Add files using upload-large-folder tool

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-623c548f-e16f-46a4-9ee1-6577a82e63e51754054052755-2025_08_01-15.14.20.520/source.csv

@@ -0,0 +1,35 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,354,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"3:14:20 PM [info] Activating crowd-code\n3:14:20 PM [info] Recording started\n3:14:20 PM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,1073,"extension-output-pdoom-org.crowd-code-#1-crowd-code",150,0,"3:14:20 PM [info] Git repository found\n3:14:20 PM [info] Git provider initialized successfully\n3:14:20 PM [info] Initial git state: [object Object]\n",Log,content
+4,703122,"experiments/sample.sh",0,0,"source .venv/bin/activate\n\ndata_dir=""$PWD/data_arrayrecord/dummy""\nckpt_dir=""$PWD/checkpoints/causal_dynamics_openai_grain_tok_restore""\n\nexport PYTHONUNBUFFERED=1\nsrun ipython --pdb sample.py -- \\n    --dyna_type ""causal"" \\n    --batch_size 1 \\n    --seq_len 2 \\n    --start_frame 1 \\n    --checkpoint $ckpt_dir \\n    --data_dir $data_dir",shellscript,tab
+5,703816,"experiments/sample.sh",283,0,"",shellscript,selection_mouse
+6,703816,"experiments/sample.sh",282,0,"",shellscript,selection_command
+7,704348,"experiments/sample.sh",337,0,"",shellscript,selection_mouse
+8,704352,"experiments/sample.sh",336,0,"",shellscript,selection_command
+9,705452,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+10,710934,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+11,711936,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+12,720157,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+13,721337,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+14,722480,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+15,723451,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+16,724245,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+17,725680,"experiments/sample.sh",337,0,"",shellscript,selection_mouse
+18,725686,"experiments/sample.sh",336,0,"",shellscript,selection_command
+19,860776,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+20,861676,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+21,863663,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+22,864281,"experiments/sample.sh",337,0,"",shellscript,selection_mouse
+23,864282,"experiments/sample.sh",336,0,"",shellscript,selection_command
+24,865295,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+25,865896,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+26,867495,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+27,1826628,"experiments/sample.sh",337,0,"",shellscript,selection_mouse
+28,1826635,"experiments/sample.sh",336,0,"",shellscript,selection_command
+29,1827817,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+30,1828589,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+31,1830307,"experiments/sample.sh",135,0,"",shellscript,selection_mouse
+32,1831392,"experiments/sample.sh",337,0,"",shellscript,selection_mouse
+33,1831396,"experiments/sample.sh",336,0,"",shellscript,selection_command
+34,1832055,"experiments/sample.sh",26,0,"",shellscript,selection_mouse
+35,1832824,"experiments/sample.sh",135,0,"",shellscript,selection_mouse

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-88e23d98-00ad-4d5b-8d4d-1f239e211eb71763045757922-2025_11_13-15.56.09.849/source.csv

@@ -0,0 +1,7 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,3,"crowd-pilot/crowd-pilot/serialization_utils.py",0,0,"#!/usr/bin/env python3\n""""""\nCommon utilities for dataset serialization scripts.\n""""""\n\nfrom __future__ import annotations\n\nfrom dataclasses import dataclass\nfrom pathlib import Path\nfrom typing import List, Optional, Tuple, Dict\n\nimport difflib\nimport re\nimport pandas as pd\nfrom datasets import Dataset, load_dataset\n\n\n_ANSI_CSI_RE = re.compile(r""\x1b\[[0-9;?]*[ -/]*[@-~]"")\n_ANSI_OSC_TERMINATED_RE = re.compile(r""\x1b\][\s\S]*?(?:\x07|\x1b\\)"")\n_ANSI_OSC_LINE_FALLBACK_RE = re.compile(r""\x1b\][^\n]*$"")\n_BRACKETED_PASTE_ENABLE = ""\x1b[?2004h""\n_BRACKETED_PASTE_DISABLE = ""\x1b[?2004l""\n_OSC_633 = ""\x1b]633;""\n_OSC_0 = ""\x1b]0;""\n\n\n@dataclass\nclass SerializeConfig:\n    output_dir: str\n    shard_size: int\n    target_chars: int\n    overlap_chars: int\n    min_session_chars: int\n    max_docs: Optional[int]\n    long_pause_threshold_ms: int\n    csv_root: Optional[str]\n    val_ratio: float\n    arrayrecord_group_size: Optional[int] = None\n\n\ndef _clean_text(text: str) -> str:\n    # Normalize line endings and strip trailing spaces; preserve tabs/newlines.\n    return text.replace(""\r\n"", ""\n"").replace(""\r"", ""\n"").rstrip()\n\n\ndef _fenced_block(path: str, language: Optional[str], content: str) -> str:\n    lang = (language or """").lower()\n    return f""```{lang}\n{content}\n```\n""\n\n\ndef _apply_change(content: str, offset: int, length: int, new_text: str) -> str:\n    # Mirrors crowd_code_player.replay_file.apply_change\n    base = str(content)\n    text = str(new_text) if pd.notna(new_text) else """"\n    text = text.replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n    if offset > len(base):\n        base = base + ("" "" * (offset - len(base)))\n    return base[:offset] + text + base[offset + length:]\n\n\ndef _apply_backspaces(text: str) -> str:\n    out: List[str] = []\n    for ch in text:\n        if ch == ""\b"":  # \x08\n            if out:\n                out.pop()\n        else:\n            out.append(ch)\n    return """".join(out)\n\n\ndef _normalize_terminal_output(raw: str) -> str:\n    """"""\n    Normalize PTY/terminal output for training:\n      - Apply backspaces (\x08)\n      - Strip OSC (window title/shell integration) first, keeping BEL/ST terminators intact\n      - Resolve carriage returns (\r) by keeping the last rewrite per line\n      - Strip CSI (coloring etc.)\n      - Finally drop any remaining BEL (\x07)\n    """"""\n    if not raw:\n        return raw\n    s = _apply_backspaces(raw)\n    # Remove OSC sequences that are properly terminated (BEL or ST)\n    s = _ANSI_OSC_TERMINATED_RE.sub("""", s)\n    # Fallback: drop any unterminated OSC up to end-of-line only\n    s = ""\n"".join(_ANSI_OSC_LINE_FALLBACK_RE.sub("""", line) for line in s.split(""\n""))\n    # Resolve carriage returns per line:\n    # - If there are multiple rewrites, keep the last non-empty chunk\n    # - If it's CRLF (ending with '\r' before '\n'), keep the content before '\r'\n    resolved_lines: List[str] = []\n    for seg in s.split(""\n""):\n        parts = seg.split(""\r"")\n        chosen = """"\n        # pick last non-empty part if available; else last part\n        for p in reversed(parts):\n            if p != """":\n                chosen = p\n                break\n        if chosen == """" and parts:\n            chosen = parts[-1]\n        resolved_lines.append(chosen)\n    s = ""\n"".join(resolved_lines)\n    # Strip ANSI escape sequences\n    s = _ANSI_CSI_RE.sub("""", s)\n    # Remove any remaining BEL beeps\n    s = s.replace(""\x07"", """")\n    return s\n\n\ndef _line_numbered_output(content: str, start_line: Optional[int] = None, end_line: Optional[int] = None) -> str:\n    # TODO (f.srambical): check whether this corresponds **exactly** to the output of cat -n {file_path} | sed -n '{vstart},{vend}p'\n    lines = content.splitlines()\n    total = len(lines)\n    if total == 0:\n        return """"\n    s = 1 if start_line is None else max(1, min(start_line, total))\n    e = total if end_line is None else max(1, min(end_line, total))\n    if e < s:\n        # FIXME (f.srambical): If this does not happen, remove the condition\n        raise ValueError(""This should never happen!"")\n        e = s\n    buf: List[str] = []\n    for idx in range(s, e + 1):\n        buf.append(f""{idx:6}\t{lines[idx - 1]}"")\n    return ""\n"".join(buf)\n\n\ndef _compute_viewport(total_lines: int, center_line: int, radius: int) -> Tuple[int, int]:\n    if total_lines <= 0:\n        return (1, 0)\n    start = max(1, center_line - radius)\n    end = min(total_lines, center_line + radius)\n    if end < start:\n        # FIXME (f.srambical): If this does not happen, remove the condition\n        raise ValueError(""This should never happen!"")\n    return (start, end)\n\n\ndef _escape_single_quotes_for_sed(text: str) -> str:\n    # Close quote, add an escaped single quote, reopen quote: '""'""'\n    return text.replace(""'"", ""'\""'\""'"")\n\n\ndef _compute_changed_block_lines(before: str, after: str) -> Tuple[int, int, List[str]]:\n    """"""\n    Return 1-based start and end line numbers in 'before' that should be replaced,\n    and the replacement lines from 'after'.\n    For pure deletions, the replacement list may be empty.\n    """"""\n    before_lines = before.splitlines()\n    after_lines = after.splitlines()\n    sm = difflib.SequenceMatcher(a=before_lines, b=after_lines, autojunk=False)\n    opcodes = [op for op in sm.get_opcodes() if op[0] != ""equal""]\n    if not opcodes:\n        # FIXME (f.srambical): clean this up\n        raise ValueError(""No diff opcodes found for content change"")\n                # No visible change; choose a safe single-line replace at end of file\n        start_line = max(1, len(before_lines))\n        end_line = start_line\n        repl = after_lines[start_line - 1:start_line] if after_lines else [""""]\n        return (start_line, end_line, repl)\n\n    first = opcodes[0]\n    last = opcodes[-1]\n    # i1/i2 refer to 'before' indices, j1/j2 to 'after'\n    start_line = (first[1] + 1) if (first[1] + 1) > 0 else 1\n    end_line = last[2] # no increment since we go from 'exclusive' to 'inclusive' indexing\n    replacement_lines = after_lines[first[3]:last[4]]\n    return (start_line, end_line, replacement_lines)\n\n\ndef _session_to_transcript(\n    df: pd.DataFrame,\n    long_pause_threshold_ms: int,\n) -> str:\n\n    file_states: Dict[str, str] = {}\n    terminal_state: str = """"\n    per_file_event_counts: Dict[str, int] = {}\n    per_file_cursor_positions: Dict[str, Tuple[int, int]] = {}  # (offset, length) for each file\n    last_time_ms: Optional[int] = None\n\n    parts: List[str] = []\n\n    for i in range(len(df)):\n        row = df.iloc[i]\n        file_path: str = row[""File""]\n        event_time: int = row[""Time""]\n        language: Optional[str] = row[""Language""]\n\n        # Long pause detection\n        if last_time_ms is not None:\n            delta = event_time - last_time_ms\n            if delta > long_pause_threshold_ms:\n                # TODO (f.srambical): think about whether we want to emit this as an observation or not\n                parts.append(f""<obs long_pause ms=\""{delta}\"" />"")\n        last_time_ms = event_time\n\n        event_type = row[""Type""]\n\n        match event_type:\n            case ""tab"":\n                # File switch event\n                parts.append(f""<act focus file=\""{file_path}\"" />"")\n                \n                # If Text is present, this is the first time opening the file\n                # and the entire file content is captured\n                text = row[""Text""]\n                if pd.notna(text):\n                    file_content = str(text).