore-after-nnx-upgrade\r\n  fix/spatiotemporal-pe-once-in-STTransformer\r\n  generate-minatar-breakout-dataset\r\n:",,terminal_output
+66,568485,"TERMINAL",0,0,"\r/\r...skipping...\r\n  train_lam_coinrun_ablation_wsd_3e-6_28747\r\n  val-loss\r\n  vizdoom-dataset\r\n  z-loss\r\n  zloss-runs\r\n~\r\n~\r\n~\r\n~\r\n~\r\n~\r\n~\r\n~\r\n(END)",,terminal_output
+67,569937,"TERMINAL",0,0,"\r/\r\rPattern not found  (press RETURN)",,terminal_output
+68,571487,"TERMINAL",0,0,"\r/",,terminal_output
+69,573023,"TERMINAL",0,0,"dd",,terminal_output
+70,573183,"TERMINAL",0,0,"aa",,terminal_output
+71,573366,"TERMINAL",0,0,"tt",,terminal_output
+72,573464,"TERMINAL",0,0,"aa",,terminal_output
+73,573890,"TERMINAL",0,0,"\r  train_lam_coinrun_ablation_wsd_3e-6_28747\r\n  val-loss\r\n  vizdoom-dataset\r\n  z-loss\r\n  zloss-runs\r\n~\r\n  train_lam_coinrun_ablation_wsd_3e-6_28747\r\n  val-loss\r\n  vizdoom-dataset\r\n  z-loss\r\n  zloss-runs\r\n~\r\n~\r\n~\r\n(END)",,terminal_output
+74,575994,"TERMINAL",0,0,"\r...skipping...\r\n  runner\r\n  runner-grain\r\n  sample-ali-branch\r\n  sample-from-different-topologies\r\n  sampling-script-add-metrics\r\n  sampling-startframe-indexing-fix\r\n  seeding-data-generation\r\n  speedup-tfrecord-preprocessing\r\n  train_lam_coinrun_ablation_wsd_3e-6_28747\r\n  val-loss\r\n  vizdoom-dataset\r\n  z-loss\r\n  zloss-runs\r\n(END)",,terminal_output
+75,582829,"TERMINAL",0,0,"\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+76,586889,"TERMINAL",0,0,"git checkout seeding-data-generation",,terminal_command
+77,586941,"TERMINAL",0,0,"]633;C",,terminal_output
+78,587175,"TERMINAL",0,0,"Switched to branch 'seeding-data-generation'\r\nYour branch is up to date with 'origin/seeding-data-generation'.\r\n]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+79,587637,"jasmine/train_dynamics.py",895,16196,"    image_height: int = 90\n    image_width: int = 160\n    data_dir: str = """"\n    save_ckpt: bool = False\n    restore_ckpt: bool = False\n    # Optimization\n    batch_size: int = 36\n    init_lr: float = 0.0\n    max_lr: float = 3e-5\n    decay_end: float = 0.0\n    wsd_decay_steps: int = (\n        10000  # NOTE: wsd_decay_steps will only be used when using a wsd-schedule\n    )\n    warmup_steps: int = 5000\n    lr_schedule: str = ""wsd""  # supported options: wsd, cos\n    # Tokenizer\n    tokenizer_dim: int = 512\n    tokenizer_ffn_dim: int = 2048\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 4\n    tokenizer_num_blocks: int = 4\n    tokenizer_num_heads: int = 8\n    tokenizer_checkpoint: str = """"\n    # LAM\n    lam_dim: int = 512\n    lam_ffn_dim: int = 2048\n    latent_action_dim: int = 32\n    num_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 4\n    lam_num_heads: int = 8\n    lam_checkpoint: str = """"\n    # Dynamics\n    dyna_type: str = ""maskgit""  # supported options: maskgit, causal\n    dyna_dim: int = 512\n    dyna_ffn_dim: int = 2048\n    dyna_num_blocks: int = 6\n    dyna_num_heads: int = 8\n    dropout: float = 0.0\n    mask_limit: float = 0.5\n    param_dtype = jnp.float32\n    dtype = jnp.bfloat16\n    use_flash_attention: bool = True\n    use_gt_actions: bool = False\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    name: str = ""train_dynamics""\n    tags: list[str] = field(default_factory=lambda: [""dynamics""])\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 25000\n    log_checkpoint_keep_period: int = 20000\n    log_gradients: bool = False\n    val_data_dir: str = """"\n    val_interval: int = 20_000\n    val_steps: int = 50\n    eval_full_frame: bool = False\n    val_maskgit_steps: int = 25\n    val_temperature: float = 1\n    val_sample_argmax: bool = False\n    wandb_id: str = """"\n\n\ndef build_model(args: Args, rng: jax.Array) -> tuple[Genie, jax.Array]:\n    rng, _rng = jax.random.split(rng)\n    rngs = nnx.Rngs(_rng)\n    genie = Genie(\n        # Tokenizer\n        in_dim=args.image_channels,\n        tokenizer_dim=args.tokenizer_dim,\n        tokenizer_ffn_dim=args.tokenizer_ffn_dim,\n        latent_patch_dim=args.latent_patch_dim,\n        num_patch_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        tokenizer_num_blocks=args.tokenizer_num_blocks,\n        tokenizer_num_heads=args.tokenizer_num_heads,\n        # LAM\n        lam_dim=args.lam_dim,\n        lam_ffn_dim=args.lam_ffn_dim,\n        latent_action_dim=args.latent_action_dim,\n        num_actions=args.num_actions,\n        lam_patch_size=args.lam_patch_size,\n        lam_num_blocks=args.lam_num_blocks,\n        lam_num_heads=args.lam_num_heads,\n        lam_co_train=not args.lam_checkpoint,\n        use_gt_actions=args.use_gt_actions,\n        # Dynamics\n        dyna_type=args.dyna_type,\n        dyna_dim=args.dyna_dim,\n        dyna_ffn_dim=args.dyna_ffn_dim,\n        dyna_num_blocks=args.dyna_num_blocks,\n        dyna_num_heads=args.dyna_num_heads,\n        dropout=args.dropout,\n        mask_limit=args.mask_limit,\n        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n        use_flash_attention=args.use_flash_attention,\n        decode=False,\n        rngs=rngs,\n    )\n    if args.use_gt_actions:\n        assert (\n            not args.lam_checkpoint\n        ), ""Cannot use LAM when using ground-truth actions.""\n    else:\n        assert genie.lam is not None\n        del genie.lam.decoder\n    return genie, rng\n\n\ndef build_optimizer(genie: Genie, args: Args) -> nnx.ModelAndOptimizer:\n    lr_schedule = get_lr_schedule(\n        args.lr_schedule,\n        args.init_lr,\n        args.max_lr,\n        args.decay_end,\n        args.num_steps,\n        args.warmup_steps,\n        args.wsd_decay_steps,\n    )\n    tx = optax.adamw(\n        learning_rate=lr_schedule,\n        b1=0.9,\n        b2=0.9,\n        weight_decay=1e-4,\n        mu_dtype=args.param_dtype,  # moments in full precision\n    )\n    optimizer = nnx.ModelAndOptimizer(genie, tx)\n    return optimizer\n\n\ndef build_mesh_and_sharding(\n    num_devices: int,\n) -> tuple[Mesh, NamedSharding, NamedSharding, NamedSharding]:\n    device_mesh_arr = create_device_mesh((num_devices,))\n    mesh = Mesh(devices=device_mesh_arr, axis_names=(""data"",))\n    replicated_sharding = NamedSharding(mesh, PartitionSpec())\n    videos_sharding = NamedSharding(mesh, PartitionSpec(""data"", None, None, None, None))\n    actions_sharding = NamedSharding(mesh, PartitionSpec(""data"", None))\n    return mesh, replicated_sharding, videos_sharding, actions_sharding\n\n\ndef shard_optimizer_states(\n    optimizer: nnx.ModelAndOptimizer, replicated_sharding: NamedSharding\n) -> None:\n    model_state = nnx.state(optimizer.model)\n    model_sharded_state = jax.lax.with_sharding_constraint(\n        model_state, replicated_sharding\n    )\n    nnx.update(optimizer.model, model_sharded_state)\n    optimizer_state = nnx.state(optimizer, nnx.optimizer.OptState)\n    optimizer_sharded_state = jax.lax.with_sharding_constraint(\n        optimizer_state, replicated_sharding\n    )\n    nnx.update(optimizer, optimizer_sharded_state)\n\n\ndef build_dataloader(args: Args, data_dir: str) -> grain.DataLoaderIterator:\n    image_shape = (args.image_height, args.image_width, args.image_channels)\n    array_record_files = [\n        os.path.join(data_dir, x)\n        for x in os.listdir(data_dir)\n        if x.endswith("".array_record"")\n    ]\n    grain_dataloader = get_dataloader(\n        array_record_files,\n        args.seq_len,\n        # NOTE: We deliberately pass the global batch size\n        # The dataloader shards the dataset across all processes\n        args.batch_size,\n        *image_shape,\n        num_workers=8,\n        prefetch_buffer_size=1,\n        seed=args.seed,\n    )\n    initial_state = grain_dataloader._create_initial_state()\n    grain_iterator = grain.DataLoaderIterator(grain_dataloader, initial_state)\n    return grain_iterator\n\n\ndef build_checkpoint_manager(args: Args) -> Optional[ocp.CheckpointManager]:\n    if args.restore_ckpt or args.save_ckpt:\n        handler_registry = ocp.handlers.DefaultCheckpointHandlerRegistry()\n        handler_registry.add(\n            ""model_state"", ocp.args.PyTreeSave, ocp.handlers.PyTreeCheckpointHandler\n        )\n        handler_registry.add(\n            ""model_state"", ocp.args.PyTreeRestore, ocp.handlers.PyTreeCheckpointHandler\n        )\n        handler_registry.add(\n            ""train_dataloader_state"",\n            grain.checkpoint.CheckpointSave,\n            cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n        )\n        handler_registry.add(\n            ""train_dataloader_state"",\n            grain.checkpoint.CheckpointRestore,\n            cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n        )\n        if args.val_data_dir:\n            handler_registry.add(\n                ""val_dataloader_state"",\n                grain.checkpoint.CheckpointSave,\n                cast(\n                    ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler\n                ),\n            )\n            handler_registry.add(\n                ""val_dataloader_state"",\n                grain.checkpoint.CheckpointRestore,\n                cast(\n                    ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler\n                ),\n            )\n        checkpoint_options = ocp.CheckpointManagerOptions(\n            save_interval_steps=args.log_checkpoint_interval,\n            max_to_keep=3,\n            keep_period=args.log_checkpoint_keep_period,\n            step_format_fixed_length=6,\n            cleanup_tmp_directories=True,\n        )\n        checkpoint_manager = ocp.CheckpointManager(\n            args.ckpt_dir,\n            options=checkpoint_options,\n            