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                    file_states[file_path] = file_content\n                    parts.append(f""// observation: file={file_path}"")\n                    parts.append(_fenced_block(file_path, language, _clean_text(file_content)))\n\n            case ""terminal_command"":\n                # Terminal command execution\n                command = row[""Text""]\n                command_str = str(command).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                parts.append(f""<act terminal_command />"")\n                parts.append(_fenced_block(file_path, ""bash"", _clean_text(command_str)))\n\n            case ""terminal_output"":\n                # Terminal output capture\n                output = row[""Text""]\n                output_str = str(output).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                parts.append(f""<obs terminal_output />"")\n                parts.append(_fenced_block(file_path, None, _clean_text(output_str)))\n\n            case ""terminal_focus"":\n                # Terminal focus event\n                parts.append(f""<act focus target=\""terminal\"" />"")\n\n            case ""git_branch_checkout"":\n                # Git branch checkout event\n                branch_info = row[""Text""]\n                branch_str = str(branch_info).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                parts.append(f""<act git_branch_checkout />"")\n                parts.append(f""// git: {_clean_text(branch_str)}"")\n\n            case ""selection_command"" | ""selection_mouse"" | ""selection_keyboard"":\n                # Handle cursor movement\n                offset = row[""RangeOffset""]\n                length = row[""RangeLength""]\n                old_cursor = per_file_cursor_positions.get(file_path, (0, 0))\n                new_cursor = (offset, length)\n                per_file_cursor_positions[file_path] = new_cursor\n                \n                # Emit cursor movement observation if position changed\n                if old_cursor != new_cursor:\n                    parts.append(f""<act cursor file=\""{file_path}\"" offset=\""{offset}\"" len=\""{length}\"" />"")\n\n            case ""content"":\n                # Handle file edit events\n                offset = row[""RangeOffset""]\n                length = row[""RangeLength""]\n                new_text = row[""Text""]\n                new_text_str = str(new_text) if pd.notna(new_text) else """"\n\n                operation = ""noop""\n                if length == 0 and new_text_str:\n                    operation = ""insert""\n                elif length > 0 and not new_text_str:\n                    operation = ""delete""\n                elif length > 0 and new_text_str:\n                    operation = ""replace""\n\n                parts.append(f""<act {operation} file=\""{file_path}\"" offset=\""{offset}\"" len=\""{length}\"" />"")\n\n                if new_text_str and (operation == ""insert"" or operation == ""replace""):\n                    parts.append(_fenced_block(file_path, language, _clean_text(new_text_str)))\n\n                before = file_states.get(file_path, """")\n                after = _apply_change(before, offset, length, new_text)\n                file_states[file_path] = after\n                per_file_event_counts[file_path] = per_file_event_counts.get(file_path, 0) + 1\n\n                # Update cursor position after edit (cursor moves to end of inserted/replaced text)\n                per_file_cursor_positions[file_path] = (offset + len(new_text_str), 0)\n\n            case _:\n                raise ValueError(f""Unknown event type: {event_type}"")\n\n    return ""\n"".join(parts).strip()\n\n\ndef session_to_bash_formatted_transcript(\n    df: pd.DataFrame,\n    viewport_radius: int = 10,\n    normalize_terminal_output: bool = True,\n) -> str:\n    r""""""\n    Serialize a session to a bash-like transcript comprised of:\n      - Commands (bash fenced blocks): cat -n, sed -i 'S,Ec\...' && cat -n | sed -n 'VSTART,VENDp'\n      - Outputs (<stdout>...</stdout>) that reflect the file state after each action\n    Tracks per-file state and a per-file viewport. Viewport only shifts when selection moves out of bounds\n    or when first initialized.\n    """"""\n    file_states: Dict[str, str] = {}\n    per_file_viewport: Dict[str, Optional[Tuple[int, int]]] = {}\n\n    parts: List[str] = []\n    terminal_output_buffer: List[str] = []\n    pending_edits_before: Dict[str, Optional[str]] = {}\n\n    def _flush_terminal_output_buffer() -> None:\n        if not terminal_output_buffer:\n            return\n        aggregated = """".join(terminal_output_buffer)\n        out = aggregated\n        if normalize_terminal_output:\n            out = _normalize_terminal_output(out)\n        cleaned = _clean_text(out)\n        if cleaned.strip():\n            parts.append(f""<stdout>\n{cleaned}\n</stdout>"")\n        terminal_output_buffer.clear()\n\n    def _flush_pending_edit_for_file(target_file: str) -> None:\n        before_snapshot = pending_edits_before.get(target_file)\n        if before_snapshot is None:\n            return\n        after_state = file_states.get(target_file, """")\n        try:\n            start_line, end_line, repl_lines = _compute_changed_block_lines(before_snapshot, after_state)\n        except ValueError:\n            pending_edits_before[target_file] = None\n            return\n        before_total_lines = len(before_snapshot.splitlines())\n        if end_line < start_line:\n            escaped_lines = [_escape_single_quotes_for_sed(line) for line in repl_lines]\n            sed_payload = ""\n"".join(escaped_lines)\n            if start_line <= max(1, before_total_lines):\n                sed_cmd = f""sed -i '{start_line}i\\\n{sed_payload}' {target_file}""\n            else:\n                sed_cmd = f""sed -i '$a\\\n{sed_payload}' {target_file}""\n        elif not repl_lines:\n            sed_cmd = f""sed -i '{start_line},{end_line}d' {target_file}""\n        else:\n            escaped_lines = [_escape_single_quotes_for_sed(line) for line in repl_lines]\n            sed_payload = ""\n"".join(escaped_lines)\n            sed_cmd = f""sed -i '{start_line},{end_line}c\\\n{sed_payload}' {target_file}""\n        total_lines = len(after_state.splitlines())\n        center = (start_line + end_line) // 2\n        vp = _compute_viewport(total_lines, center, viewport_radius)\n        per_file_viewport[target_file] = vp\n        vstart, vend = vp\n        chained_cmd = f""{sed_cmd} && cat -n {target_file} | sed -n '{vstart},{vend}p'""\n        parts.append(_fenced_block(target_file, ""bash"", _clean_text(chained_cmd)))\n        viewport_output = _line_numbered_output(after_state, vstart, vend)\n        parts.append(f""<stdout>\n{viewport_output}\n</stdout>"")\n        pending_edits_before[target_file] = None\n\n    def _flush_all_pending_edits() -> None:\n        for fname in list(pending_edits_before.keys()):\n            _flush_pending_edit_for_file(fname)\n\n    for i in range(len(df)):\n        row = df.iloc[i]\n        file_path: str = row[""File""]\n        event_type = row[""Type""]\n\n        if i % 100 == 0:\n            breakpoint()\n        \n        match event_type:\n            case ""tab"":\n                _flush_all_pending_edits()\n                _flush_terminal_output_buffer()\n                text = row[""Text""]\n                if pd.notna(text):\n                    content = str(text).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                    file_states[file_path] = content\n                    # First open with full file capture\n                    cmd = f""cat -n {file_path}""\n                    parts.append(_fenced_block(file_path, ""bash"", _clean_text(cmd)))\n                    output = _line_numbered_output(content)\n                    parts.append(f""<stdout>\n{output}\n</stdout>"")\n                else:\n                    # File switch without content snapshot: show current viewport only\n                    content = file_states.get(file_path, """")\n                    total_lines = len(content.splitlines())\n                    vp = per_file_viewport.get(file_path)\n                    if not vp or vp[1] == 0:\n                        vp = _compute_viewport(total_lines, 1, viewport_radius)\n                        per_file_viewport[file_path] = vp\n                    if vp:\n                        vstart, vend = vp\n                        cmd = f""cat -n {file_path} | sed -n '{vstart},{vend}p'""\n                        parts.append(_fenced_block(file_path, ""bash"", _clean_text(cmd)))\n                        viewport_output = _line_numbered_output(content, vstart, vend)\n                        parts.append(f""<stdout>\n{viewport_output}\n</stdout>"")\n\n            case ""content"":\n                _flush_terminal_output_buffer()\n                offset = int(row[""RangeOffset""])\n                length = int(row[""RangeLength""])\n                new_text = row[""Text""]\n                before = file_states.get(file_path, """")\n                after = _apply_change(before, offset, length, new_text)\n                if pending_edits_before.get(file_path) is None:\n                    pending_edits_before[file_path] = before\n                file_states[file_path] = after\n\n            case ""selection_command"" | ""selection_mouse"" | ""selection_keyboard"":\n                # During an edit burst (pending edits), suppress flush and viewport emissions\n                if pending_edits_before.get(file_path) is None:\n                    _flush_terminal_output_buffer()\n                else:\n                    # Skip emitting viewport while edits are pending to avoid per-keystroke sed/cat spam\n                    break\n                offset = int(row[""RangeOffset""])\n                content = file_states.get(file_path, """")\n                total_lines = len(content.splitlines())\n                target_line = content[:offset].count(""\n"") + 1\n                vp = per_file_viewport.get(file_path)\n                should_emit = False\n                if not vp or vp[1] == 0:\n                    vp = _compute_viewport(total_lines, target_line, viewport_radius)\n                    per_file_viewport[file_path] = vp\n                    should_emit = True\n                else:\n                    vstart, vend = vp\n                    if target_line < vstart or target_line > vend:\n                        vp = _compute_viewport(total_lines, target_line, viewport_radius)\n                        per_file_viewport[file_path] = vp\n                        should_emit = True\n                if should_emit and vp:\n                    vstart, vend = vp\n                    cmd = f""cat -n {file_path} | sed -n '{vstart},{vend}p'""\n                    parts.append(_fenced_block(file_path, ""bash"", _clean_text(cmd)))\n                    viewport_output = _line_numbered_output(content, vstart, vend)\n                    parts.append(f""<stdout>\n{viewport_output}\n</stdout>"")\n\n            case ""terminal_command"":\n                _flush_all_pending_edits()\n                _flush_terminal_output_buffer()\n                command = row[""Text""]\n                command_str = str(command).