handler_registry=handler_registry,\n        )\n        return checkpoint_manager\n    else:\n        return None\n\n\ndef restore_or_initialize_components(\n    args: Args,\n    checkpoint_manager: Optional[ocp.CheckpointManager],\n    optimizer: nnx.ModelAndOptimizer,\n    train_iterator: grain.DataLoaderIterator,\n    rng: jax.Array,\n    replicated_sharding: NamedSharding,\n    val_iterator: Optional[grain.DataLoaderIterator],\n    restore_step: Optional[int] = None,\n) -> tuple[\n    int,\n    nnx.ModelAndOptimizer,\n    grain.DataLoaderIterator,\n    grain.DataLoaderIterator,\n    jax.Array,\n]:\n    step = 0\n    if checkpoint_manager and restore_step is None:\n        restore_step = checkpoint_manager.latest_step()\n    if args.restore_ckpt:\n        assert checkpoint_manager is not None\n        abstract_optimizer = nnx.eval_shape(lambda: optimizer)\n        abstract_optimizer_state = nnx.state(abstract_optimizer)\n        if val_iterator:\n            restore_args = ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                train_dataloader_state=grain.checkpoint.CheckpointRestore(train_iterator),  # type: ignore\n                val_dataloader_state=grain.checkpoint.CheckpointRestore(val_iterator),  # type: ignore\n            )\n        else:\n            restore_args = ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                train_dataloader_state=grain.checkpoint.CheckpointRestore(train_iterator),  # type: ignore\n            )\n        restored = checkpoint_manager.restore(\n            checkpoint_manager.latest_step(), args=restore_args\n        )\n        restored_optimizer_state = restored[""model_state""]\n        nnx.update(optimizer, restored_optimizer_state)\n        train_iterator = restored[""train_dataloader_state""]\n        if val_iterator:\n            val_iterator = restored[""val_dataloader_state""]\n        step = checkpoint_manager.latest_step() or 0\n        print(f""Restored dataloader and model state from step {step}"")\n    else:\n        # Restore from pre-trained tokenizer (and LAM)\n        rng, _rng = jax.random.split(rng)\n        optimizer = restore_genie_components(optimizer, replicated_sharding, _rng, args)\n    return step, optimizer, train_iterator, val_iterator, rng\n\n\ndef _calculate_step_metrics(\n    outputs: dict[str, jax.Array],\n    gt: jax.Array,\n    num_actions: int,\n    num_patch_latents: int,\n) -> tuple[jax.Array, dict]:\n    mask = outputs[""mask""]\n    outputs[""token_logits""] = outputs[""token_logits""].astype(jnp.float32)\n    ce_loss = optax.softmax_cross_entropy_with_integer_labels(\n        outputs[""token_logits""], outputs[""video_tokens""]\n    )\n    ce_loss = (mask * ce_loss).sum() / mask.sum()\n    acc = outputs[""token_logits""].argmax(-1) == outputs[""video_tokens""]\n    acc = (mask * acc).sum() / mask.sum()\n    select_probs = jax.nn.softmax(outputs[""token_logits""])\n    gt_val = gt.clip(0, 1).reshape(-1, *gt.shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = jnp.asarray(pix.psnr(gt_val, recon)).mean()\n    ssim = jnp.asarray(pix.ssim(gt_val, recon)).mean()\n    _, index_counts_tokenizer = jnp.unique_counts(\n        jnp.ravel(outputs[""video_tokens""]),\n        size=num_patch_latents,\n        fill_value=0,\n    )\n    codebook_usage_tokenizer = (index_counts_tokenizer != 0).mean()\n    metrics = dict(\n        cross_entropy_loss=ce_loss,\n        masked_token_accuracy=acc,\n        select_logit=outputs[""token_logits""].max(-1).mean(),\n        select_p=select_probs.max(-1).mean(),\n        entropy=jax.scipy.special.entr(select_probs).sum(-1).mean(),\n        psnr=psnr,\n        ssim=ssim,\n        codebook_usage_tokenizer=codebook_usage_tokenizer,\n    )\n    if ""lam_indices"" in outputs.keys():\n        _, index_counts_lam = jnp.unique_counts(\n            jnp.ravel(outputs[""lam_indices""]),\n            size=num_actions,\n            fill_value=0,\n        )\n        codebook_usage_lam = (index_counts_lam != 0).mean()\n        metrics[""codebook_usage_lam""] = codebook_usage_lam\n    return ce_loss, metrics\n\n\ndef main(args: Args) -> None:\n    jax.distributed.initialize()\n    num_devices = jax.device_count()\n    if num_devices == 0:\n        raise ValueError(""No JAX devices found."")\n    print(f""Running on {num_devices} devices."")\n\n    if args.batch_size % num_devices != 0:\n        raise ValueError(\n            f""Global batch size {args.batch_size} must be divisible by ""\n            f""number of devices {num_devices}.""\n        )\n\n    rng = jax.random.key(args.seed)\n\n    # --- Initialize model ---\n    genie, rng = build_model(args, rng)\n    _, params, _ = nnx.split(genie, nnx.Param, ...)\n    param_counts = count_parameters_by_component(params)\n\n    if args.log and jax.process_index() == 0:\n        wandb_init_kwargs = {\n            ""entity"": args.entity,\n            ""project"": args.project,\n            ""name"": args.name,\n            ""tags"": args.tags,\n            ""group"": ""debug"",\n            ""config"": args,\n        }\n\n        if args.wandb_id:\n            wandb_init_kwargs.update(\n                {\n                    ""id"": args.wandb_id,\n                    ""resume"": ""allow"",\n                }\n            )\n        wandb.init(**wandb_init_kwargs)\n\n        wandb.config.update({""model_param_count"": param_counts})\n\n    print(""Parameter counts:"")\n    print(param_counts)\n\n    # --- Initialize optimizer ---\n    optimizer = build_optimizer(genie, args)\n    del genie\n\n    # FIXME: switch to create_hybrid_device_mesh for runs spanning multiple nodes\n    _, replicated_sharding, videos_sharding, actions_sharding = build_mesh_and_sharding(\n        num_devices\n    )\n\n    shard_optimizer_states(optimizer, replicated_sharding)\n\n    # --- Initialize checkpoint manager ---\n    checkpoint_manager = build_checkpoint_manager(args)\n\n    # --- Create DataLoaderIterator from dataloader ---\n    train_iterator = build_dataloader(args, args.data_dir)\n    val_iterator = None\n    if args.val_data_dir:\n        val_iterator = build_dataloader(args, args.val_data_dir)\n\n    # --- Restore checkpoint ---\n    step, optimizer, train_iterator, val_iterator, rng = (\n        restore_or_initialize_components(\n            args,\n            checkpoint_manager,\n            optimizer,\n            train_iterator,\n            rng,\n            replicated_sharding,\n            val_iterator,\n        )\n    )\n\n    # --- Define loss and train step (close over args) ---\n    def dynamics_loss_fn(\n        model: Genie,\n        inputs: dict,\n    ) -> tuple[jax.Array, tuple[jax.Array, dict]]:\n        gt = jnp.asarray(inputs[""videos""], dtype=jnp.float32) / 255.0\n        inputs[""videos""] = gt.astype(args.dtype)\n        outputs = model(inputs)\n        ce_loss, metrics = _calculate_step_metrics(\n            outputs, gt, args.num_actions, args.num_patch_latents\n        )\n        return ce_loss, (outputs[""recon""], metrics)\n",python,content
+80,588234,"jasmine/train_dynamics.py",0,0,"Switched from branch 'change-default-to-wsd' to 'seeding-data-generation'",python,git_branch_checkout
+81,589551,"TERMINAL",0,0,"git pull",,terminal_command
+82,589600,"TERMINAL",0,0,"]633;C",,terminal_output
+83,591670,"TERMINAL",0,0,"remote: Enumerating objects: 11, done.\r\nremote: Counting objects:   9% (1/11)\rremote: Counting objects:  18% (2/11)\rremote: Counting objects:  27% (3/11)\rremote: Counting objects:  36% (4/11)\rremote: Counting objects:  45% (5/11)\rremote: Counting objects:  54% (6/11)\rremote: Counting objects:  63% (7/11)\rremote: Counting objects:  72% (8/11)\rremote: Counting objects:  81% (9/11)\rremote: Counting objects:  90% (10/11)\rremote: Counting objects: 100% (11/11)\rremote: Counting objects: 100% (11/11), done.\r\nremote: Compressing objects:  50% (1/2)\rremote: Compressing objects: 100% (2/2)\rremote: Compressing objects: 100% (2/2), done.\r\nremote: Total 6 (delta 4), reused 5 (delta 4), pack-reused 0 (from 0)\r\nUnpacking objects:  16% (1/6)\rUnpacking objects:  33% (2/6)\rUnpacking objects:  50% (3/6)\rUnpacking objects:  66% (4/6)\rUnpacking objects:  83% (5/6)\rUnpacking objects: 100% (6/6)\rUnpacking objects: 100% (6/6), 1.33 KiB | 52.00 KiB/s, done.\r\n",,terminal_output
+84,591826,"TERMINAL",0,0,"From github.com:p-doom/jasmine\r\n   6a0b4aa..b8c92e7  main       -> origin/main\r\n",,terminal_output
+85,591904,"TERMINAL",0,0,"Already up to date.\r\n]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+86,593016,"slurm/jobs/franz/berlin/coinrun/mila_submission/ablations/coinrun_dynamics_no_flash_attn.sh",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=24:00:00\n#SBATCH --cpus-per-task=8\n#SBATCH --gres=gpu:1\n#SBATCH --output=/fast/project/HFMI_SynergyUnit/jafar_ws/logs/franz/coinrun/dynamics/%x_%j.log\n#SBATCH --error=/fast/project/HFMI_SynergyUnit/jafar_ws/logs/franz/coinrun/dynamics/%x_%j.log\n#SBATCH --job-name=dynamics_coinrun_mila_submission_no_flash_attn\n#SBATCH --requeue\n#SBATCH --signal=b:usr1@300 # 5 min before timeout\n\n# --- signal trap to requeue job before timeout ---\nrequeue_job() {\n    echo ""[$(date)] caught sigusr1 (timeout warning), requeueing slurm job $SLURM_JOB_ID...""\n    # optional: trigger checkpoint saving here\n    # e.g., touch $checkpoint_dir/requeue_trigger\n    scontrol requeue $SLURM_JOB_ID\n    exit 0\n}\n\ntrap requeue_job sigusr1\n\n# set checkpoint flag based on restart count\nrestart_count=$(scontrol show job $SLURM_JOB_ID | grep -o 'Restarts=[0-9]*' | cut -d'=' -f2)\n\nif [ $restart_count -eq 0 ]; then\n    restore_ckpt_flag=""--no-restore-ckpt""\nelse\n    restore_ckpt_flag=""--restore-ckpt""\nfi\n\n\n\n# Log the sbatch script\ncat $0\n\nsource .venv/bin/activate\n\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\ntags=""coinrun dynamics 500m_dataset mila_submission ablation no-flash-attn""\n\narray_records_dir=""/fast/project/HFMI_SynergyUnit/jafar_ws/data/coinrun/array_records_500m_seed_w_increment""\ntokenizer_ckpt_dir=""/fast/project/HFMI_SynergyUnit/jafar_ws/checkpoints/coinrun/tokenizer/tokenizer_coinrun_mila_submission_29736/""\nCHECKPOINT_DIR=""/fast/project/HFMI_SynergyUnit/jafar_ws/checkpoints/coinrun/dynamics/${job_name}/${slurm_job_id}""\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\n\nsrun python jasmine/train_dynamics.py \\n    --no-use-flash-attention \\n    --save_ckpt \\n    $restore_ckpt_flag \\n    --wandb_id $SLURM_JOB_ID \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --name=""${job_name}_${slurm_job_id}"" \\n    --tags ${tags} \\n    --entity instant-uv \\n    --project jafar \\n    --tokenizer_checkpoint=""${tokenizer_ckpt_dir}"" \\n    --val_data_dir=""${array_records_dir}/val"" \\n    --data_dir=""${array_records_dir}/train"" &\n\nchild_pid=$!\n\nwait $child_pid\n\n",shellscript,tab