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                parts.append(_fenced_block(file_path, ""bash"", _clean_text(command_str)))\n\n            case ""terminal_output"":\n                output = row[""Text""]\n                raw_output = str(output).replace(""\\n"", ""\n"").replace(""\\r"", ""\r"")\n                terminal_output_buffer.append(raw_output)\n\n            case ""terminal_focus"" | ""git_branch_checkout"":\n                _flush_all_pending_edits()\n                _flush_terminal_output_buffer()\n                # FIXME (f.srambical): handle these events \n                pass\n\n            case _:\n                _flush_all_pending_edits()\n                _flush_terminal_output_buffer()\n                raise ValueError(f""Unknown event type: {event_type}"")\n\n    _flush_all_pending_edits()\n    _flush_terminal_output_buffer()\n    return ""\n"".join(parts).strip()\n\ndef load_hf_csv(hf_path: str, split: str) -> Dataset:\n    loaded = load_dataset(hf_path, split=split)\n\n    assert isinstance(loaded, Dataset), ""Expected a Dataset from load_dataset""\n    return loaded\n\n\ndef _discover_local_sessions(root: Path) -> List[Path]:\n    # Recursively find all CSV files\n    paths: List[Path] = []\n    for p in root.rglob(""*.csv""):\n        if p.is_file():\n            paths.append(p)\n    paths.sort()\n    return paths\n\n\ndef _chunk_text(text: str, target_chars: int, overlap_chars: int) -> List[str]:\n    """"""Split a long text into overlapping chunks near target length.""""""\n    if target_chars <= 0:\n        return [text]\n    n = len(text)\n    if n <= target_chars:\n        return [text]\n\n    chunks: List[str] = []\n    start = 0\n    # Ensure sane overlap\n    overlap = max(0, min(overlap_chars, target_chars // 2))\n    while start < n:\n        end_target = min(start + target_chars, n)\n        if end_target < n:\n            end = end_target\n        else:\n            end = n\n        chunk = text[start:end].strip()\n        chunks.append(chunk)\n        if end == n:\n            break\n        # advance with overlap\n        start = max(0, end - overlap)\n        if start >= n:\n            break\n    return chunks\n\n\n",python,tab
+2,318,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"3:56:09 PM [info] Activating crowd-code\n3:56:09 PM [info] Recording started\n3:56:09 PM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,550,"extension-output-pdoom-org.crowd-code-#1-crowd-code",150,0,"3:56:10 PM [info] Git repository found\n3:56:10 PM [info] Git provider initialized successfully\n3:56:10 PM [info] Initial git state: [object Object]\n",Log,content
+4,230600,"TERMINAL",0,0,"",,terminal_focus
+5,230602,"crowd-pilot/crowd-pilot/serialization_utils.py",0,0,"",python,tab
+6,266711,"TERMINAL",0,0,"source /home/franz.srambical/crowd-pilot/.venv/bin/activate",,terminal_command

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-8e7b7877-c553-4d5c-a7c5-433adcd8112b1754287948136-2025_08_04-08.12.35.154/source.csv

@@ -0,0 +1,286 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,3,"utils/nn.py",0,0,"import math\nfrom typing import Tuple, Callable, List\n\nfrom flax import nnx\nimport jax\nimport jax.numpy as jnp\nimport einops\n\n\nclass PositionalEncoding(nnx.Module):\n    """"""https://uvadlc-notebooks.readthedocs.io/en/latest/tutorial_notebooks/JAX/tutorial6/Transformers_and_MHAttention.html""""""\n\n    def __init__(self, d_model: int, max_len: int = 5000):\n        self.d_model = d_model\n        self.max_len = max_len\n\n        pe = jnp.zeros((self.max_len, self.d_model))\n        position = jnp.arange(0, self.max_len, dtype=jnp.float32)[:, None]\n        div_term = jnp.exp(\n            jnp.arange(0, self.d_model, 2) * (-math.log(10000.0) / self.d_model)\n        )\n        pe = pe.at[:, 0::2].set(jnp.sin(position * div_term))\n        pe = pe.at[:, 1::2].set(jnp.cos(position * div_term))\n        self.pe = nnx.Variable(pe)\n\n    def __call__(self, x: jax.Array) -> jax.Array:\n        x = x + self.pe[: x.shape[2]]\n        return x\n\n\nclass STBlock(nnx.Module):\n    def __init__(\n        self,\n        dim: int,\n        ffn_dim: int,\n        num_heads: int,\n        dropout: float,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        rngs: nnx.Rngs,\n    ):\n        self.dim = dim\n        self.ffn_dim = ffn_dim\n        self.num_heads = num_heads\n        self.dropout = dropout\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n\n        self.spatial_pos_enc = PositionalEncoding(self.dim)\n        self.spatial_norm = nnx.LayerNorm(\n            num_features=self.dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.spatial_attention = nnx.MultiHeadAttention(\n            num_heads=self.num_heads,\n            in_features=self.dim,\n            qkv_features=self.dim,\n            dropout_rate=self.dropout,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            attention_fn=_create_flash_attention_fn(\n                self.use_flash_attention, is_causal=False\n            ),\n            rngs=rngs,\n            decode=False,\n        )\n\n        self.temporal_pos_enc = PositionalEncoding(self.dim)\n        self.temporal_norm = nnx.LayerNorm(\n            num_features=self.dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.temporal_attention = nnx.MultiHeadAttention(\n            num_heads=self.num_heads,\n            in_features=self.dim,\n            qkv_features=self.dim,\n            dropout_rate=self.dropout,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            attention_fn=_create_flash_attention_fn(\n                self.use_flash_attention, is_causal=True\n            ),\n            rngs=rngs,\n            decode=False,\n        )\n\n        self.ffn_norm = nnx.LayerNorm(\n            num_features=self.dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.ffn_dense1 = nnx.Linear(\n            in_features=self.dim,\n            out_features=self.ffn_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.ffn_dense2 = nnx.Linear(\n            in_features=self.ffn_dim,\n            out_features=self.dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n    @nnx.remat\n    def __call__(self, x_BTNM: jax.Array) -> jax.Array:\n        # --- Spatial attention ---\n        z_BTNM = self.spatial_pos_enc(x_BTNM)\n        z_BTNM = self.spatial_norm(z_BTNM)\n        z_BTNM = self.spatial_attention(z_BTNM)\n        x_BTNM = x_BTNM + z_BTNM\n\n        # --- Temporal attention ---\n        x_BNTM = x_BTNM.swapaxes(1, 2)\n        z_BNTM = self.temporal_pos_enc(x_BNTM)\n        z_BNTM = self.temporal_norm(z_BNTM)\n        z_BNTM = self.temporal_attention(z_BNTM)\n        x_BNTM = x_BNTM + z_BNTM\n        x_BTNM = x_BNTM.swapaxes(1, 2)\n\n        # --- Feedforward ---\n        z_BTNM = self.ffn_norm(x_BTNM)\n        z_BTND = self.ffn_dense1(z_BTNM)\n        z_BTND = jax.nn.gelu(z_BTND)\n        z_BTNM = self.ffn_dense2(z_BTND)\n        x_BTNM = x_BTNM + z_BTNM\n\n        return x_BTNM\n\n\nclass STTransformer(nnx.Module):\n    """"""\n    Dimension keys:\n        B: batch size\n        T: number of frames\n        N: number of patches per frame\n        I: number of input features\n        M: model dimension\n        D: FFN dimension\n        O: number of output features\n    """"""\n    def __init__(\n        self,\n        input_dim: int,\n        model_dim: int,\n        ffn_dim: int,\n        out_dim: int,\n        num_blocks: int,\n        num_heads: int,\n        dropout: float,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        rngs: nnx.Rngs,\n    ):\n        self.input_dim = input_dim\n        self.model_dim = model_dim\n        self.ffn_dim = ffn_dim\n        self.out_dim = out_dim\n        self.num_blocks = num_blocks\n        self.num_heads = num_heads\n        self.dropout = dropout\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n\n        self.input_norm1 = nnx.LayerNorm(\n            num_features=self.input_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.input_dense = nnx.Linear(\n            in_features=self.input_dim,\n            out_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.input_norm2 = nnx.LayerNorm(\n            num_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n        self.blocks = []\n        for _ in range(self.num_blocks):\n            self.blocks.append(\n                STBlock(\n                    dim=self.model_dim,\n                    ffn_dim=self.ffn_dim,\n                    num_heads=self.num_heads,\n                    dropout=self.dropout,\n                    param_dtype=self.param_dtype,\n                    dtype=self.dtype,\n                    use_flash_attention=self.use_flash_attention,\n                    rngs=rngs,\n                )\n            )\n\n        self.output_dense = nnx.Linear(\n            in_features=self.model_dim,\n            out_features=self.out_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n    def __call__(self, x_BTNI: jax.Array) -> jax.Array:\n        x_BTNI = self.input_norm1(x_BTNI)\n        x_BTNM = self.input_dense(x_BTNI)\n        x_BTNM = self.input_norm2(x_BTNM)\n\n        for block in self.blocks:\n            x_BTNM = block(x_BTNM)\n\n        x_BTNO = self.output_dense(x_BTNM)\n        return x_BTNO\n\nclass TransformerBlock(nnx.Module):\n    def __init__(\n        self,\n        model_dim: int,\n        ffn_dim: int,\n        num_heads: int,\n        dropout: float,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        decode: bool,\n        rngs: nnx.Rngs,\n    ):\n        self.model_dim = model_dim\n        self.ffn_dim = ffn_dim\n        self.num_heads = num_heads\n        self.dropout = dropout\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n        self.decode = decode\n\n        self.temporal_pos_enc = PositionalEncoding(self.model_dim)\n        self.spatial_pos_enc = PositionalEncoding(self.model_dim)\n        self.temporal_norm = nnx.LayerNorm(\n            num_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.spatial_norm = nnx.LayerNorm(\n            num_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.ffn_norm = nnx.LayerNorm(\n            num_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.temporal_attention = nnx.MultiHeadAttention(\n            num_heads=self.num_heads,\n            in_features=self.model_dim,\n            qkv_features=self.model_dim,\n            dropout_rate=self.dropout,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            attention_fn=_create_flash_attention_fn(\n                self.use_flash_attention, is_causal=True\n            ),\n            rngs=rngs,\n            decode=self.decode,\n        )\n        self.spatial_attention = nnx.MultiHeadAttention(\n            num_heads=self.num_heads,\n            in_features=self.model_dim,\n            qkv_features=self.model_dim,\n            dropout_rate=self.dropout,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            attention_fn=_create_flash_attention_fn(\n                self.use_flash_attention, is_causal=True\n            ),\n            rngs=rngs,\n            decode=self.decode,\n        )\n        self.ffn_dense1 = nnx.Linear(\n            in_features=self.model_dim,\n            out_features=self.ffn_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.ffn_dense2 = nnx.Linear(\n            in_features=self.ffn_dim,\n            out_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n    @nnx.remat\n    def __call__(self, x_BTNM: jax.Array, pos_index: Tuple[jax.Array, jax.Array] | None = None) -> jax.Array:\n        # --- Spatial attention ---\n        B, T, N, M = x_BTNM.shape\n        z_FNM = einops.rearrange(x_BTNM, ""b t n m -> (b t) n m"")\n        z_FNM = self.spatial_norm(z_FNM)\n        if self.decode:\n            assert pos_index is not None\n            z_FM = z_FNM[:, pos_index[1]]\n            z_F1M = jnp.reshape(z_FM, (B * T, 1, M))\n            z_F1M = self.spatial_attention(z_F1M)\n            z_FM = jnp.reshape(z_F1M, (B * T, M))\n            z_FNM = z_FNM.at[:, pos_index[1], :].set(z_FM)\n        else:\n            z_FNM = self.spatial_attention(z_FNM)\n        z_BTNM = einops.rearrange(z_FNM, ""(b t) n m -> b t n m"", t=T)\n        x_BTNM = x_BTNM + z_BTNM\n        # --- Temporal attention ---\n        z_PTM = einops.rearrange(x_BTNM, ""b t n m -> (b n) t m"")\n        z_PTM = self.temporal_norm(z_PTM)\n        if self.decode:\n            assert pos_index is not None\n            z_PM = z_PTM[:, pos_index[0]]\n            z_P1M = jnp.reshape(z_PM, (B * N, 1, M))\n            z_P1M = self.temporal_attention(z_P1M)\n            z_PM = jnp.reshape(z_P1M, (B * N, M))\n            z_PTM = z_PTM.at[:, pos_index[0], :].set(z_PM)\n        else:\n            z_PTM = self.temporal_attention(z_PTM)\n        z_BTNM = einops.rearrange(z_PTM, ""(b n) t m -> b t n m"", n=N)\n        x_BTNM = x_BTNM + z_BTNM\n        # --- Feedforward ---\n        z_BTNM = self.ffn_norm(x_BTNM)\n        z_BTND = self.ffn_dense1(z_BTNM)\n        z_BTND = jax.nn.gelu(z_BTND)\n        z_BTNM = self.ffn_dense2(z_BTND)\n        x_BTNM = x_BTNM + z_BTNM\n\n        return x_BTNM\n\nclass Transformer(nnx.Module):\n    """"""\n    Dimension keys:\n        B: batch size\n        T: number of frames\n        N: number of patches per frame\n        I: number of input features\n        M: model dimension\n        D: FFN dimension\n        O: number of output features\n        F: number of frames in batch\n        P: number of patch positions in batch\n    """"""\n    def __init__(\n        self,\n        input_dim: int,\n        model_dim: int,\n        ffn_dim: int,\n        out_dim: int,\n        num_blocks: int,\n        num_heads: int,\n        dropout: float,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        decode: bool,\n        rngs: nnx.Rngs,\n    ):\n        self.input_dim = input_dim\n        self.model_dim = model_dim\n        self.ffn_dim = ffn_dim\n        self.out_dim = out_dim\n        self.num_blocks = num_blocks\n        self.num_heads = num_heads\n        self.dropout = dropout\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n\n        self.pos_enc = PositionalEncoding(self.model_dim)\n        self.input_norm1 = nnx.LayerNorm(\n            num_features=self.input_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.input_dense = nnx.Linear(\n            in_features=self.input_dim,\n            out_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n        self.input_norm2 = nnx.LayerNorm(\n            num_features=self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n        self.blocks: List[TransformerBlock] = []\n        for _ in range(self.num_blocks):\n            self.blocks.append(\n                TransformerBlock(\n                    model_dim=self.model_dim,\n                    ffn_dim=self.ffn_dim,\n                    num_heads=self.num_heads,\n                    dropout=self.dropout,\n                    param_dtype=self.param_dtype,\n                    dtype=self.dtype,\n                    use_flash_attention=self.use_flash_attention,\n                    decode=decode,\n                    rngs=rngs,\n                )\n            )\n        self.output_dense = nnx.Linear(\n            in_features=self.model_dim,\n            out_features=self.out_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n    def __call__(self, x_BTNI: jax.Array, pos_index: Tuple[jax.Array, jax.Array] | None = None) -> jax.Array:\n        x_BTNI = self.input_norm1(x_BTNI)\n        x_BTNM = self.input_dense(x_BTNI)\n        x_BTNM = self.input_norm2(x_BTNM)\n        x_BTNM = self.pos_enc(x_BTNM)\n\n        for block in self.blocks:\n            x_BTNM = block(x_BTNM, pos_index)\n\n        x_BTNV = self.output_dense(x_BTNM)\n        return x_BTNV\n\ndef normalize(x: jax.Array) -> jax.Array:\n    return x / (jnp.linalg.norm(x, ord=2, axis=-1, keepdims=True) + 1e-8)\n\n\nclass VectorQuantizer(nnx.Module):\n    """"""\n    Dimension keys:\n        D: B * T * N\n        K: number of latents\n        L: latent dimension\n    """"""\n    def __init__(\n        self, latent_dim: int, num_latents: int, dropout: float, rngs: nnx.Rngs\n    ):\n        self.latent_dim = latent_dim\n        self.num_latents = num_latents\n        self.dropout = dropout\n\n        self.codebook = nnx.Param(\n            normalize(\n                nnx.initializers.lecun_uniform()(\n                    rngs.params(), (self.num_latents, self.latent_dim)\n                )\n            )\n        )\n        self.drop = nnx.Dropout(self.dropout, rngs=rngs)\n\n    def __call__(\n        self, x_DL: jax.Array, training: bool\n    ) -> Tuple[jax.Array, jax.Array, jax.Array, jax.Array]:\n        # --- Compute distances ---\n        x_DL = normalize(x_DL)\n        normalized_codebook_KL = normalize(self.codebook.value)\n        distance_DK = -jnp.matmul(x_DL, normalized_codebook_KL.T)\n        if training:\n            distance_DK = self.drop(distance_DK)\n\n        # --- Get indices and embeddings ---\n        indices_D = jnp.argmin(distance_DK, axis=-1)\n        z_DL = self.codebook[indices_D]\n\n        # --- Straight through estimator ---\n        z_q_DL = x_DL + jax.lax.stop_gradient(z_DL - x_DL)\n        return z_q_DL, z_DL, x_DL, indices_D\n\n    def get_codes(self, indices_E: jax.Array) -> jax.Array:\n        return self.codebook[indices_E]\n\n\ndef _create_flash_attention_fn(use_flash_attention: bool, is_causal: bool) -> Callable:\n    """"""\n    Create an attention function that uses flash attention if enabled.\n\n    flax.nnx.MultiHeadAttention provides tensors with shape (batch..., length, num_heads, head_dim),\n    but jax.nn.dot_product_attention expects (batch, length, num_heads, head_dim). We reshape to\n    ensure compatibility. cuDNN's flash attention additionally requires a sequence length that\n    is a multiple of 4. We pad the sequence length to the nearest multiple of 4 and mask\n    accordingly. Note that cuDNN requires the mask to be broadcast before calling the attention\n    function due to strict shape checking.\n    """"""\n\n    def attention_fn(query_BTHD, key_BSHD, value_BSHD, bias=None, mask_B111=None, **kwargs):\n        implementation = ""cudnn"" if use_flash_attention else None\n\n        def _merge_batch_dims(x):\n            return einops.rearrange(x, ""... l h k -> (...) l h k"")\n\n        def _pad(x, pad_size):\n            return jnp.pad(x, ((0, 0), (0, pad_size), (0, 0), (0, 0)))\n\n        original_shape = query_BTHD.shape\n        T = query_BTHD.shape[-3]\n        S = key_BSHD.shape[-3]\n\n        # Pad to nearest multiple of 4\n        Q = ((T + 3) // 4) * 4\n        pad_size_Q = Q - T\n        K = ((S + 3) // 4) * 4\n        pad_size_K = K - S\n\n        query_BQHD = _pad(_merge_batch_dims(query_BTHD), pad_size_Q)\n        key_BKHD = _pad(_merge_batch_dims(key_BSHD), pad_size_K)\n        value_BKHD = _pad(_merge_batch_dims(value_BSHD), pad_size_K)\n        B = query_BQHD.shape[0]\n\n        attention_mask = jnp.ones((Q, K), dtype=jnp.bool_)\n        attention_mask = attention_mask.at[Q:, :].set(False)\n        attention_mask = attention_mask.at[:, K:].set(False)\n\n        # Handle causal mask for cached decoder self-attention (from nnx.MultiHeadAttention)\n        if mask_B111 is not None:\n            # FIXME (f.srambical): Why do we need this?\n            mask_B111 = _merge_batch_dims(mask_B111)\n            # We need to broadcast T and S dimensions to target_seq_len since cudnn attention strictly checks the mask shape\n            # https://github.com/jax-ml/jax/issues/28974\n            # https://github.com/jax-ml/jax/blob/08c7677393672ccb85c10f1ed0bd506905c3c994/jax/_src/cudnn/fused_attention_stablehlo.py#L1830\n            # https://github.com/jax-ml/jax/blob/08c7677393672ccb85c10f1ed0bd506905c3c994/jax/_src/cudnn/fused_attention_stablehlo.py#L337\n            mask_B1TS = einops.repeat(mask_B111, ""... 1 1 -> ... t s"", t=Q, s=K)\n            mask_B1TS = mask_B111.astype(jnp.bool)\n        else:\n            mask_11TS = attention_mask[jnp.newaxis, jnp.newaxis, :, :]\n            mask_B1TS = jnp.broadcast_to(mask_11TS, (B, 1, Q, K))\n\n        bias_4d = _merge_batch_dims(bias) if bias is not None else None\n\n        # NOTE: jax.nn.dot_product_attention does not support dropout\n        output_4d = jax.nn.dot_product_attention(\n            query=query_BQHD,\n            key=key_BKHD,\n            value=value_BKHD,\n            bias=bias_4d,\n            mask=mask_B1TS,\n            implementation=implementation,\n            is_causal=is_causal,\n        )\n        return output_4d[..., :T, :, :].reshape(original_shape)\n\n    return attention_fn\n",python,tab
+2,156,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"8:12:35 AM [info] Activating crowd-code\n8:12:35 AM [info] Recording started\n8:12:35 AM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,270,"extension-output-pdoom-org.crowd-code-#1-crowd-code",150,0,"8:12:35 AM [info] Git repository found\n8:12:35 AM [info] Git provider initialized successfully\n8:12:35 AM [info] Initial git state: [object Object]\n",Log,content
+4,27027,"TERMINAL",0,0,"",,terminal_focus
+5,27028,"utils/nn.py",0,0,"",python,tab
+6,27772,"TERMINAL",0,0,"source /home/franz.srambical/jafar/.venv/bin/activate",,terminal_command
+7,30804,"TERMINAL",0,0,"salloc --gpus=1 --ntasks-per-node=1 --cpus-per-task=1 --mem=100G",,terminal_command
+8,30871,"TERMINAL",0,0,"]633;Csalloc: Granted job allocation 14895\r\n",,terminal_output
+9,30978,"TERMINAL",0,0,"salloc: Waiting for resource configuration\r\n",,terminal_output
+10,31985,"TERMINAL",0,0,"salloc: Nodes hai005 are ready for job\r\n",,terminal_output
+11,32407,"TERMINAL",0,0,"Running inside SLURM, Job ID 14895.\r\n",,terminal_output
+12,32483,"TERMINAL",0,0,"]0;franz.srambical@hai-login2:~/jafar[?2004h[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output
+13,33111,"TERMINAL",0,0,"n",,terminal_output
+14,33239,"TERMINAL",0,0,"s",,terminal_output
+15,33338,"TERMINAL",0,0,"y",,terminal_output
+16,33448,"TERMINAL",0,0,"s",,terminal_output
+17,33607,"TERMINAL",0,0,"",,terminal_output
+18,33984,"TERMINAL",0,0,"\r\nnsys     nsys-ui  \r\n[franz.srambical@hai005.haicore.berlin:~/jafar] $ nsys",,terminal_output
+19,34150,"TERMINAL",0,0,"\r\nnsys     nsys-ui  \r\n[franz.srambical@hai005.haicore.berlin:~/jafar] $ nsys",,terminal_output
+20,35164,"TERMINAL",0,0," ",,terminal_output
+21,37061,"TERMINAL",0,0,"p",,terminal_output
+22,37138,"TERMINAL",0,0,"r",,terminal_output
+23,37229,"TERMINAL",0,0,"o",,terminal_output
+24,37328,"TERMINAL",0,0,"f",,terminal_output
+25,37429,"TERMINAL",0,0,"i",,terminal_output
+26,37529,"TERMINAL",0,0,"l",,terminal_output
+27,37652,"TERMINAL",0,0,"e",,terminal_output
+28,37716,"TERMINAL",0,0," ",,terminal_output
+29,37984,"TERMINAL",0,0,"-",,terminal_output