+87,599160,"slurm/jobs/franz/berlin/coinrun/mila_submission/ablations/coinrun_dynamics_no_flash_attn.sh",0,0,"",shellscript,tab
+88,618818,"slurm/jobs/mihir/horeka/preprocessing/coinrun_chunked_500m.sh",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=48:00:00\n#SBATCH --partition=large\n#SBATCH --cpus-per-task=16\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/preprocess/coinrun/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/preprocess/coinrun/%x_%j.log\n#SBATCH --job-name=generate_coinrun_chunked_500m\n\ncd data\nsource .venv/bin/activate\n\npython jasmine_data/coinrun/generate_coinrun_dataset.py \\n    --min_episode_length 1000 \\n    --num_episodes_train 500000 \\n    --num_episodes_val 100 \\n    --num_episodes_test 100 \\n    --chunk_size 100 \\n    --output_dir /hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_coinrun/coinrun_episodes_500m_gt_actions_split",shellscript,tab
+89,620149,"slurm/jobs/mihir/horeka/preprocessing/coinrun_chunked_500m.sh",84,0,"",shellscript,selection_mouse
+90,626578,"TERMINAL",0,0,"fqueue",,terminal_command
+91,626654,"TERMINAL",0,0,"]633;C[?1049h(B[?7hEvery 1.0s: squeue -o ""%.10i %.16P %.30j %.8u %.8T %.10M %.9l %.6D %R""hkn1993.localdomain: Tue Sep 30 23:38:17 2025JOBIDPARTITIONNAME     USER    STATE\t   TIME TIME_LIMI  NODES NODELIST(REASON)3533426\tacceleratedtrain_dynamics_maskgit tum_cte0  RUNNING    8:34:10 2-00:00:00\t1 hkn04233532468\tacceleratedtrain_dynamics_maskgit tum_cte0  RUNNING   11:48:47 2-00:00:00\t1 hkn08073532465\tacceleratedtrain_dynamics_maskgit tum_cte0  RUNNING   11:49:30 2-00:00:00\t1 hkn05073532466\tacceleratedtrain_dynamics_maskgit tum_cte0  RUNNING   11:49:30 2-00:00:00\t1 hkn05073532383\tacceleratedtrain_tokenizer_default tum_cte0  RUNNING   11:56:30 2-00:00:00\t1 hkn08073529698\tacceleratedtrain_tokenizer_default tum_cte0  RUNNING 1-03:02:24 2-00:00:00\t1 hkn04293533705large  generate_coinrun_chunked_500m tum_cte0  RUNNING    7:39:28 2-00:00:00\t1 hkn1901",,terminal_output
+92,627711,"TERMINAL",0,0,"81811159",,terminal_output
+93,628911,"TERMINAL",0,0,"929222630",,terminal_output
+94,630136,"TERMINAL",0,0,"2035033371",,terminal_output
+95,630950,"TERMINAL",0,0,"14144482",,terminal_output
+96,631765,"TERMINAL",0,0,"25255593",,terminal_output
+97,632927,"TERMINAL",0,0,"363666304",,terminal_output
+98,633868,"TERMINAL",0,0,"47477715",,terminal_output
+99,634861,"TERMINAL",0,0,"58588826",,terminal_output
+100,635903,"TERMINAL",0,0,"69699937",,terminal_output
+101,636841,"TERMINAL",0,0,"720740404048",,terminal_output
+102,638031,"TERMINAL",0,0,"81811159",,terminal_output
+103,639070,"TERMINAL",0,0,"929222640",,terminal_output
+104,639992,"TERMINAL",0,0,"3039:0033371",,terminal_output
+105,640996,"TERMINAL",0,0,"14144482",,terminal_output
+106,642067,"TERMINAL",0,0,"25255593",,terminal_output
+107,643080,"TERMINAL",0,0,"363666404",,terminal_output
+108,644074,"TERMINAL",0,0,"47477715",,terminal_output
+109,645185,"TERMINAL",0,0,"58588826",,terminal_output
+110,646129,"TERMINAL",0,0,"630750505048",,terminal_output
+111,647356,"TERMINAL",0,0,"81811159",,terminal_output
+112,648341,"TERMINAL",0,0,"929222650",,terminal_output
+113,649414,"TERMINAL",0,0,"4031033371",,terminal_output
+114,650256,"TERMINAL",0,0,"14144482",,terminal_output
+115,651388,"TERMINAL",0,0,"25255593",,terminal_output
+116,652436,"TERMINAL",0,0,"363666504",,terminal_output
+117,653356,"TERMINAL",0,0,"47477715",,terminal_output
+118,654437,"TERMINAL",0,0,"58588826",,terminal_output
+119,655565,"TERMINAL",0,0,"69699937",,terminal_output
+120,656887,"TERMINAL",0,0,"740750:0050:007:0048",,terminal_output
+121,657814,"TERMINAL",0,0,"81811159",,terminal_output
+122,658992,"TERMINAL",0,0,"929222640:00",,terminal_output
+123,659647,"TERMINAL",0,0,"5032033371",,terminal_output
+124,660677,"TERMINAL",0,0,"14144482",,terminal_output
+125,661700,"TERMINAL",0,0,"25255593",,terminal_output
+126,662725,"TERMINAL",0,0,"3636663:004",,terminal_output
+127,663718,"TERMINAL",0,0,"47477715",,terminal_output
+128,664849,"TERMINAL",0,0,"58588826",,terminal_output
+129,665850,"TERMINAL",0,0,"69699937",,terminal_output
+130,666807,"TERMINAL",0,0,"750710101048",,terminal_output
+131,667899,"TERMINAL",0,0,"81811159",,terminal_output
+132,669026,"TERMINAL",0,0,"929222610",,terminal_output
+133,669900,"TERMINAL",0,0,"9:0033033371",,terminal_output
+134,671025,"TERMINAL",0,0,"14144482",,terminal_output
+135,672136,"TERMINAL",0,0,"25255593",,terminal_output
+136,673075,"TERMINAL",0,0,"363666104",,terminal_output
+137,674101,"TERMINAL",0,0,"47477715",,terminal_output
+138,675068,"TERMINAL",0,0,"58588826",,terminal_output
+139,676258,"TERMINAL",0,0,"65:00720202048",,terminal_output
+140,677267,"TERMINAL",0,0,"81811159",,terminal_output
+141,678449,"TERMINAL",0,0,"929222620",,terminal_output
+142,679415,"TERMINAL",0,0,"1034033371",,terminal_output
+143,680344,"TERMINAL",0,0,"14144482",,terminal_output
+144,681599,"TERMINAL",0,0,"25255593",,terminal_output
+145,682379,"TERMINAL",0,0,"363666204",,terminal_output
+146,683425,"TERMINAL",0,0,"47477715",,terminal_output
+147,684518,"TERMINAL",0,0,"58588826",,terminal_output
+148,685418,"TERMINAL",0,0,"69699937",,terminal_output
+149,686483,"TERMINAL",0,0,"710730303048",,terminal_output
+150,687774,"TERMINAL",0,0,"81811159",,terminal_output
+151,688675,"TERMINAL",0,0,"929222630",,terminal_output
+152,689740,"TERMINAL",0,0,"2035033371",,terminal_output
+153,690769,"TERMINAL",0,0,"14144482",,terminal_output
+154,692025,"TERMINAL",0,0,"25255593",,terminal_output
+155,693019,"TERMINAL",0,0,"363666304",,terminal_output
+156,694009,"TERMINAL",0,0,"47477715",,terminal_output
+157,695301,"TERMINAL",0,0,"58588826",,terminal_output
+158,696209,"TERMINAL",0,0,"69699937",,terminal_output
+159,697140,"TERMINAL",0,0,"720740404048",,terminal_output
+160,698116,"TERMINAL",0,0,"81811159",,terminal_output
+161,699175,"TERMINAL",0,0,"929222640",,terminal_output
+162,700317,"TERMINAL",0,0,"30350:0033371",,terminal_output
+163,701220,"TERMINAL",0,0,"14144482",,terminal_output
+164,702245,"TERMINAL",0,0,"25255593",,terminal_output
+165,703351,"TERMINAL",0,0,"363666404",,terminal_output
+166,704812,"TERMINAL",0,0,"47477715",,terminal_output
+167,705412,"TERMINAL",0,0,"58588826",,terminal_output
+168,706329,"TERMINAL",0,0,"69699937",,terminal_output
+169,707303,"TERMINAL",0,0,"730750505048",,terminal_output
+170,708423,"TERMINAL",0,0,"829222650",,terminal_output
+171,709366,"TERMINAL",0,0,"4031033371",,terminal_output
+172,710421,"TERMINAL",0,0,"14144482",,terminal_output
+173,711573,"TERMINAL",0,0,"25255593",,terminal_output
+174,712376,"TERMINAL",0,0,"363666504",,terminal_output
+175,713357,"TERMINAL",0,0,"47477715",,terminal_output
+176,714455,"TERMINAL",0,0,"58588826",,terminal_output
+177,715393,"TERMINAL",0,0,"69699937",,terminal_output
+178,716488,"TERMINAL",0,0,"74071:001:008:0048",,terminal_output
+179,717629,"TERMINAL",0,0,"81811159",,terminal_output
+180,718646,"TERMINAL",0,0,"92922261:00",,terminal_output
+181,719715,"TERMINAL",0,0,"5032033371",,terminal_output
+182,720684,"TERMINAL",0,0,"14144482",,terminal_output
+183,721698,"TERMINAL",0,0,"25255593",,terminal_output
+184,723099,"TERMINAL",0,0,"3636664:004",,terminal_output
+185,723893,"TERMINAL",0,0,"47477715",,terminal_output
+186,724967,"TERMINAL",0,0,"58588826",,terminal_output
+187,726008,"TERMINAL",0,0,"69699937",,terminal_output
+188,727209,"TERMINAL",0,0,"750710101048",,terminal_output
+189,728171,"TERMINAL",0,0,"81811159",,terminal_output
+190,729299,"TERMINAL",0,0,"929222610",,terminal_output
+191,730130,"TERMINAL",0,0,"40:0033033371",,terminal_output
+192,731280,"TERMINAL",0,0,"14144482",,terminal_output
+193,732432,"TERMINAL",0,0,"25255593",,terminal_output
+194,733689,"TERMINAL",0,0,"363666104",,terminal_output
+195,734686,"TERMINAL",0,0,"47477715",,terminal_output
+196,735860,"TERMINAL",0,0,"58588826",,terminal_output
+197,736768,"TERMINAL",0,0,"69699937",,terminal_output
+198,737915,"TERMINAL",0,0,"76:00720202048",,terminal_output
+199,738779,"TERMINAL",0,0,"81811159",,terminal_output
+200,739643,"TERMINAL",0,0,"9340333721",,terminal_output
+201,740758,"TERMINAL",0,0,"114144482",,terminal_output
+202,741872,"TERMINAL",0,0,"25255593",,terminal_output
+203,742820,"TERMINAL",0,0,"363666204",,terminal_output
+204,743944,"TERMINAL",0,0,"47477715",,terminal_output
+205,744905,"TERMINAL",0,0,"58588826",,terminal_output
+206,746140,"TERMINAL",0,0,"69699937",,terminal_output
+207,746742,"TERMINAL",0,0,"710730303048",,terminal_output
+208,747510,"TERMINAL",0,0,"81811159",,terminal_output
+209,748521,"TERMINAL",0,0,"929222630",,terminal_output
+210,749496,"TERMINAL",0,0,"2035033371",,terminal_output
+211,750793,"TERMINAL",0,0,"14144482",,terminal_output
+212,751539,"TERMINAL",0,0,"25255593",,terminal_output
+213,752600,"TERMINAL",0,0,"363666304",,terminal_output
+214,753732,"TERMINAL",0,0,"47477715",,terminal_output
+215,754701,"TERMINAL",0,0,"58588826",,terminal_output
+216,755819,"TERMINAL",0,0,"69699937",,terminal_output
+217,756746,"TERMINAL",0,0,"720740404048",,terminal_output
+218,757855,"TERMINAL",0,0,"81811159",,terminal_output
+219,758865,"TERMINAL",0,0,"929222640",,terminal_output
+220,759856,"TERMINAL",0,0,"3031:0033371",,terminal_output
+221,760853,"TERMINAL",0,0,"14144482",,terminal_output
+222,762069,"TERMINAL",0,0,"25255593",,terminal_output
+223,763081,"TERMINAL",0,0,"363666404",,terminal_output
+224,764063,"TERMINAL",0,0,"47477715",,terminal_output
+225,764925,"TERMINAL",0,0,"58588826",,terminal_output
+226,766090,"TERMINAL",0,0,"69699937",,terminal_output

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-94fb4d7e-812c-4d36-984a-6626015fa6fd1750854950642-2025_06_25-14.36.16.983/source.csv

@@ -0,0 +1,42 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,4,"train_tokenizer.py",0,0,"from dataclasses import dataclass\nimport os\nimport time\n\nimport einops\nfrom flax.training import orbax_utils\nfrom flax.training.train_state import TrainState\nfrom jax.sharding import Mesh, PartitionSpec, NamedSharding\nfrom jax.experimental.mesh_utils import create_device_mesh\nimport optax\nimport orbax\nfrom orbax.checkpoint import PyTreeCheckpointer\nimport numpy as np\nimport dm_pix as pix\nimport jax\nimport jax.numpy as jnp\nimport tyro\nimport wandb\n\nfrom models.tokenizer import TokenizerVQVAE\nfrom utils.dataloader import get_dataloader\n\nts = int(time.time())\n\n\n@dataclass\nclass Args:\n    # Experiment\n    num_steps: int = 300_000\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_height: int = 90\n    image_width: int = 160\n    data_dir: str = ""data_tfrecords/coinrun""\n    checkpoint: str = """"\n    # Optimization\n    vq_beta: float = 0.25\n    batch_size: int = 48\n    min_lr: float = 3e-4\n    max_lr: float = 3e-4\n    warmup_steps: int = 10000\n    # Tokenizer\n    model_dim: int = 512\n    latent_dim: int = 32\n    num_latents: int = 1024\n    patch_size: int = 4\n    num_blocks: int = 8\n    num_heads: int = 8\n    dropout: float = 0.0\n    codebook_dropout: float = 0.01\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 10000\n    log_gradients: bool = False\n\n\nargs = tyro.cli(Args)\n\n\ndef tokenizer_loss_fn(params, state, inputs):\n    # --- Compute loss ---\n    outputs = state.apply_fn(\n        params, inputs, training=True, rngs={""dropout"": inputs[""rng""]}\n    )\n    mse = jnp.square(inputs[""videos""] - outputs[""recon""]).mean()\n    q_loss = jnp.square(jax.lax.stop_gradient(outputs[""emb""]) - outputs[""z""]).mean()\n    commitment_loss = jnp.square(\n        outputs[""emb""] - jax.lax.stop_gradient(outputs[""z""])\n    ).mean()\n    loss = mse + q_loss + args.vq_beta * commitment_loss\n\n    # --- Compute validation metrics ---\n    gt = inputs[""videos""].clip(0, 1).reshape(-1, *inputs[""videos""].shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = pix.psnr(gt, recon).mean()\n    ssim = pix.ssim(gt, recon).mean()\n    _, index_counts = jnp.unique_counts(\n        jnp.ravel(outputs[""indices""]), size=args.num_latents, fill_value=0\n    )\n    codebook_usage = (index_counts != 0).mean()\n    metrics = dict(\n        loss=loss,\n        mse=mse,\n        q_loss=q_loss,\n        commitment_loss=commitment_loss,\n        psnr=psnr,\n        ssim=ssim,\n        codebook_usage=codebook_usage,\n    )\n    return loss, (outputs[""recon""], metrics)\n\n\n@jax.jit\ndef train_step(state, inputs):\n    grad_fn = jax.value_and_grad(tokenizer_loss_fn, has_aux=True, allow_int=True)\n    (loss, (recon, metrics)), grads = grad_fn(state.params, state, inputs)\n    state = state.apply_gradients(grads=grads)\n    if args.log_gradients:\n        metrics[""encoder_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""encoder""]\n        )\n        metrics[""vq_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""vq""]\n        )\n        metrics[""decoder_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""decoder""]\n        )\n    return state, loss, recon, metrics\n\n\nif __name__ == ""__main__"":\n    jax.distributed.initialize()\n    num_devices = jax.device_count()\n    if num_devices == 0:\n        raise ValueError(""No JAX devices found."")\n    print(f""Running on {num_devices} devices."")\n\n    if args.batch_size % num_devices != 0:\n        raise ValueError(\n            f""Global batch size {args.batch_size} must be divisible by ""\n            f""number of devices {num_devices}.""\n        )\n\n    per_device_batch_size_for_init = args.batch_size // num_devices\n\n    rng = jax.random.PRNGKey(args.seed)\n    if args.log and jax.process_index() == 0:\n        wandb.init(entity=args.entity, project=args.project, group=""debug"", config=args)\n\n    # --- Initialize model ---\n    tokenizer = TokenizerVQVAE(\n        in_dim=args.image_channels,\n        model_dim=args.model_dim,\n        latent_dim=args.latent_dim,\n        num_latents=args.num_latents,\n        patch_size=args.patch_size,\n        num_blocks=args.num_blocks,\n        num_heads=args.num_heads,\n        dropout=args.dropout,\n        codebook_dropout=args.codebook_dropout,\n    )\n    rng, _rng = jax.random.split(rng)\n    image_shape = (args.image_height, args.image_width, args.image_channels)\n    inputs = dict(\n        videos=jnp.zeros(\n            (per_device_batch_size_for_init, args.seq_len, *image_shape),\n            dtype=jnp.float32,\n        ),\n    )\n    init_params = tokenizer.init(_rng, inputs)\n\n    # --- Initialize optimizer ---\n    lr_schedule = optax.warmup_cosine_decay_schedule(\n        args.min_lr, args.max_lr, args.warmup_steps, args.num_steps\n    )\n    tx = optax.adamw(learning_rate=lr_schedule, b1=0.9, b2=0.9, weight_decay=1e-4)\n    train_state = TrainState.create(apply_fn=tokenizer.apply, params=init_params, tx=tx)\n\n    # FIXME: switch to create_hybrid_device_mesh for runs spanning multiple nodes\n    device_mesh_arr = create_device_mesh((num_devices,))\n    mesh = Mesh(devices=device_mesh_arr, axis_names=(""data"",))\n\n    replicated_sharding = NamedSharding(mesh, PartitionSpec())\n    train_state = jax.device_put(train_state, replicated_sharding)\n\n    # --- Load checkpoint ---\n    step = 0\n    if args.checkpoint:\n        restore_target = {""model"": train_state}\n        restore_args = orbax_utils.restore_args_from_target(restore_target)\n        train_state.params[""params""].update(\n            PyTreeCheckpointer()\n            .restore(args.checkpoint, item=restore_target, restore_args=restore_args)[\n                ""model""\n            ]\n            .params[""params""]\n        )\n        # Assume checkpoint is of the form tokenizer_<timestamp>_<step>\n        step += int(args.checkpoint.split(""_"")[-1])\n\n    # --- TRAIN LOOP ---\n    tfrecord_files = [\n        os.path.join(args.data_dir, x)\n        for x in os.listdir(args.data_dir)\n        if x.endswith("".tfrecord"")\n    ]\n    dataloader = get_dataloader(\n        # NOTE: We deliberately pass the global batch size\n        # The dataloader shards the dataset across all processes\n        tfrecord_files,\n        args.seq_len,\n        args.batch_size,\n        *image_shape,\n    )\n    print(f""Starting training from step {step}..."")\n    while step < args.num_steps:\n        # for videos in dataloader:\n        # npy_path = ""overfit_dir/single_sample_corner.npy""\n        npy_path = ""overfit_dir/single_batch_12_elems.npy""\n        videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):\n            # --- Train step ---\n            rng, _rng = jax.random.split(rng)\n\n            videos_sharding = NamedSharding(\n                mesh, PartitionSpec(""data"", None, None, None, None)\n            )\n            videos = jax.make_array_from_process_local_data(videos_sharding, videos)\n\n            inputs = dict(videos=videos, rng=_rng)\n            train_state, loss, recon, metrics = train_step(train_state, inputs)\n            print(f""Step {step}, loss: {loss}"")\n            step += 1\n\n            # --- Logging ---\n            if args.log and jax.process_index() == 0:\n                if step % args.log_interval == 0:\n                    wandb.log({""loss"": loss, ""step"": step, **metrics})\n                if step % args.log_image_interval == 0:\n                    gt_seq = inputs[""videos""][0]\n                    recon_seq = recon[0].clip(0, 1)\n                    comparison_seq = jnp.concatenate((gt_seq, recon_seq), axis=1)\n                    comparison_seq = einops.rearrange(\n                        comparison_seq * 255, ""t h w c -> h (t w) c""\n                    )\n                    log_images = dict(\n                        image=wandb.Image(np.asarray(gt_seq[0])),\n                        recon=wandb.Image(np.asarray(recon_seq[0])),\n                        true_vs_recon=wandb.Image(\n                            np.asarray(comparison_seq.astype(np.uint8))\n                        ),\n                    )\n                    wandb.log(log_images)\n            if step % args.log_checkpoint_interval == 0:\n                ckpt = {""model"": train_state}\n                orbax_checkpointer = orbax.checkpoint.PyTreeCheckpointer()\n                save_args = orbax_utils.save_args_from_target(ckpt)\n                orbax_checkpointer.save(\n                    os.path.join(os.getcwd(), args.ckpt_dir, f""tokenizer_{ts}_{step}""),\n                    ckpt,\n                    save_args=save_args,\n                )\n            if step >= args.num_steps:\n                break\n",python,tab