+30,38182,"TERMINAL",0,0,"o",,terminal_output
+31,38288,"TERMINAL",0,0," ",,terminal_output
+32,40833,"TERMINAL",0,0,"m",,terminal_output
+33,41751,"TERMINAL",0,0,"",,terminal_output
+34,41984,"TERMINAL",0,0,"f",,terminal_output
+35,42303,"TERMINAL",0,0,"",,terminal_output
+36,42423,"TERMINAL",0,0,"t",,terminal_output
+37,42500,"TERMINAL",0,0,"e",,terminal_output
+38,42559,"TERMINAL",0,0,"s",,terminal_output
+39,42623,"TERMINAL",0,0,"t",,terminal_output
+40,42845,"TERMINAL",0,0,"_",,terminal_output
+41,43219,"TERMINAL",0,0,"p",,terminal_output
+42,43371,"TERMINAL",0,0,"r",,terminal_output
+43,43445,"TERMINAL",0,0,"o",,terminal_output
+44,43592,"TERMINAL",0,0,"f",,terminal_output
+45,43670,"TERMINAL",0,0,"i",,terminal_output
+46,43725,"TERMINAL",0,0,"l",,terminal_output
+47,43811,"TERMINAL",0,0,"e",,terminal_output
+48,44081,"TERMINAL",0,0," ",,terminal_output
+49,45881,"TERMINAL",0,0,"-",,terminal_output
+50,46075,"TERMINAL",0,0,"-f",,terminal_output
+51,46200,"TERMINAL",0,0,"o",,terminal_output
+52,46313,"TERMINAL",0,0,"r",,terminal_output
+53,46568,"TERMINAL",0,0,"c",,terminal_output
+54,46633,"TERMINAL",0,0,"e",,terminal_output
+55,46935,"TERMINAL",0,0,"_",,terminal_output
+56,47620,"TERMINAL",0,0,"",,terminal_output
+57,48781,"TERMINAL",0,0,"-",,terminal_output
+58,48978,"TERMINAL",0,0,"o",,terminal_output
+59,49100,"TERMINAL",0,0,"v",,terminal_output
+60,49293,"TERMINAL",0,0,"e",,terminal_output
+61,49368,"TERMINAL",0,0,"r",,terminal_output
+62,49569,"TERMINAL",0,0,"w",,terminal_output
+63,49643,"TERMINAL",0,0,"r",,terminal_output
+64,49754,"TERMINAL",0,0,"i",,terminal_output
+65,49862,"TERMINAL",0,0,"t",,terminal_output
+66,49947,"TERMINAL",0,0,"e",,terminal_output
+67,51426,"TERMINAL",0,0," ",,terminal_output
+68,51499,"TERMINAL",0,0,"t",,terminal_output
+69,52077,"TERMINAL",0,0,"r",,terminal_output
+70,52423,"TERMINAL",0,0,"y",,terminal_output
+71,52541,"TERMINAL",0,0,"e",,terminal_output
+72,52617,"TERMINAL",0,0," ",,terminal_output
+73,53355,"TERMINAL",0,0,"",,terminal_output
+74,53510,"TERMINAL",0,0,"",,terminal_output
+75,53668,"TERMINAL",0,0,"",,terminal_output
+76,54103,"TERMINAL",0,0,"u",,terminal_output
+77,54185,"TERMINAL",0,0,"e",,terminal_output
+78,54328,"TERMINAL",0,0," ",,terminal_output
+79,54478,"TERMINAL",0,0,"-",,terminal_output
+80,54649,"TERMINAL",0,0,"-",,terminal_output
+81,56255,"TERMINAL",0,0,"t",,terminal_output
+82,56397,"TERMINAL",0,0,"r",,terminal_output
+83,56544,"TERMINAL",0,0,"a",,terminal_output
+84,56661,"TERMINAL",0,0,"c",,terminal_output
+85,56781,"TERMINAL",0,0,"e",,terminal_output
+86,56938,"TERMINAL",0,0,"=",,terminal_output
+87,57153,"TERMINAL",0,0,"c",,terminal_output
+88,57312,"TERMINAL",0,0,"u",,terminal_output
+89,57395,"TERMINAL",0,0,"d",,terminal_output
+90,57446,"TERMINAL",0,0,"a",,terminal_output
+91,57835,"TERMINAL",0,0,",",,terminal_output
+92,59494,"TERMINAL",0,0,"n",,terminal_output
+93,59551,"TERMINAL",0,0,"v",,terminal_output
+94,59742,"TERMINAL",0,0,"t",,terminal_output
+95,59932,"TERMINAL",0,0,"x",,terminal_output
+96,61804,"TERMINAL",0,0," ",,terminal_output
+97,63065,"TERMINAL",0,0,"b",,terminal_output
+98,63194,"TERMINAL",0,0,"as",,terminal_output
+99,63280,"TERMINAL",0,0,"h",,terminal_output
+100,63429,"TERMINAL",0,0," e",,terminal_output
+101,63600,"TERMINAL",0,0,"x",,terminal_output
+102,63677,"TERMINAL",0,0,"p",,terminal_output
+103,63794,"TERMINAL",0,0,"e",,terminal_output
+104,63864,"TERMINAL",0,0,"ri",,terminal_output
+105,63916,"TERMINAL",0,0,"m",,terminal_output
+106,64042,"TERMINAL",0,0,"ents/",,terminal_output
+107,65215,"TERMINAL",0,0,"t",,terminal_output
+108,65607,"TERMINAL",0,0,"r",,terminal_output
+109,65719,"TERMINAL",0,0,"ai",,terminal_output
+110,65771,"TERMINAL",0,0,"n",,terminal_output
+111,65935,"TERMINAL",0,0,"",,terminal_output
+112,66603,"TERMINAL",0,0,"",,terminal_output
+113,67080,"TERMINAL",0,0,"",,terminal_output
+114,67235,"TERMINAL",0,0,"",,terminal_output
+115,67378,"TERMINAL",0,0,"",,terminal_output
+116,67520,"TERMINAL",0,0,"",,terminal_output
+117,67649,"TERMINAL",0,0,"",,terminal_output
+118,67836,"TERMINAL",0,0,"d",,terminal_output
+119,67899,"TERMINAL",0,0,"y",,terminal_output
+120,68013,"TERMINAL",0,0,"namics_grain_",,terminal_output
+121,68464,"TERMINAL",0,0,"t",,terminal_output
+122,68565,"TERMINAL",0,0,"ok",,terminal_output
+123,68674,"TERMINAL",0,0,"_",,terminal_output
+124,69397,"TERMINAL",0,0,"r",,terminal_output
+125,69607,"TERMINAL",0,0,"estore.sh ",,terminal_output
+126,70404,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+127,71800,"TERMINAL",0,0,"Collecting data...\r\n",,terminal_output
+128,77965,"TERMINAL",0,0,"WARNING:2025-08-04 08:13:53,008:jax._src.distributed:127: JAX detected proxy variable(s) in the environment as distributed setup: QUADD_INJECTION_PROXY. On some systems, this may cause a hang of distributed.initialize and you may need to unset these ENV variable(s)\r\nWARNING:jax._src.distributed:JAX detected proxy variable(s) in the environment as distributed setup: QUADD_INJECTION_PROXY. On some systems, this may cause a hang of distributed.initialize and you may need to unset these ENV variable(s)\r\n",,terminal_output
+129,78646,"TERMINAL",0,0,"Running on 1 devices.\r\n",,terminal_output
+130,86522,"TERMINAL",0,0,"Counting all components: ['dynamics', 'lam', 'tokenizer']\r\nParameter counts:\r\n{'dynamics': 26555392, 'lam': 35115232, 'tokenizer': 33750256, 'total': 95420880}\r\n",,terminal_output
+131,88929,"TERMINAL",0,0,"WARNING:absl:Metadata file does not exist: /home/franz.srambical/jafar/checkpoints/causal_dynamics_openai_grain_tok_restore/000290/_CHECKPOINT_METADATA\r\n",,terminal_output
+132,89797,"TERMINAL",0,0,"/fast/home/franz.srambical/jafar/.venv/lib/python3.10/site-packages/orbax/checkpoint/_src/serialization/type_handlers.py:1256: UserWarning: Sharding info not provided when restoring. Populating sharding info from sharding file. Please note restoration time will be slightly increased due to reading from file. Note also that this option is unsafe when restoring on a different topology than the checkpoint was saved with.\r\n  warnings.warn(\r\n",,terminal_output
+133,96833,"TERMINAL",0,0,"Starting training from step 0...\r\n",,terminal_output
+134,98035,"TERMINAL",0,0,"2025-08-04 08:14:13.078544: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\r\n",,terminal_output
+135,98192,"TERMINAL",0,0,"WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\r\nE0000 00:00:1754288053.237108 3090181 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\r\n",,terminal_output
+136,98265,"TERMINAL",0,0,"E0000 00:00:1754288053.280784 3090181 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\r\n",,terminal_output
+137,98709,"TERMINAL",0,0,"W0000 00:00:1754288053.643365 3090181 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\r\nW0000 00:00:1754288053.643382 3090181 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\r\nW0000 00:00:1754288053.643385 3090181 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\r\nW0000 00:00:1754288053.643387 3090181 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.\r\n",,terminal_output
+138,104271,"TERMINAL",0,0,"2025-08-04 08:14:19.312081: E external/xla/xla/backends/profiler/gpu/cupti_error_manager.cc:213] cuptiSubscribe: error 39: CUPTI_ERROR_MULTIPLE_SUBSCRIBERS_NOT_SUPPORTED\r\n2025-08-04 08:14:19.312098: E external/xla/xla/backends/profiler/gpu/cupti_error_manager.cc:242] cuptiGetResultString: ignored due to a previous error.\r\nE0804 08:14:19.312101 3090181 cupti_tracer.cc:1204] function cupti_interface_->Subscribe( &subscriber_, (CUpti_CallbackFunc)ApiCallback, this)failed with error \r\n",,terminal_output
+139,131807,"TERMINAL",0,0,"2025-08-04 08:14:46.853849: W external/xla/xla/service/gpu/autotuning/dot_search_space.cc:200] All configs were filtered out because none of them sufficiently match the hints. Maybe the hints set does not contain a good representative set of valid configs?Working around this by using the full hints set instead.\r\n2025-08-04 08:14:46.854633: W external/xla/xla/service/gpu/autotuning/dot_search_space.cc:200] All configs were filtered out because none of them sufficiently match the hints. Maybe the hints set does not contain a good representative set of valid configs?Working around this by using the full hints set instead.\r\n2025-08-04 08:14:46.855920: W external/xla/xla/service/gpu/autotuning/dot_search_space.cc:200] All configs were filtered out because none of them sufficiently match the hints. Maybe the hints set does not contain a good representative set of valid configs?Working around this by using the full hints set instead.\r\n2025-08-04 08:14:46.855946: W external/xla/xla/service/gpu/autotuning/dot_search_space.cc:200] All configs were filtered out because none of them sufficiently match the hints. Maybe the hints set does not contain a good representative set of valid configs?Working around this by using the full hints set instead.\r\n",,terminal_output
+140,156766,"TERMINAL",0,0,"Step 0, loss: 16.796998977661133\r\n",,terminal_output
+141,195251,"TERMINAL",0,0,"Step 1, loss: 1.9303642511367798\r\n",,terminal_output
+142,196253,"TERMINAL",0,0,"Step 2, loss: 2.342648506164551\r\n",,terminal_output
+143,197253,"TERMINAL",0,0,"Step 3, loss: 2.199798107147217\r\n",,terminal_output
+144,198255,"TERMINAL",0,0,"Step 4, loss: 1.6089359521865845\r\nSaved checkpoint at step 5\r\n",,terminal_output
+145,199252,"TERMINAL",0,0,"2025-08-04 08:15:53.706165: E external/xla/xla/backends/profiler/gpu/cupti_error_manager.cc:157] cuptiFinalize: ignored due to a previous error.\r\n2025-08-04 08:15:53.706187: E external/xla/xla/backends/profiler/gpu/cupti_error_manager.cc:242] cuptiGetResultString: ignored due to a previous error.\r\nE0804 08:15:53.706190 3090181 cupti_tracer.cc:1317] function cupti_interface_->Finalize()failed with error \r\n2025-08-04 08:15:53.707018: E external/xla/xla/backends/profiler/gpu/cupti_error_manager.cc:150] cuptiGetTimestamp: ignored due to a previous error.\r\n",,terminal_output
+146,210246,"TERMINAL",0,0,"2025-08-04 08:16:04.208428: E external/xla/xla/backends/profiler/gpu/cupti_error_manager.cc:150] cuptiGetTimestamp: ignored due to a previous error.\r\n",,terminal_output
+147,249252,"TERMINAL",0,0,"/home/franz.srambical/.local/share/uv/python/cpython-3.10.18-linux-x86_64-gnu/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 10 leaked shared_memory objects to clean up at shutdown\r\n  warnings.warn('resource_tracker: There appear to be %d '\r\n",,terminal_output
+148,253252,"TERMINAL",0,0,"Generating '/var/tmp/nsys-report-4202.qdstrm'\r\n",,terminal_output