+2,1687,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"2:36:16 PM [info] Activating crowd-code\n2:36:16 PM [info] Recording started\n2:36:16 PM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,1689,"train_tokenizer.py",0,0,"",python,tab
+4,1690,"train_tokenizer.py",4056,0,"",python,selection_mouse
+5,5861,"TERMINAL",0,0,"/bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt",,terminal_command
+6,10311,"train_tokenizer.py",4053,0,"",python,selection_mouse
+7,24016,"train_tokenizer.py",4056,0,"",python,selection_mouse
+8,24380,"models/tokenizer.py",0,0,"from typing import Dict, Any, Tuple\n\nimport flax.linen as nn\n\nfrom utils.preprocess import patchify, unpatchify\nfrom utils.nn import STTransformer, VectorQuantizer\n\n\nclass TokenizerVQVAE(nn.Module):\n    """"""ST-ViVit VQ-VAE""""""\n\n    in_dim: int\n    model_dim: int\n    latent_dim: int\n    num_latents: int\n    patch_size: int\n    num_blocks: int\n    num_heads: int\n    dropout: float\n    codebook_dropout: float\n\n    def setup(self):\n        self.encoder = STTransformer(\n            self.model_dim,\n            self.latent_dim,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n        )\n        self.vq = VectorQuantizer(\n            self.latent_dim,\n            self.num_latents,\n            self.codebook_dropout,\n        )\n        self.out_dim = self.in_dim * self.patch_size**2\n        self.decoder = STTransformer(\n            self.model_dim,\n            self.out_dim,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n        )\n\n    def __call__(self, batch: Dict[str, Any], training: bool = True) -> Dict[str, Any]:\n        H, W = batch[""videos""].shape[2:4]\n        outputs = self.vq_encode(batch[""videos""], training)\n        recon = self.decoder(outputs[""z_q""])  # (B, T, H_down * W_down, C)\n        recon = nn.sigmoid(recon)\n        outputs[""recon""] = unpatchify(recon, self.patch_size, H, W)\n        return outputs\n\n    def vq_encode(self, videos: Any, training: bool = True) -> Dict[str, Any]:\n        # --- Preprocess + encode ---\n        B, T = videos.shape[:2]\n        x = patchify(videos, self.patch_size)\n        N = x.shape[2]\n        x = self.encoder(x)  # (B, T, N, E)\n\n        # --- Vector quantize ---\n        x = x.reshape(B * T * N, self.latent_dim)\n        z_q, z, emb, indices = self.vq(x, training)\n        z_q = z_q.reshape(B, T, N, self.latent_dim)\n        indices = indices.reshape(B, T, N)\n        return dict(z_q=z_q, z=z, emb=emb, indices=indices)\n\n    def decode(self, indices: Any, video_hw: Tuple[int, int]):\n        z = self.vq.codebook[indices]\n        recon = self.decoder(z)\n        recon = nn.sigmoid(recon)\n        return unpatchify(recon, self.patch_size, *video_hw)\n",python,tab
+9,27086,"models/tokenizer.py",388,0,"",python,selection_mouse
+10,27247,"models/tokenizer.py",384,16,"codebook_dropout",python,selection_mouse
+11,28185,"models/tokenizer.py",332,0,"",python,selection_mouse
+12,28723,"models/tokenizer.py",370,0,"",python,selection_mouse
+13,29516,"models/tokenizer.py",312,0,"",python,selection_mouse
+14,29672,"models/tokenizer.py",306,10,"patch_size",python,selection_mouse
+15,46767,"models/tokenizer.py",1329,0,"",python,selection_mouse
+16,47134,"utils/preprocess.py",0,0,"import dm_pix as pix\nimport einops\nimport jax\nimport jax.numpy as jnp\n\n\ndef patchify(videos: jax.Array, size: int) -> jax.Array:\n    B, T, H, W, C = videos.shape\n    x = jnp.pad(videos, ((0, 0), (0, 0), (0, -H % size), (0, -W % size), (0, 0)))\n    return einops.rearrange(\n        x, ""b t (hn hp) (wn wp) c -> b t (hn wn) (hp wp c)"", hp=size, wp=size\n    )\n\n\ndef unpatchify(patches: jax.Array, size: int, h_out: int, w_out: int) -> jax.Array:\n    h_pad = -h_out % size\n    hn = (h_out + h_pad) // size\n    x = einops.rearrange(\n        patches,\n        ""b t (hn wn) (hp wp c) -> b t (hn hp) (wn wp) c"",\n        hp=size,\n        wp=size,\n        hn=hn,\n    )\n    return x[:, :, :h_out, :w_out]\n",python,tab
+17,162466,"utils/preprocess.py",527,0,"",python,selection_mouse
+18,162481,"utils/preprocess.py",526,0,"",python,selection_command
+19,163693,"utils/preprocess.py",527,0,"",python,selection_mouse
+20,163700,"utils/preprocess.py",526,0,"",python,selection_command
+21,164243,"utils/preprocess.py",544,0,"",python,selection_mouse
+22,164257,"utils/preprocess.py",543,0,"",python,selection_command
+23,165045,"utils/preprocess.py",579,0,"",python,selection_mouse
+24,165785,"utils/preprocess.py",619,0,"",python,selection_mouse
+25,165797,"utils/preprocess.py",618,0,"",python,selection_command
+26,166268,"utils/preprocess.py",544,0,"",python,selection_mouse
+27,166273,"utils/preprocess.py",543,0,"",python,selection_command
+28,166801,"utils/preprocess.py",602,0,"",python,selection_mouse
+29,166805,"utils/preprocess.py",601,0,"",python,selection_command
+30,166969,"utils/preprocess.py",602,0,"",python,selection_mouse
+31,166976,"utils/preprocess.py",601,0,"",python,selection_command
+32,167114,"utils/preprocess.py",545,58,"        ""b t (hn wn) (hp wp c) -> b t (hn hp) (wn wp) c"",\n",python,selection_mouse
+33,167121,"utils/preprocess.py",546,57,"       ""b t (hn wn) (hp wp c) -> b t (hn hp) (wn wp) c"",\n",python,selection_command
+34,167809,"utils/preprocess.py",602,0,"",python,selection_mouse
+35,167828,"utils/preprocess.py",601,0,"",python,selection_command
+36,400812,"train_tokenizer.py",0,0,"",python,tab
+37,413351,"train_tokenizer.py",1116,0,"",python,selection_mouse
+38,2135599,"train_tokenizer.py",0,0,"Switched from branch 'mihir-tmp' to 'main'",python,git_branch_checkout
+39,2140593,"train_tokenizer.py",0,0,"Switched from branch 'main' to 'tmp'",python,git_branch_checkout
+40,11451341,"train_tokenizer.py",0,0,"Switched from branch 'tmp' to 'main'",python,git_branch_checkout
+41,11466426,"train_tokenizer.py",0,0,"Switched from branch 'main' to 'add-wandb-name-and-tags'",python,git_branch_checkout

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-d33d9128-8aa8-4382-a7f1-61cc99198a8e1750839147762-2025_06_25-10.21.30.519/source.csv

@@ -0,0 +1,178 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,2413,"TERMINAL",0,0,"queue",,terminal_command
+3,2443,"TERMINAL",0,0,"[?25l[?2004l\r]633;E;watch -n1 squeue --me;895d5730-3b47-4a5d-840c-5d137f58d793]633;C[?25h",,terminal_output
+4,8529,"TERMINAL",0,0,"squeue --me",,terminal_command
+5,8573,"TERMINAL",0,0,"[?25l[?2004l\r]633;E;squeue --me;895d5730-3b47-4a5d-840c-5d137f58d793]633;C[?25h",,terminal_output
+6,17811,"TERMINAL",0,0,"time squeue --me",,terminal_command
+7,17844,"TERMINAL",0,0,"[?25l[?2004l\r]633;E;squeue --me;895d5730-3b47-4a5d-840c-5d137f58d793]633;C[?25h",,terminal_output
+8,37681,"utils/dataloader_new.py",0,0,"import functools\nimport jax\n\nimport tensorflow as tf\n\n# reserve GPU memory for JAX only if tensorflow is built with GPU support\ntry:\n  tf.config.experimental.set_visible_devices([], ""GPU"")\nexcept tf.errors.NotFoundError:\n  pass\n\n\n# --- TensorFlow function for processing: slicing, normalization ---\ndef _tf_process_episode(episode_tensor, seq_len, image_h, image_w, image_c):\n    """"""\n    Processes a raw episode tensor in TensorFlow.\n    Takes a full episode, extracts a random sequence, and normalizes it.\n    Args:\n        episode_tensor: A TensorFlow tensor representing a full video episode.\n                        Expected shape: (dynamic_length, image_h, image_w, image_c)\n                        Expected dtype: e.g., tf.uint8 (raw pixel values)\n        seq_len: The desired length of the sub-sequence to extract.\n        image_h: The height of each frame.\n        image_w: The width of each frame.\n        image_c: The number of channels in each frame.\n    Returns:\n        A TensorFlow tensor representing the processed video sequence.\n        Shape: (seq_len, image_h, image_w, image_c)\n        Dtype: tf.float32 (normalized pixel values)\n    """"""\n    current_episode_len = tf.shape(episode_tensor)[0]\n\n    max_start_idx = current_episode_len - seq_len\n\n    start_idx = tf.random.uniform(\n        shape=(), minval=0, maxval=max_start_idx + 1, dtype=tf.int32\n    )\n\n    seq = episode_tensor[start_idx : start_idx + seq_len]\n\n    seq = tf.cast(seq, tf.float32) / 255.0\n\n    # Ensure the final shape is statically known for batching.\n    # tf.reshape is robust, but tf.ensure_shape or set_shape can also be used if confident.\n    processed_sequence = tf.reshape(seq, [seq_len, image_h, image_w, image_c])\n\n    return processed_sequence\n\n\ndef _parse_tfrecord_fn(example_proto, image_h, image_w, image_c):\n    feature_description = {\n        ""height"": tf.io.FixedLenFeature([], tf.int64),\n        ""width"": tf.io.FixedLenFeature([], tf.int64),\n        ""channels"": tf.io.FixedLenFeature([], tf.int64),\n        ""sequence_length"": tf.io.FixedLenFeature([], tf.int64),\n        ""raw_video"": tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example_proto, feature_description)\n\n    video_shape = (example[""sequence_length""], image_h, image_w, image_c)\n\n    episode_tensor = tf.io.decode_raw(example[""raw_video""], out_type=tf.uint8)\n    episode_tensor = tf.reshape(episode_tensor, video_shape)\n\n    episode_tensor = tf.ensure_shape(episode_tensor, [None, image_h, image_w, image_c])\n    return episode_tensor\n\n\ndef get_dataloader(\n    tfrecord_paths: list[str],  # List of TFRecord file paths\n    seq_len: int,\n    global_batch_size: int,\n    image_h: int,\n    image_w: int,\n    image_c: int,\n    shuffle_buffer_size: int = 10,\n    num_parallel_calls: int = tf.data.AUTOTUNE,\n    seed: int = 42,\n):\n    """"""\n    Creates a tf.data.Dataset pipeline from TFRecord files.\n    """"""\n    if not tfrecord_paths:\n        raise ValueError(""tfrecord_paths list cannot be empty."")\n\n    process_id = jax.process_index()\n    num_processes = jax.process_count()\n\n    assert (\n        global_batch_size % num_processes == 0\n    ), ""Global batch size {global_batch_size} \\n        must be divisible by the number of JAX processes {num_processes} for proper sharding.""\n    per_process_batch_size = global_batch_size // num_processes\n\n    # Create a dataset of just the paths (filenames)\n    path_dataset = tf.data.Dataset.from_tensor_slices(tfrecord_paths)\n    breakpoint()\n\n    dataset = tf.data.TFRecordDataset(\n        tfrecord_paths, num_parallel_reads=tf.data.AUTOTUNE\n    )\n\n    dataset = dataset.shard(num_shards=num_processes, index=process_id)\n\n    # (f.srambical) NOTE: For TFRecords, it's often good to have a large shuffle buffer.\n    if shuffle_buffer_size > 0:\n        dataset = dataset.shuffle(\n            buffer_size=shuffle_buffer_size, seed=seed, reshuffle_each_iteration=True\n        )\n    parse_fn = functools.partial(\n        _parse_tfrecord_fn, image_h=image_h, image_w=image_w, image_c=image_c\n    )\n    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    tf_process_fn = functools.partial(\n        _tf_process_episode,\n        seq_len=seq_len,\n        image_h=image_h,\n        image_w=image_w,\n        image_c=image_c,\n    )\n    dataset = dataset.map(tf_process_fn, num_parallel_calls=num_parallel_calls)\n\n    dataset = dataset.repeat(None)\n    dataset = dataset.batch(per_process_batch_size, drop_remainder=True)\n    dataset = dataset.prefetch(tf.data.AUTOTUNE)\n\n    return dataset.as_numpy_iterator()\n",python,tab
+9,37682,"utils/dataloader_new.py",3487,0,"",python,selection_mouse
+10,38347,"utils/dataloader_new.py",3657,0,"",python,selection_mouse
+11,39063,"utils/dataloader_new.py",3526,0,"",python,selection_mouse
+12,39065,"utils/dataloader_new.py",3525,0,"",python,selection_command
+13,39637,"utils/dataloader_new.py",3487,0,"",python,selection_mouse
+14,99817,"utils/dataloader_new.py",4112,0,"",python,selection_mouse
+15,100375,"utils/dataloader_new.py",3544,0,"",python,selection_mouse
+16,100858,"utils/dataloader_new.py",3487,0,"",python,selection_mouse
+17,149529,"utils/dataloader_new.py",4205,0,"",python,selection_mouse
+18,149531,"utils/dataloader_new.py",4204,0,"",python,selection_command
+19,150257,"utils/dataloader.py",0,0,"import functools\nimport jax\n\nimport tensorflow as tf\n\n# reserve GPU memory for JAX only if tensorflow is built with GPU support\ntry:\n  tf.config.experimental.set_visible_devices([], ""GPU"")\nexcept tf.errors.NotFoundError:\n  pass\n\n\n# --- TensorFlow function for processing: slicing, normalization ---\ndef _tf_process_episode(episode_tensor, seq_len, image_h, image_w, image_c):\n    """"""\n    Processes a raw episode tensor in TensorFlow.\n    Takes a full episode, extracts a random sequence, and normalizes it.\n    Args:\n        episode_tensor: A TensorFlow tensor representing a full video episode.\n                        Expected shape: (dynamic_length, image_h, image_w, image_c)\n                        Expected dtype: e.g., tf.uint8 (raw pixel values)\n        seq_len: The desired length of the sub-sequence to extract.\n        image_h: The height of each frame.\n        image_w: The width of each frame.\n        image_c: The number of channels in each frame.\n    Returns:\n        A TensorFlow tensor representing the processed video sequence.\n        Shape: (seq_len, image_h, image_w, image_c)\n        Dtype: tf.float32 (normalized pixel values)\n    """"""\n    current_episode_len = tf.shape(episode_tensor)[0]\n\n    max_start_idx = current_episode_len - seq_len\n\n    start_idx = tf.random.uniform(\n        shape=(), minval=0, maxval=max_start_idx + 1, dtype=tf.int32\n    )\n\n    seq = episode_tensor[start_idx : start_idx + seq_len]\n\n    seq = tf.cast(seq, tf.float32) / 255.0\n\n    # Ensure the final shape is statically known for batching.\n    # tf.reshape is robust, but tf.ensure_shape or set_shape can also be used if confident.\n    processed_sequence = tf.reshape(seq, [seq_len, image_h, image_w, image_c])\n\n    return processed_sequence\n\n\ndef _parse_tfrecord_fn(example_proto, image_h, image_w, image_c):\n    feature_description = {\n        ""height"": tf.io.FixedLenFeature([], tf.int64),\n        ""width"": tf.io.FixedLenFeature([], tf.int64),\n        ""channels"": tf.io.FixedLenFeature([], tf.int64),\n        ""sequence_length"": tf.io.FixedLenFeature([], tf.int64),\n        ""raw_video"": tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example_proto, feature_description)\n\n    video_shape = (example[""sequence_length""], image_h, image_w, image_c)\n\n    episode_tensor = tf.io.decode_raw(example[""raw_video""], out_type=tf.uint8)\n    episode_tensor = tf.reshape(episode_tensor, video_shape)\n\n    episode_tensor = tf.ensure_shape(episode_tensor, [None, image_h, image_w, image_c])\n    return episode_tensor\n\n\ndef get_dataloader(\n    tfrecord_paths: list[str],  # List of TFRecord file paths\n    seq_len: int,\n    global_batch_size: int,\n    image_h: int,\n    image_w: int,\n    image_c: int,\n    shuffle_buffer_size: int = 10,\n    num_parallel_calls: int = tf.data.AUTOTUNE,\n    seed: int = 42,\n):\n    """"""\n    Creates a tf.data.Dataset pipeline from TFRecord files.\n    """"""\n    if not tfrecord_paths:\n        raise ValueError(""tfrecord_paths list cannot be empty."")\n\n    process_id = jax.process_index()\n    num_processes = jax.process_count()\n\n    assert (\n        global_batch_size % num_processes == 0\n    ), ""Global batch size {global_batch_size} \\n        must be divisible by the number of JAX processes {num_processes} for proper sharding.""\n    per_process_batch_size = global_batch_size // num_processes\n\n    dataset = tf.data.TFRecordDataset(\n        tfrecord_paths, num_parallel_reads=tf.data.AUTOTUNE\n    )\n\n    dataset = dataset.shard(num_shards=num_processes, index=process_id)\n\n    # (f.srambical) NOTE: For TFRecords, it's often good to have a large shuffle buffer.\n    if shuffle_buffer_size > 0:\n        dataset = dataset.shuffle(\n            buffer_size=shuffle_buffer_size, seed=seed, reshuffle_each_iteration=True\n        )\n    parse_fn = functools.partial(\n        _parse_tfrecord_fn, image_h=image_h, image_w=image_w, image_c=image_c\n    )\n    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    tf_process_fn = functools.partial(\n        _tf_process_episode,\n        seq_len=seq_len,\n        image_h=image_h,\n        image_w=image_w,\n        image_c=image_c,\n    )\n    dataset = dataset.map(tf_process_fn, num_parallel_calls=num_parallel_calls)\n\n    dataset = dataset.repeat(None)\n    dataset = dataset.batch(per_process_batch_size, drop_remainder=True)\n    dataset = dataset.prefetch(tf.data.AUTOTUNE)\n\n    return dataset.as_numpy_iterator()\n",python,tab
+20,153641,"TERMINAL",0,0,"time squeue --me",,terminal_command
+21,153696,"TERMINAL",0,0,"[?25l[?2004l\r]633;E;squeue --me;895d5730-3b47-4a5d-840c-5d137f58d793]633;C[?25h",,terminal_output
+22,220337,"TERMINAL",0,0,"time squeue --me",,terminal_command
+23,220388,"TERMINAL",0,0,"[?25l[?2004l\r]633;E;squeue --me;895d5730-3b47-4a5d-840c-5d137f58d793]633;C[?25h",,terminal_output
+24,275710,"TERMINAL",0,0,"sbatch scripts/train_tokenizer_overfit_sample.sbatch",,terminal_command
+25,275761,"TERMINAL",0,0,"[?25l\r]633;A(jafar) ]0;mahajanm@atcremers51: ~/Projects/jafarmahajanm@atcremers51:~/Projects/jafar$ ]633;Bsbatch scripts/train_tokenizer_overfit_sample.sbatch\r\n[?2004l\r]633;E;sbatch scripts/train_tokenizer_overfit_sample.sbatch;895d5730-3b47-4a5d-840c-5d137f58d793]633;C[?25h",,terminal_output
+26,412543,"utils/dataloader.py",0,0,"",python,tab
+27,413008,"utils/dataloader_new.py",0,0,"",python,tab
+28,413008,"utils/dataloader_new.py",3500,0,"",python,selection_mouse
+29,413622,"utils/dataloader_new.py",3487,0,"",python,selection_mouse
+30,416146,"utils/dataloader.py",0,0,"",python,tab
+31,416147,"utils/dataloader.py",2110,0,"",python,selection_mouse
+32,425788,".gitignore",0,0,"*.pyc\n*.npy\n*.png\n*.gif\n\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt",ignore,tab
+33,425789,".gitignore",68,0,"",ignore,selection_command
+34,427457,".gitignore",73,0,"",ignore,selection_mouse
+35,427465,".gitignore",72,0,"",ignore,selection_command
+36,428954,".gitignore",73,0,"\n",ignore,content
+37,431180,".gitignore",74,0,"s",ignore,content
+38,431181,".gitignore",75,0,"",ignore,selection_keyboard
+39,431239,".gitignore",75,0,"l",ignore,content
+40,431240,".gitignore",76,0,"",ignore,selection_keyboard
+41,431493,".gitignore",76,0,"u",ignore,content
+42,431494,".gitignore",77,0,"",ignore,selection_keyboard
+43,431628,".gitignore",77,0,"r",ignore,content
+44,431629,".gitignore",78,0,"",ignore,selection_keyboard
+45,431813,".gitignore",78,0,"m",ignore,content
+46,431814,".gitignore",79,0,"",ignore,selection_keyboard
+47,432911,".gitignore",79,0,"*",ignore,content
+48,432913,".gitignore",80,0,"",ignore,selection_keyboard
+49,433224,".gitignore",80,0,".",ignore,content
+50,433226,".gitignore",81,0,"",ignore,selection_keyboard
+51,433442,".gitignore",81,0,"o",ignore,content
+52,433443,".gitignore",82,0,"",ignore,selection_keyboard
+53,433656,".gitignore",82,0,"u",ignore,content
+54,433657,".gitignore",83,0,"",ignore,selection_keyboard
+55,433856,".gitignore",83,0,"t",ignore,content
+56,433857,".gitignore",84,0,"",ignore,selection_keyboard
+57,433984,".gitignore",84,0,"\n",ignore,content
+58,438435,".gitignore",85,0,"d",ignore,content
+59,438437,".gitignore",86,0,"",ignore,selection_keyboard
+60,438616,".gitignore",86,0,"a",ignore,content
+61,438617,".gitignore",87,0,"",ignore,selection_keyboard