+149,255249,"TERMINAL",0,0,"\r[1/1] [0%                          ] test_profile.nsys-rep\r[1/1] [0%                          ] test_profile.nsys-rep\r[1/1] [10%                         ] test_profile.nsys-rep\r[1/1] [8%                          ] test_profile.nsys-rep\r[1/1] [7%                          ] test_profile.nsys-rep\r[1/1] [6%                          ] test_profile.nsys-rep\r[1/1] [12%                         ] test_profile.nsys-rep\r[1/1] [10%                         ] test_profile.nsys-rep\r[1/1] [9%                          ] test_profile.nsys-rep\r[1/1] [8%                          ] test_profile.nsys-rep\r[1/1] [7%                          ] test_profile.nsys-rep\r[1/1] [6%                          ] test_profile.nsys-rep\r[1/1] [8%                          ] test_profile.nsys-rep\r[1/1] [10%                         ] test_profile.nsys-rep\r[1/1] [11%                         ] test_profile.nsys-rep\r[1/1] [12%                         ] test_profile.nsys-rep\r[1/1] [13%                         ] test_profile.nsys-rep\r[1/1] [12%                         ] test_profile.nsys-rep\r[1/1] [13%                         ] test_profile.nsys-rep\r[1/1] [12%                         ] test_profile.nsys-rep\r[1/1] [13%                         ] test_profile.nsys-rep\r[1/1] [14%                         ] test_profile.nsys-rep\r[1/1] [13%                         ] test_profile.nsys-rep\r[1/1] [12%                         ] test_profile.nsys-rep\r[1/1] [13%                         ] test_profile.nsys-rep\r[1/1] [12%                         ] test_profile.nsys-rep\r[1/1] [11%                         ] test_profile.nsys-rep\r[1/1] [10%                         ] test_profile.nsys-rep\r[1/1] [9%                          ] test_profile.nsys-rep\r[1/1] [8%                          ] test_profile.nsys-rep\r[1/1] [7%                          ] test_profile.nsys-rep\r[1/1] [6%                          ] test_profile.nsys-rep\r[1/1] [7%                          ] test_profile.nsys-rep\r[1/1] [8%                          ] test_profile.nsys-rep\r[1/1] [9%                          ] test_profile.nsys-rep\r[1/1] [11%                         ] test_profile.nsys-rep\r[1/1] [13%                         ] test_profile.nsys-rep\r[1/1] [=15%                        ] test_profile.nsys-rep\r[1/1] [=17%                        ] test_profile.nsys-rep\r[1/1] [==19%                       ] test_profile.nsys-rep\r[1/1] [==21%                       ] test_profile.nsys-rep\r[1/1] [===22%                      ] test_profile.nsys-rep\r[1/1] [===23%                      ] test_profile.nsys-rep\r[1/1] [====25%                     ] test_profile.nsys-rep\r[1/1] [====27%                     ] test_profile.nsys-rep\r[1/1] [=====30%                    ] test_profile.nsys-rep\r[1/1] [=====32%                    ] test_profile.nsys-rep\r[1/1] [======34%                   ] test_profile.nsys-rep\r[1/1] [=======36%                  ] test_profile.nsys-rep\r[1/1] [=======37%                  ] test_profile.nsys-rep\r[1/1] [=======38%                  ] test_profile.nsys-rep\r[1/1] [=======39%                  ] test_profile.nsys-rep\r[1/1] [========40%                 ] test_profile.nsys-rep\r[1/1] [========41%                 ] test_profile.nsys-rep\r[1/1] [========42%                 ] test_profile.nsys-rep\r[1/1] [=========43%                ] test_profile.nsys-rep\r[1/1] [=========44%                ] test_profile.nsys-rep\r[1/1] [=========45%                ] test_profile.nsys-rep\r[1/1] [==========47%               ] test_profile.nsys-rep\r[1/1] [==========49%               ] test_profile.nsys-rep\r[1/1] [===========51%              ] test_profile.nsys-rep\r[1/1] [===========53%              ] test_profile.nsys-rep\r[1/1] [============55%             ] test_profile.nsys-rep\r[1/1] [============57%             ] test_profile.nsys-rep\r[1/1] [=============59%            ] test_profile.nsys-rep\r[1/1] [==============61%           ] test_profile.nsys-rep\r[1/1] [==============63%           ] test_profile.nsys-rep\r[1/1] [===============65%          ] test_profile.nsys-rep\r[1/1] [================68%         ] test_profile.nsys-rep\r[1/1] [================70%         ] test_profile.nsys-rep\r[1/1] [=================72%        ] test_profile.nsys-rep\r[1/1] [=================74%        ] test_profile.nsys-rep\r[1/1] [==================76%       ] test_profile.nsys-rep\r[1/1] [==================77%       ] test_profile.nsys-rep\r[1/1] [===================82%      ] test_profile.nsys-rep\r[1/1] [========================100%] test_profile.nsys-rep",,terminal_output
+150,260255,"TERMINAL",0,0,"\r[1/1] [========================100%] test_profile.nsys-rep\r\nGenerated:\r\n    /fast/home/franz.srambical/jafar/test_profile.nsys-rep\r\n]0;franz.srambical@hai-login2:~/jafar[?2004h[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output
+151,341900,"TERMINAL",0,0,"\r[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output
+152,955976,"TERMINAL",0,0,"n",,terminal_output
+153,956091,"TERMINAL",0,0,"c",,terminal_output
+154,956287,"TERMINAL",0,0,"u",,terminal_output
+155,961697,"TERMINAL",0,0," ",,terminal_output
+156,961792,"TERMINAL",0,0,"-",,terminal_output
+157,961973,"TERMINAL",0,0,"o",,terminal_output
+158,962062,"TERMINAL",0,0," ",,terminal_output
+159,963985,"TERMINAL",0,0,"t",,terminal_output
+160,964066,"TERMINAL",0,0,"e",,terminal_output
+161,964152,"TERMINAL",0,0,"s",,terminal_output
+162,964204,"TERMINAL",0,0,"t",,terminal_output
+163,964485,"TERMINAL",0,0,"_",,terminal_output
+164,964669,"TERMINAL",0,0,"p",,terminal_output
+165,964751,"TERMINAL",0,0,"r",,terminal_output
+166,964865,"TERMINAL",0,0,"o",,terminal_output
+167,964979,"TERMINAL",0,0,"f",,terminal_output
+168,965104,"TERMINAL",0,0,"i",,terminal_output
+169,965164,"TERMINAL",0,0,"l",,terminal_output
+170,965256,"TERMINAL",0,0,"e",,terminal_output
+171,967201,"TERMINAL",0,0," ",,terminal_output
+172,967345,"TERMINAL",0,0,"-",,terminal_output
+173,967491,"TERMINAL",0,0,"-",,terminal_output
+174,967546,"TERMINAL",0,0,"s",,terminal_output
+175,967605,"TERMINAL",0,0,"e",,terminal_output
+176,967696,"TERMINAL",0,0,"t",,terminal_output
+177,967801,"TERMINAL",0,0," ",,terminal_output
+178,968014,"TERMINAL",0,0,"f",,terminal_output
+179,968095,"TERMINAL",0,0,"u",,terminal_output
+180,968192,"TERMINAL",0,0,"l",,terminal_output
+181,968385,"TERMINAL",0,0,"l",,terminal_output
+182,968479,"TERMINAL",0,0," ",,terminal_output
+183,970729,"TERMINAL",0,0,"bas",,terminal_output
+184,970834,"TERMINAL",0,0,"h",,terminal_output
+185,970969,"TERMINAL",0,0," e",,terminal_output
+186,971152,"TERMINAL",0,0,"x",,terminal_output
+187,971246,"TERMINAL",0,0,"p",,terminal_output
+188,971509,"TERMINAL",0,0,"erim",,terminal_output
+189,971611,"TERMINAL",0,0,"ents/",,terminal_output
+190,972533,"TERMINAL",0,0,"t",,terminal_output
+191,972669,"TERMINAL",0,0,"r",,terminal_output
+192,972766,"TERMINAL",0,0,"ai",,terminal_output
+193,972843,"TERMINAL",0,0,"n",,terminal_output
+194,973130,"TERMINAL",0,0,"",,terminal_output
+195,973284,"TERMINAL",0,0,"",,terminal_output
+196,973415,"TERMINAL",0,0,"",,terminal_output
+197,973572,"TERMINAL",0,0,"",,terminal_output
+198,973721,"TERMINAL",0,0,"",,terminal_output
+199,973900,"TERMINAL",0,0,"d",,terminal_output
+200,973998,"TERMINAL",0,0,"y",,terminal_output
+201,974129,"TERMINAL",0,0,"namics_grain_",,terminal_output
+202,974856,"TERMINAL",0,0,"t",,terminal_output
+203,975063,"TERMINAL",0,0,"ok_",,terminal_output
+204,976053,"TERMINAL",0,0,"r",,terminal_output
+205,976210,"TERMINAL",0,0,"estore.sh ",,terminal_output
+206,976605,"TERMINAL",0,0,"\r\n[?2004l\rbash: ncu: command not found\r\n]0;franz.srambical@hai-login2:~/jafar[?2004h[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output
+207,978326,"TERMINAL",0,0,"n",,terminal_output
+208,978392,"TERMINAL",0,0,"c",,terminal_output
+209,978620,"TERMINAL",0,0,"",,terminal_output
+210,978862,"TERMINAL",0,0,"\r\nncclras           ncdu              ncurses6-config   ncursesw6-config  \r\n[franz.srambical@hai005.haicore.berlin:~/jafar] $ nc",,terminal_output
+211,979116,"TERMINAL",0,0,"u",,terminal_output
+212,981520,"TERMINAL",0,0,"",,terminal_output
+213,982234,"TERMINAL",0,0,"u",,terminal_output
+214,982369,"TERMINAL",0,0,"rses",,terminal_output
+215,982539,"TERMINAL",0,0,"",,terminal_output
+216,983359,"TERMINAL",0,0,"",,terminal_output
+217,992859,"TERMINAL",0,0,"m",,terminal_output
+218,992923,"TERMINAL",0,0,"o",,terminal_output
+219,993010,"TERMINAL",0,0,"d",,terminal_output
+220,993130,"TERMINAL",0,0,"u",,terminal_output
+221,993284,"TERMINAL",0,0,"l",,terminal_output
+222,993401,"TERMINAL",0,0,"e",,terminal_output
+223,993573,"TERMINAL",0,0,"",,terminal_output
+224,993929,"TERMINAL",0,0,"\r\nmodule              modulemd-validator  \r\n[franz.srambical@hai005.haicore.berlin:~/jafar] $ module",,terminal_output
+225,995094,"TERMINAL",0,0," ",,terminal_output
+226,995401,"TERMINAL",0,0,"",,terminal_output
+227,995876,"TERMINAL",0,0,"l",,terminal_output
+228,996199,"TERMINAL",0,0,"",,terminal_output
+229,996597,"TERMINAL",0,0,"l",,terminal_output
+230,996669,"TERMINAL",0,0,"i",,terminal_output
+231,996885,"TERMINAL",0,0,"t",,terminal_output
+232,997246,"TERMINAL",0,0,"s",,terminal_output
+233,997333,"TERMINAL",0,0,"t",,terminal_output
+234,997504,"TERMINAL",0,0," ",,terminal_output
+235,997813,"TERMINAL",0,0,"\r\n[?2004l\rNo modules loaded\r\n]0;franz.srambical@hai-login2:~/jafar[?2004h[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output
+236,1000434,"TERMINAL",0,0,"module list ",,terminal_output
+237,1001046,"TERMINAL",0,0,"",,terminal_output
+238,1001216,"TERMINAL",0,0,"",,terminal_output
+239,1001368,"TERMINAL",0,0,"",,terminal_output
+240,1001521,"TERMINAL",0,0,"",,terminal_output
+241,1001685,"TERMINAL",0,0,"",,terminal_output
+242,1001849,"TERMINAL",0,0,"s",,terminal_output
+243,1001964,"TERMINAL",0,0,"p",,terminal_output
+244,1002024,"TERMINAL",0,0,"i",,terminal_output
+245,1002120,"TERMINAL",0,0,"d",,terminal_output
+246,1002330,"TERMINAL",0,0,"e",,terminal_output
+247,1002404,"TERMINAL",0,0,"r",,terminal_output
+248,1002594,"TERMINAL",0,0," ",,terminal_output
+249,1004453,"TERMINAL",0,0,"n",,terminal_output
+250,1004545,"TERMINAL",0,0,"s",,terminal_output
+251,1004621,"TERMINAL",0,0,"i",,terminal_output
+252,1004704,"TERMINAL",0,0,"g",,terminal_output
+253,1004804,"TERMINAL",0,0,"h",,terminal_output
+254,1004899,"TERMINAL",0,0,"t",,terminal_output
+255,1004994,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+256,1006202,"TERMINAL",0,0,"\r\n----------------------------------------------------------------------------------------------------------------\r\n  insight: insight/0.20.5 (E)\r\n----------------------------------------------------------------------------------------------------------------\r\n    This extension is provided by the following modules. To access the extension you must load one of the following \r\nmodules. Note that any module names in parentheses show the module location in the software hierarchy.\r\n\r\n\r\n       R-bundle-CRAN/2024.11-foss-2024a\r\n\r\n\r\nNames marked by a trailing (E) are extensions provided by another module.\r\n\r\n\r\n\r\n \r\n\r\n]0;franz.srambical@hai-login2:~/jafar[?2004h[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output
+257,1021117,"TERMINAL",0,0,"mo",,terminal_output
+258,1021320,"TERMINAL",0,0,"d",,terminal_output
+259,1021376,"TERMINAL",0,0,"u",,terminal_output
+260,1021517,"TERMINAL",0,0,"l",,terminal_output
+261,1021569,"TERMINAL",0,0,"e",,terminal_output
+262,1021691,"TERMINAL",0,0," ",,terminal_output
+263,1021831,"TERMINAL",0,0,"l",,terminal_output
+264,1021979,"TERMINAL",0,0,"o",,terminal_output
+265,1022033,"TERMINAL",0,0,"a",,terminal_output
+266,1022113,"TERMINAL",0,0,"d",,terminal_output
+267,1022183,"TERMINAL",0,0," ",,terminal_output
+268,1022297,"TERMINAL",0,0,"n",,terminal_output
+269,1022373,"TERMINAL",0,0,"s",,terminal_output
+270,1022465,"TERMINAL",0,0,"i",,terminal_output
+271,1022552,"TERMINAL",0,0,"g",,terminal_output
+272,1022649,"TERMINAL",0,0,"h",,terminal_output
+273,1022777,"TERMINAL",0,0,"t",,terminal_output