+62,438716,".gitignore",87,0,"t",ignore,content
+63,438718,".gitignore",88,0,"",ignore,selection_keyboard
+64,438807,".gitignore",88,0,"a",ignore,content
+65,438808,".gitignore",89,0,"",ignore,selection_keyboard
+66,439038,".gitignore",89,0,"\n",ignore,content
+67,439447,".gitignore",90,0,"d",ignore,content
+68,439448,".gitignore",91,0,"",ignore,selection_keyboard
+69,439656,".gitignore",91,0,"a",ignore,content
+70,439657,".gitignore",92,0,"",ignore,selection_keyboard
+71,439748,".gitignore",92,0,"t",ignore,content
+72,439749,".gitignore",93,0,"",ignore,selection_keyboard
+73,439893,".gitignore",93,0,"a",ignore,content
+74,439893,".gitignore",94,0,"",ignore,selection_keyboard
+75,440360,".gitignore",94,0,"_",ignore,content
+76,440361,".gitignore",95,0,"",ignore,selection_keyboard
+77,441127,".gitignore",95,0,"t",ignore,content
+78,441128,".gitignore",96,0,"",ignore,selection_keyboard
+79,441589,".gitignore",96,0,"f",ignore,content
+80,441589,".gitignore",97,0,"",ignore,selection_keyboard
+81,442445,".gitignore",97,0,"r",ignore,content
+82,442446,".gitignore",98,0,"",ignore,selection_keyboard
+83,442647,".gitignore",98,0,"e",ignore,content
+84,442647,".gitignore",99,0,"",ignore,selection_keyboard
+85,442812,".gitignore",99,0,"c",ignore,content
+86,442813,".gitignore",100,0,"",ignore,selection_keyboard
+87,442939,".gitignore",100,0,"o",ignore,content
+88,442940,".gitignore",101,0,"",ignore,selection_keyboard
+89,443066,".gitignore",101,0,"r",ignore,content
+90,443067,".gitignore",102,0,"",ignore,selection_keyboard
+91,443268,".gitignore",102,0,"d",ignore,content
+92,443269,".gitignore",103,0,"",ignore,selection_keyboard
+93,443390,".gitignore",103,0,"s",ignore,content
+94,443391,".gitignore",104,0,"",ignore,selection_keyboard
+95,445534,".gitignore",104,0,"\n",ignore,content
+96,448327,".gitignore",105,0,"l",ignore,content
+97,448329,".gitignore",106,0,"",ignore,selection_keyboard
+98,448514,".gitignore",106,0,"o",ignore,content
+99,448516,".gitignore",107,0,"",ignore,selection_keyboard
+100,449427,".gitignore",107,0,"g",ignore,content
+101,449429,".gitignore",108,0,"",ignore,selection_keyboard
+102,449520,".gitignore",108,0,"s",ignore,content
+103,449523,".gitignore",109,0,"",ignore,selection_keyboard
+104,450133,".gitignore",109,0,"\n",ignore,content
+105,459064,".gitignore",110,0,"s",ignore,content
+106,459066,".gitignore",111,0,"",ignore,selection_keyboard
+107,459282,".gitignore",111,0,"c",ignore,content
+108,459283,".gitignore",112,0,"",ignore,selection_keyboard
+109,459506,".gitignore",112,0,"r",ignore,content
+110,459507,".gitignore",113,0,"",ignore,selection_keyboard
+111,459696,".gitignore",113,0,"u",ignore,content
+112,459697,".gitignore",114,0,"",ignore,selection_keyboard
+113,460010,".gitignore",114,0,"t",ignore,content
+114,460011,".gitignore",115,0,"",ignore,selection_keyboard
+115,460236,".gitignore",114,1,"",ignore,content
+116,460364,".gitignore",113,1,"",ignore,content
+117,460529,".gitignore",113,0,"i",ignore,content
+118,460531,".gitignore",114,0,"",ignore,selection_keyboard
+119,460757,".gitignore",114,0,"p",ignore,content
+120,460758,".gitignore",115,0,"",ignore,selection_keyboard
+121,460904,".gitignore",115,0,"t",ignore,content
+122,460905,".gitignore",116,0,"",ignore,selection_keyboard
+123,461098,".gitignore",116,0,"s",ignore,content
+124,461099,".gitignore",117,0,"",ignore,selection_keyboard
+125,468232,"utils/dataloader.py",0,0,"",python,tab
+126,468284,"utils/dataloader.py",2726,0,"",python,selection_command
+127,472265,".gitignore",0,0,"",ignore,tab
+128,946611,"scripts/batch_sizes.md",0,0,"## Batchsizes for jafar with Minecraft dataset\n\n### Tokenizer\n| GPU VRAM | Batch Size | Learning Rate |\n|----------|------------|---------------|\n| -        | 1          | 4.3e-5        | \n| 12 GB    | 6          | 1e-4          | \n| 24 GB    | 12         | 1.5e-4        | ?? \n| 40 GB    | 12         |               |\n| 48 GB    | 24         |               | ?? why does this work on cremers\n| 80 GB    | 48         | 3e-4          |\n\n### LAM\n| GPU VRAM | Batch Size | Learning Rate |\n|----------|------------|---------------|\n| -        | 1          | 5e-6/5e-7     | \n| 12 GB    | 6          |               | \n| 24 GB    | 12         |               |\n| 48 GB    | 24         |               |\n| -        | 36         |  3e-5/3e-6    |\n| 80 GB    | 48         |               |\n",markdown,tab
+129,965032,"scripts/train_tokenizer_overfit_batch.sbatch",0,0,"#!/bin/bash\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=1\n#SBATCH --time=00:30:00\n#SBATCH --cpus-per-task=5\n#SBATCH --gres=gpu:1,VRAM:24G\n#SBATCH --mem=50G\n#SBATCH --output=logs/logs_training/%x_%j.log\n#SBATCH --error=logs/logs_training/%x_%j.log\n#SBATCH --job-name=train_tokenizer_minecraft_overfit_batch\n\n# Log the sbatch script\ncat $0\n\ntf_records_dir=""/storage/user/mahajanm/Projects/world-modeling/knoms_tfrecords_500/""\nws_dir='/storage/user/mahajanm/Projects/world-modeling'\njob_name=$SLURM_JOB_NAME\nslurm_job_id=$SLURM_JOB_ID\n\nCHECKPOINT_DIR=$ws_dir/checkpoints/$job_name_$slurm_job_id\nmkdir -p $CHECKPOINT_DIR\n\nenv | grep SLURM\n\nSLURM_STEP_NODELIST=$SLURM_NODELIST python train_tokenizer.py \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --batch_size=1 \\n    --min_lr=1e-4 \\n    --max_lr=1e-4 \\n    --log_image_interval=3 \\n    --log \\n    --entity instant-uv \\n    --project jafar \\n    --data_dir $tf_records_dir\n",shellscript,tab
+130,1334999,"scripts/batch_sizes.md",0,0,"",markdown,tab
+131,1767929,"TERMINAL",0,0,"queue",,terminal_command
+132,1767986,"TERMINAL",0,0,"\r\n[?2004l\r]633;E;watch -n1 squeue --me;895d5730-3b47-4a5d-840c-5d137f58d793]633;C",,terminal_output
+133,1805007,"TERMINAL",0,0,"[?1049h(B[?7h]4;8;rgb:54/54/54\]4;9;rgb:FF/54/54\]4;10;rgb:54/FF/54\]4;11;rgb:FF/FF/54\]4;12;rgb:54/54/FF\]4;13;rgb:FF/54/FF\]4;14;rgb:54/FF/FF\(BEvery 1.0s: squeue --meatcremers51: Wed Jun 25 10:50:58 2025slurm_load_jobs error: Slurm backup controller in standby mode(B",,terminal_output
+134,1829578,".gitignore",0,0,"",ignore,tab
+135,1830784,".gitignore",98,19,"ecords\nlogs\nscripts",ignore,selection_mouse
+136,1830831,".gitignore",73,44,"\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+137,1830832,".gitignore",50,67,"ndb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+138,1830833,".gitignore",25,92,"wandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+139,1830844,".gitignore",24,93,"\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+140,1830891,".gitignore",18,99,"*.gif\n\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+141,1830892,".gitignore",12,105,"*.png\n*.gif\n\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+142,1830896,".gitignore",6,111,"*.npy\n*.png\n*.gif\n\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+143,1830945,".gitignore",1,116,".pyc\n*.npy\n*.png\n*.gif\n\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+144,1830992,".gitignore",0,117,"*.pyc\n*.npy\n*.png\n*.gif\n\nwandb_key\ncheckpoints/\nwandb/\n__pycache__/\n*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+145,1832428,".gitignore",89,0,"",ignore,selection_mouse
+146,1833143,".gitignore",117,0,"",ignore,selection_mouse
+147,1833305,".gitignore",109,8,"\nscripts",ignore,selection_mouse
+148,1833330,".gitignore",94,23,"_tfrecords\nlogs\nscripts",ignore,selection_mouse
+149,1833380,".gitignore",88,29,"a\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+150,1833381,".gitignore",76,41,"urm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+151,1833431,".gitignore",75,42,"lurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+152,1833531,".gitignore",69,48,"ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+153,1833581,".gitignore",68,49,"*ckpt\nslurm*.out\ndata\ndata_tfrecords\nlogs\nscripts",ignore,selection_mouse
+154,2047870,"utils/dataloader_new.py",0,0,"",python,tab
+155,2047872,"utils/dataloader_new.py",3023,0,"",python,selection_mouse
+156,2053549,"utils/dataloader.py",0,0,"",python,tab
+157,2053567,"utils/dataloader.py",2726,0,"",python,selection_command
+158,2060763,"utils/dataloader.py",0,0,"",python,tab
+159,2062569,"utils/dataloader_new.py",0,0,"",python,tab
+160,2065700,"train_tokenizer.py",0,0,"from dataclasses import dataclass\nimport os\nimport time\n\nimport einops\nfrom flax.training import orbax_utils\nfrom flax.training.train_state import TrainState\nfrom jax.sharding import Mesh, PartitionSpec, NamedSharding\nfrom jax.experimental.mesh_utils import create_device_mesh\nimport optax\nimport orbax\nfrom orbax.checkpoint import PyTreeCheckpointer\nimport numpy as np\nimport dm_pix as pix\nimport jax\nimport jax.numpy as jnp\nimport tyro\nimport wandb\n\nfrom models.tokenizer import TokenizerVQVAE\nfrom utils.dataloader import get_dataloader\n\nts = int(time.time())\n\n\n@dataclass\nclass Args:\n    # Experiment\n    num_steps: int = 300_000\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_height: int = 90\n    image_width: int = 160\n    data_dir: str = ""data_tfrecords/coinrun""\n    checkpoint: str = """"\n    # Optimization\n    vq_beta: float = 0.25\n    batch_size: int = 48\n    min_lr: float = 3e-4\n    max_lr: float = 3e-4\n    warmup_steps: int = 10000\n    # Tokenizer\n    model_dim: int = 512\n    latent_dim: int = 32\n    num_latents: int = 1024\n    patch_size: int = 4\n    num_blocks: int = 8\n    num_heads: int = 8\n    dropout: float = 0.0\n    codebook_dropout: float = 0.01\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 10000\n    log_gradients: bool = False\n\n\nargs = tyro.cli(Args)\n\n\ndef tokenizer_loss_fn(params, state, inputs):\n    # --- Compute loss ---\n    outputs = state.apply_fn(\n        params,\n        inputs,\n        training=True,\n        