+274,1023018,"TERMINAL",0,0,"_",,terminal_output
+275,1023253,"TERMINAL",0,0,"oc",,terminal_output
+276,1023589,"TERMINAL",0,0,"",,terminal_output
+277,1023704,"TERMINAL",0,0,"",,terminal_output
+278,1023778,"TERMINAL",0,0,"c",,terminal_output
+279,1023900,"TERMINAL",0,0,"o",,terminal_output
+280,1024023,"TERMINAL",0,0,"m",,terminal_output
+281,1024155,"TERMINAL",0,0,"p",,terminal_output
+282,1024259,"TERMINAL",0,0,"u",,terminal_output
+283,1024395,"TERMINAL",0,0,"te",,terminal_output
+284,1024488,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+285,1025856,"TERMINAL",0,0,"Lmod has detected the following error:  The following module(s) are unknown: ""nsight_compute""\r\n\r\nPlease check the spelling or version number. Also try ""module spider ...""\r\nIt is also possible your cache file is out-of-date; it may help to try:\r\n  $ module --ignore_cache load ""nsight_compute""\r\n\r\nAlso make sure that all modulefiles written in TCL start with the string #%Module\r\n\r\n\r\n\r\n]0;franz.srambical@hai-login2:~/jafar[?2004h[franz.srambical@hai005.haicore.berlin:~/jafar] $ ",,terminal_output

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-91d7eee6-8ee7-4f50-ba4c-546c6e0f99451755438444066-2025_08_17-15.47.30.74/source.csv

@@ -0,0 +1,24 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,3,"train_lam.py",0,0,"from dataclasses import dataclass, field\nimport os\nfrom typing import cast\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 models.lam import LatentActionModel\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\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    vq_beta: float = 0.25\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    vq_reset_thresh: int = 50\n    # LAM\n    model_dim: int = 512\n    ffn_dim: int = 2048\n    latent_dim: int = 32\n    num_latents: int = 6\n    patch_size: int = 16\n    num_blocks: int = 4\n    num_heads: int = 8\n    dropout: float = 0.0\n    codebook_dropout: float = 0.0\n    param_dtype = jnp.float32\n    dtype = jnp.bfloat16\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    name: str = ""train_lam""\n    tags: list[str] = field(default_factory=lambda: [""lam""])\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 10000\n    log_checkpoint_keep_period: int = 20000\n    wandb_id: str = """"\n    use_flash_attention: bool = True\n\n\nargs = tyro.cli(Args)\n\n\ndef lam_loss_fn(\n    model: LatentActionModel, inputs: dict\n) -> tuple[jax.Array, tuple[jax.Array, jax.Array, dict]]:\n    # --- Compute loss ---\n    gt = jnp.asarray(inputs[""videos""], dtype=jnp.float32) / 255.0\n    inputs[""videos""] = gt.astype(args.dtype)\n    model.train()\n    outputs = model(inputs, training=True)\n    outputs[""recon""] = outputs[""recon""].astype(jnp.float32)\n    gt_future_frames = gt[:, 1:]\n    mse = jnp.square(gt_future_frames - 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 = gt_future_frames.clip(0, 1).reshape(-1, *gt_future_frames.shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = jnp.asarray(pix.psnr(gt, recon)).mean()\n    ssim = jnp.asarray(pix.ssim(gt, recon)).mean()\n    count_fn = jax.vmap(lambda i: (outputs[""indices""] == i).sum())\n    index_counts = count_fn(jnp.arange(args.num_latents))\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=(index_counts != 0).mean(),\n    )\n    return loss, (outputs[""recon""], index_counts, metrics)\n\n\n@nnx.jit\ndef train_step(\n    lam: LatentActionModel,\n    optimizer: nnx.Optimizer,\n    inputs: dict,\n    action_last_active: jax.Array,\n    rng: jax.Array,\n) -> tuple[jax.Array, jax.Array, jax.Array, dict]:\n    def loss_fn(\n        model: LatentActionModel,\n    ) -> tuple[jax.Array, tuple[jax.Array, jax.Array, dict]]:\n        return lam_loss_fn(model, inputs)\n\n    # --- Update model ---\n    (loss, (recon, idx_counts, metrics)), grads = nnx.value_and_grad(\n        loss_fn, has_aux=True\n    )(lam)\n    optimizer.update(grads)\n\n    # --- Reset inactive latent actions ---\n    codebook = lam.vq.codebook\n    num_codes = len(codebook)\n    active_codes = idx_counts != 0.0\n    action_last_active = jnp.where(active_codes, 0, action_last_active + 1)\n    p_code = active_codes / active_codes.sum()\n    reset_idxs = jax.random.choice(rng, num_codes, shape=(num_codes,), p=p_code)\n    do_reset = action_last_active >= args.vq_reset_thresh\n    new_codebook = jnp.where(\n        jnp.expand_dims(do_reset, -1), codebook[reset_idxs], codebook.value\n    )\n    lam.vq.codebook.value = new_codebook\n    action_last_active = jnp.where(do_reset, 0, action_last_active)\n    return loss, recon, action_last_active, 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.key(args.seed)\n\n    # --- Initialize model ---\n    rng, _rng = jax.random.split(rng)\n    rngs = nnx.Rngs(_rng)\n    lam = LatentActionModel(\n        in_dim=args.image_channels,\n        model_dim=args.model_dim,\n        ffn_dim=args.ffn_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        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n        use_flash_attention=args.use_flash_attention,\n        rngs=rngs,\n    )\n\n    # Count parameters\n    _, params, _ = nnx.split(lam, 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    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.Optimizer(lam, 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    videos_sharding = NamedSharding(mesh, PartitionSpec(""data"", None, None, None, None))\n\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    # --- Initialize checkpoint manager ---\n    step = 0\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        ""dataloader_state"",\n        grain.checkpoint.CheckpointSave,\n        cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n    )\n    handler_registry.add(\n        ""dataloader_state"",\n        grain.checkpoint.CheckpointRestore,\n        cast(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\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    image_shape = (args.image_height, args.image_width, args.image_channels)\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        abstract_optimizer = nnx.eval_shape(lambda: optimizer)\n        abstract_optimizer_state = nnx.state(abstract_optimizer)\n        restored = checkpoint_manager.restore(\n            checkpoint_manager.latest_step(),\n            args=ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                dataloader_state=grain.checkpoint.CheckpointRestore(grain_iterator),  # type: ignore\n            ),\n        )\n        restored_optimizer_state = restored[""model_state""]\n        nnx.update(optimizer, restored_optimizer_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\n    # --- TRAIN LOOP ---\n    dataloader = (\n        jax.make_array_from_process_local_data(videos_sharding, elem)\n        for elem in grain_iterator\n    )\n    print(f""Starting training from step {step}..."")\n    action_last_active = jnp.zeros(args.num_latents, dtype=jnp.int32)\n    while step < args.num_steps:\n        for videos in dataloader:\n            # --- Train step ---\n            rng, _rng = jax.random.split(rng)\n\n            inputs = dict(videos=videos, rng=_rng)\n            rng, _rng = jax.random.split(rng)\n            loss, recon, action_last_active, metrics = train_step(\n                lam, optimizer, inputs, action_last_active, _rng\n            )\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, 1:].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                    # 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            # --- Checkpointing ---\n            if args.save_ckpt and step % args.log_checkpoint_interval == 0:\n                optimizer_state = nnx.state(optimizer)\n                checkpoint_manager.save(\n                    step,\n                    args=ocp.args.Composite(\n                        model_state=ocp.args.PyTreeSave(optimizer_state),  # type: ignore\n                        dataloader_state=grain.checkpoint.CheckpointSave(  # type: ignore\n                            grain_iterator  # type: ignore\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
+2,131,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"3:47:30 PM [info] Activating crowd-code\n3:47:30 PM [info] Recording started\n3:47:30 PM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,176,"extension-output-pdoom-org.crowd-code-#1-crowd-code",150,0,"3:47:30 PM [info] Git repository found\n3:47:30 PM [info] Git provider initialized successfully\n3:47:30 PM [info] Initial git state: [object Object]\n",Log,content
+4,457,"TERMINAL",0,0,"",,terminal_focus
+5,627,"train_lam.py",0,0,"",python,tab
+6,174755,"train_lam.py",6883,63,"        mu_dtype=args.dtype,\n",python,content
+7,175106,"train_lam.py",0,0,"Switched from branch 'momentum-in-fp32' to 'main'",python,git_branch_checkout
+8,186959,"slurm/dev/mihir/horeka/yolo-runs/train_dynamics_new_arch.sbatch",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=2\n#SBATCH --ntasks-per-node=4\n#SBATCH --time=48:00:00\n#SBATCH --partition=accelerated\n#SBATCH --cpus-per-task=5\n#SBATCH --gres=gpu:4\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/%x_%j.log\n#SBATCH --job-name=train_dyn_new_arch-bugfixed-temporal-shift\n\n# Log the sbatch script\ncat $0\n\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\nsource .venv/bin/activate\n\narray_records_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_new/open_ai_minecraft_arrayrecords_chunked\n\njob_name=$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\n\nenv | grep SLURM\n\nsrun python train_dynamics.py \\n    --save_ckpt \\n    --num_steps=50000 \\n    --warmup_steps=2500 \\n    --wsd_decay_steps=5000 \\n    --ckpt_dir $CHECKPOINT_DIR \\n    --batch_size=96 \\n    --max_lr=1e-4 \\n    --log_image_interval=1000 \\n    --log \\n    --log_checkpoint_interval=1000 \\n    --name=dynamics-new-arch-bugfix-temporal-shift-$slurm_job_id \\n    --tags dynamics new-arch bug-fix \\n    --entity instant-uv \\n    --project jafar \\n    --tokenizer_checkpoint=$tokenizer_ckpt_dir \\n    --data_dir $array_records_dir \\n ",shellscript,tab