rngs={""params"": inputs[""rng""], ""dropout"": inputs[""dropout_rng""]},\n    )\n    mse = jnp.square(inputs[""videos""] - outputs[""recon""]).mean()\n    q_loss = jnp.square(jax.lax.stop_gradient(outputs[""emb""]) - outputs[""z""]).mean()\n    commitment_loss = jnp.square(\n        outputs[""emb""] - jax.lax.stop_gradient(outputs[""z""])\n    ).mean()\n    loss = mse + q_loss + args.vq_beta * commitment_loss\n\n    # --- Compute validation metrics ---\n    gt = inputs[""videos""].clip(0, 1).reshape(-1, *inputs[""videos""].shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = pix.psnr(gt, recon).mean()\n    ssim = pix.ssim(gt, recon).mean()\n    _, index_counts = jnp.unique_counts(\n        jnp.ravel(outputs[""indices""]), size=args.num_latents, fill_value=0\n    )\n    codebook_usage = (index_counts != 0).mean()\n    metrics = dict(\n        loss=loss,\n        mse=mse,\n        q_loss=q_loss,\n        commitment_loss=commitment_loss,\n        psnr=psnr,\n        ssim=ssim,\n        codebook_usage=codebook_usage,\n    )\n    return loss, (outputs[""recon""], metrics)\n\n\n@jax.jit\ndef train_step(state, inputs):\n    grad_fn = jax.value_and_grad(tokenizer_loss_fn, has_aux=True, allow_int=True)\n    (loss, (recon, metrics)), grads = grad_fn(state.params, state, inputs)\n    state = state.apply_gradients(grads=grads)\n    if args.log_gradients:\n        metrics[""encoder_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""encoder""]\n        )\n        metrics[""vq_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""vq""]\n        )\n        metrics[""decoder_gradients_std/""] = jax.tree.map(\n            lambda x: x.std(), grads[""params""][""decoder""]\n        )\n    return state, loss, recon, metrics\n\n\nif __name__ == ""__main__"":\n    jax.distributed.initialize()\n    num_devices = jax.device_count()\n    if num_devices == 0:\n        raise ValueError(""No JAX devices found."")\n    print(f""Running on {num_devices} devices."")\n\n    if args.batch_size % num_devices != 0:\n        raise ValueError(\n            f""Global batch size {args.batch_size} must be divisible by ""\n            f""number of devices {num_devices}.""\n        )\n\n    per_device_batch_size_for_init = args.batch_size // num_devices\n\n    rng = jax.random.PRNGKey(args.seed)\n    if args.log and jax.process_index() == 0:\n        wandb.init(entity=args.entity, project=args.project, group=""debug"", config=args)\n\n    # --- Initialize model ---\n    tokenizer = TokenizerVQVAE(\n        in_dim=args.image_channels,\n        model_dim=args.model_dim,\n        latent_dim=args.latent_dim,\n        num_latents=args.num_latents,\n        patch_size=args.patch_size,\n        num_blocks=args.num_blocks,\n        num_heads=args.num_heads,\n        dropout=args.dropout,\n        codebook_dropout=args.codebook_dropout,\n    )\n    rng, _rng = jax.random.split(rng)\n    image_shape = (args.image_height, args.image_width, args.image_channels)\n    inputs = dict(\n        videos=jnp.zeros(\n            (per_device_batch_size_for_init, args.seq_len, *image_shape),\n            dtype=jnp.float32,\n        ),\n    )\n    init_params = tokenizer.init(_rng, inputs)\n\n    # --- Initialize optimizer ---\n    lr_schedule = optax.warmup_cosine_decay_schedule(\n        args.min_lr, args.max_lr, args.warmup_steps, args.num_steps\n    )\n    tx = optax.adamw(learning_rate=lr_schedule, b1=0.9, b2=0.9, weight_decay=1e-4)\n    train_state = TrainState.create(apply_fn=tokenizer.apply, params=init_params, tx=tx)\n\n    # FIXME: switch to create_hybrid_device_mesh for runs spanning multiple nodes\n    device_mesh_arr = create_device_mesh((num_devices,))\n    mesh = Mesh(devices=device_mesh_arr, axis_names=(""data"",))\n\n    replicated_sharding = NamedSharding(mesh, PartitionSpec())\n    train_state = jax.device_put(train_state, replicated_sharding)\n\n    # --- Load checkpoint ---\n    step = 0\n    if args.checkpoint:\n        restore_target = {""model"": train_state}\n        restore_args = orbax_utils.restore_args_from_target(restore_target)\n        train_state.params[""params""].update(\n            PyTreeCheckpointer()\n            .restore(args.checkpoint, item=restore_target, restore_args=restore_args)[\n                ""model""\n            ]\n            .params[""params""]\n        )\n        # Assume checkpoint is of the form tokenizer_<timestamp>_<step>\n        step += int(args.checkpoint.split(""_"")[-1])\n\n    # --- TRAIN LOOP ---\n    tfrecord_files = [\n        os.path.join(args.data_dir, x)\n        for x in os.listdir(args.data_dir)\n        if x.endswith("".tfrecord"")\n    ]\n    dataloader = get_dataloader(\n        # NOTE: We deliberately pass the global batch size\n        # The dataloader shards the dataset across all processes\n        tfrecord_files,\n        args.seq_len,\n        args.batch_size,\n        *image_shape,\n    )\n    print(f""Starting training from step {step}..."")\n    while step < args.num_steps:\n        # for videos in dataloader:\n        npy_path = ""overfit_dir/single_sample_corner.npy""\n        videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):\n            # --- Train step ---\n            rng, _rng, _rng_dropout = jax.random.split(rng, 3)\n\n            videos_sharding = NamedSharding(\n                mesh, PartitionSpec(""data"", None, None, None, None)\n            )\n            videos = jax.make_array_from_process_local_data(videos_sharding, videos)\n\n            inputs = dict(videos=videos, rng=_rng, dropout_rng=_rng_dropout)\n            start_time = time.time()\n            train_state, loss, recon, metrics = train_step(train_state, inputs)\n            elapsed_time = (time.time() - start_time) * 1000\n            print(f""Step {step}, loss: {loss}, step time: {elapsed_time}ms"")\n            step += 1\n\n            # --- Logging ---\n            if args.log:\n                if step % args.log_interval == 0 and jax.process_index() == 0:\n                    wandb.log(\n                        {\n                            ""loss"": loss,\n                            ""step"": step,\n                            ""step_time_ms"": elapsed_time,\n                            **metrics,\n                        }\n                    )\n                if step % args.log_image_interval == 0:\n                    gt_seq = inputs[""videos""][0]\n                    recon_seq = recon[0].clip(0, 1)\n                    comparison_seq = jnp.concatenate((gt_seq, recon_seq), axis=1)\n                    comparison_seq = einops.rearrange(\n                        comparison_seq * 255, ""t h w c -> h (t w) c""\n                    )\n                    # NOTE: Process-dependent control flow deliberately happens\n                    # after indexing operation since it must not contain code\n                    # sections that lead to cross-accelerator communication.\n                    if jax.process_index() == 0:\n                        log_images = dict(\n                            image=wandb.Image(np.asarray(gt_seq[0])),\n                            recon=wandb.Image(np.asarray(recon_seq[0])),\n                            true_vs_recon=wandb.Image(\n                                np.asarray(comparison_seq.astype(np.uint8))\n                            ),\n                        )\n                        wandb.log(log_images)\n            if step % args.log_checkpoint_interval == 0:\n                ckpt = {""model"": train_state}\n                orbax_checkpointer = orbax.checkpoint.PyTreeCheckpointer()\n                save_args = orbax_utils.save_args_from_target(ckpt)\n                orbax_checkpointer.save(\n                    os.path.join(os.getcwd(), args.ckpt_dir, f""tokenizer_{ts}_{step}""),\n                    ckpt,\n                    save_args=save_args,\n                )\n            if step >= args.num_steps:\n                break\n",python,tab
+161,2065786,"train_tokenizer.py",6527,0,"",python,selection_command
+162,2067915,"utils/dataloader_new.py",0,0,"",python,tab
+163,2077118,"train_tokenizer.py",0,0,"",python,tab
+164,2091508,"train_tokenizer.py",6721,0,"",python,selection_mouse
+165,2091509,"train_tokenizer.py",6720,0,"",python,selection_command
+166,2091620,"train_tokenizer.py",6720,1,":",python,selection_mouse
+167,2091620,"train_tokenizer.py",6721,0,"",python,selection_command
+168,2091634,"train_tokenizer.py",6679,42,"ape: "", videos.shape)\n        while(True):",python,selection_mouse
+169,2091648,"train_tokenizer.py",6676,45," shape: "", videos.shape)\n        while(True):",python,selection_mouse
+170,2091698,"train_tokenizer.py",6673,48,"tch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+171,2091699,"train_tokenizer.py",6632,89,"eos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+172,2091699,"train_tokenizer.py",6629,92,"videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+173,2091748,"train_tokenizer.py",6625,96,"    videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+174,2091749,"train_tokenizer.py",6624,97,"     videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+175,2091749,"train_tokenizer.py",6564,157,"       npy_path = ""overfit_dir/single_sample_corner.npy""\n        videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+176,2091800,"train_tokenizer.py",6563,158,"        npy_path = ""overfit_dir/single_sample_corner.npy""\n        videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+177,2091939,"train_tokenizer.py",6527,194,"        # for videos in dataloader:\n        npy_path = ""overfit_dir/single_sample_corner.npy""\n        videos = np.load(npy_path)\n        print(""batch shape: "", videos.shape)\n        while(True):",python,selection_mouse
+178,2212633,"train_tokenizer.py",6818,0,"",python,selection_mouse