+9,188008,"train_lam.py",0,0,"from dataclasses import dataclass, field\nimport os\nfrom typing import cast\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 models.lam import LatentActionModel\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\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    vq_beta: float = 0.25\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    vq_reset_thresh: int = 50\n    # LAM\n    model_dim: int = 512\n    ffn_dim: int = 2048\n    latent_dim: int = 32\n    num_latents: int = 6\n    patch_size: int = 16\n    num_blocks: int = 4\n    num_heads: int = 8\n    dropout: float = 0.0\n    codebook_dropout: float = 0.0\n    param_dtype = jnp.float32\n    dtype = jnp.bfloat16\n    # Logging\n    log: bool = False\n    entity: str = """"\n    project: str = """"\n    name: str = ""train_lam""\n    tags: list[str] = field(default_factory=lambda: [""lam""])\n    log_interval: int = 5\n    log_image_interval: int = 250\n    ckpt_dir: str = """"\n    log_checkpoint_interval: int = 10000\n    log_checkpoint_keep_period: int = 20000\n    wandb_id: str = """"\n    use_flash_attention: bool = True\n\n\nargs = tyro.cli(Args)\n\n\ndef lam_loss_fn(\n    model: LatentActionModel, inputs: dict\n) -> tuple[jax.Array, tuple[jax.Array, jax.Array, dict]]:\n    # --- Compute loss ---\n    gt = jnp.asarray(inputs[""videos""], dtype=jnp.float32) / 255.0\n    inputs[""videos""] = gt.astype(args.dtype)\n    model.train()\n    outputs = model(inputs, training=True)\n    outputs[""recon""] = outputs[""recon""].astype(jnp.float32)\n    gt_future_frames = gt[:, 1:]\n    mse = jnp.square(gt_future_frames - 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 = gt_future_frames.clip(0, 1).reshape(-1, *gt_future_frames.shape[2:])\n    recon = outputs[""recon""].clip(0, 1).reshape(-1, *outputs[""recon""].shape[2:])\n    psnr = jnp.asarray(pix.psnr(gt, recon)).mean()\n    ssim = jnp.asarray(pix.ssim(gt, recon)).mean()\n    count_fn = jax.vmap(lambda i: (outputs[""indices""] == i).sum())\n    index_counts = count_fn(jnp.arange(args.num_latents))\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=(index_counts != 0).mean(),\n    )\n    return loss, (outputs[""recon""], index_counts, metrics)\n\n\n@nnx.jit\ndef train_step(\n    lam: LatentActionModel,\n    optimizer: nnx.Optimizer,\n    inputs: dict,\n    action_last_active: jax.Array,\n    rng: jax.Array,\n) -> tuple[jax.Array, jax.Array, jax.Array, dict]:\n    def loss_fn(\n        model: LatentActionModel,\n    ) -> tuple[jax.Array, tuple[jax.Array, jax.Array, dict]]:\n        return lam_loss_fn(model, inputs)\n\n    # --- Update model ---\n    (loss, (recon, idx_counts, metrics)), grads = nnx.value_and_grad(\n        loss_fn, has_aux=True\n    )(lam)\n    optimizer.update(grads)\n\n    # --- Reset inactive latent actions ---\n    codebook = lam.vq.codebook\n    num_codes = len(codebook)\n    active_codes = idx_counts != 0.0\n    action_last_active = jnp.where(active_codes, 0, action_last_active + 1)\n    p_code = active_codes / active_codes.sum()\n    reset_idxs = jax.random.choice(rng, num_codes, shape=(num_codes,), p=p_code)\n    do_reset = action_last_active >= args.vq_reset_thresh\n    new_codebook = jnp.where(\n        jnp.expand_dims(do_reset, -1), codebook[reset_idxs], codebook.value\n    )\n    lam.vq.codebook.value = new_codebook\n    action_last_active = jnp.where(do_reset, 0, action_last_active)\n    return loss, recon, action_last_active, 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.key(args.seed)\n\n    # --- Initialize model ---\n    rng, _rng = jax.random.split(rng)\n    rngs = nnx.Rngs(_rng)\n    lam = LatentActionModel(\n        in_dim=args.image_channels,\n        model_dim=args.model_dim,\n        ffn_dim=args.ffn_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        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n        use_flash_attention=args.use_flash_attention,\n        rngs=rngs,\n    )\n\n    # Count parameters\n    _, params, _ = nnx.split(lam, 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    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.dtype,\n    )\n    optimizer = nnx.Optimizer(lam, 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    videos_sharding = NamedSharding(mesh, PartitionSpec(""data"", None, None, None, None))\n\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    # --- Initialize checkpoint manager ---\n    step = 0\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        ""dataloader_state"",\n        grain.checkpoint.CheckpointSave,\n        cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n    )\n    handler_registry.add(\n        ""dataloader_state"",\n        grain.checkpoint.CheckpointRestore,\n        cast(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\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    image_shape = (args.image_height, args.image_width, args.image_channels)\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        abstract_optimizer = nnx.eval_shape(lambda: optimizer)\n        abstract_optimizer_state = nnx.state(abstract_optimizer)\n        restored = checkpoint_manager.restore(\n            checkpoint_manager.latest_step(),\n            args=ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                dataloader_state=grain.checkpoint.CheckpointRestore(grain_iterator),  # type: ignore\n            ),\n        )\n        restored_optimizer_state = restored[""model_state""]\n        nnx.update(optimizer, restored_optimizer_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\n    # --- TRAIN LOOP ---\n    dataloader = (\n        jax.make_array_from_process_local_data(videos_sharding, elem)\n        for elem in grain_iterator\n    )\n    print(f""Starting training from step {step}..."")\n    action_last_active = jnp.zeros(args.num_latents, dtype=jnp.int32)\n    while step < args.num_steps:\n        for videos in dataloader:\n            # --- Train step ---\n            rng, _rng = jax.random.split(rng)\n\n            inputs = dict(videos=videos, rng=_rng)\n            rng, _rng = jax.random.split(rng)\n            loss, recon, action_last_active, metrics = train_step(\n                lam, optimizer, inputs, action_last_active, _rng\n            )\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, 1:].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                    # 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            # --- Checkpointing ---\n            if args.save_ckpt and step % args.log_checkpoint_interval == 0:\n                optimizer_state = nnx.state(optimizer)\n                checkpoint_manager.save(\n                    step,\n                    args=ocp.args.Composite(\n                        model_state=ocp.args.PyTreeSave(optimizer_state),  # type: ignore\n                        dataloader_state=grain.checkpoint.CheckpointSave(  # type: ignore\n                            grain_iterator  # type: ignore\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
+10,188073,"train_lam.py",583,0,"jax.config.update(""jax_transfer_guard"", ""disallow"")\n",python,content
+11,193582,"train_dynamics.py",0,0,"from dataclasses import dataclass, field\nimport os\nfrom typing import cast\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.lr_utils import get_lr_schedule\nfrom utils.parameter_utils import count_parameters_by_component\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 = 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_latent_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    # 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(\n    model: Genie, inputs: dict\n) -> tuple[jax.Array, tuple[jax.Array, dict]]:\n    """"""Compute masked dynamics loss""""""\n    # gt = jnp.asarray(inputs[""videos""], dtype=jnp.float32) / 255.0\n    # inputs[""videos""] = gt.astype(args.dtype)\n    model.train()\n    outputs = model(inputs, training=True)\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 = 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, recon)).mean()\n    # ssim = jnp.asarray(pix.ssim(gt, recon)).mean()\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=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_lam=codebook_usage_lam,\n        codebook_usage_tokenizer=codebook_usage_tokenizer,\n    )\n    return ce_loss, (None, metrics)\n\n\n@nnx.jit\ndef train_step(\n    model: Genie, optimizer: nnx.Optimizer, inputs: dict\n) -> tuple[jax.Array, jax.Array, dict]:\n    """"""Update state and compute metrics""""""\n\n    def loss_fn(model: Genie) -> tuple[jax.Array, tuple[jax.Array, dict]]:\n        return dynamics_loss_fn(model, inputs)\n\n    (loss, (recon, metrics)), grads = nnx.value_and_grad(loss_fn, has_aux=True)(model)\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\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.key(args.seed)\n\n    # --- Initialize model ---\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_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_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\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    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.dtype,\n    )\n    optimizer = nnx.Optimizer(genie, tx)\n    del genie\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    videos_sharding = NamedSharding(mesh, PartitionSpec(""data"", None, None, None, None))\n\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    # --- Initialize checkpoint manager ---\n    step = 0\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        ""dataloader_state"",\n        grain.checkpoint.CheckpointSave,\n        cast(ocp.handlers.CheckpointHandler, grain.checkpoint.CheckpointHandler),\n    )\n    handler_registry.add(\n        ""dataloader_state"",\n        grain.checkpoint.CheckpointRestore,\n        cast(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\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    image_shape = (args.image_height, args.image_width, args.image_channels)\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        abstract_optimizer = nnx.eval_shape(lambda: optimizer)\n        abstract_optimizer_state = nnx.state(abstract_optimizer)\n        restored = checkpoint_manager.restore(\n            checkpoint_manager.latest_step(),\n            args=ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n                dataloader_state=grain.checkpoint.CheckpointRestore(grain_iterator),  # type: ignore\n            ),\n        )\n        restored_optimizer_state = restored[""model_state""]\n        nnx.update(optimizer, restored_optimizer_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        optimizer = restore_genie_components(optimizer, replicated_sharding, rng, args)\n        # NOTE: We have to remove the (unused) tokenizer vq dropout due flax.nnx lazily initializing modules.\n        # Specifically, the first dynamics model checkpoint will contain the vq dropout module,\n        # but the first full restore will fail due to nnx not initializing the module when\n        # dropout is set to 0.0.\n        del optimizer.model.tokenizer.vq.drop\n\n    # --- TRAIN LOOP ---\n    # dataloader = (\n    #     jax.make_array_from_process_local_data(videos_sharding, elem)\n    #     for elem in grain_iterator\n    # )\n    # jax.config.update(""jax_transfer_guard"", ""disallow"")\n    print(f""Starting training from step {step}..."")\n    should_break = False\n    while step < args.num_steps and not should_break:\n        for _ in range(50):\n            # --- Train step ---\n            rng, _rng_mask = jax.random.split(rng, 2)\n            inputs = dict(mask_rng=_rng_mask)\n            loss, recon, metrics = train_step(optimizer.model, optimizer, 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                    pass\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                optimizer_state = nnx.state(optimizer)\n                checkpoint_manager.save(\n                    step,\n                    args=ocp.args.Composite(\n                        model_state=ocp.args.PyTreeSave(optimizer_state),  # type: ignore\n                        dataloader_state=grain.checkpoint.CheckpointSave(  # type: ignore\n                            grain_iterator  # type: ignore\n                        ),\n                    ),\n                )\n                print(f""Saved checkpoint at step {step}"")\n            if step >= 50:\n                should_break = True\n                break\n\n    checkpoint_manager.close()\n",python,tab
+12,217408,"train_dynamics.py",8607,0,"",python,selection_command
+13,218226,"train_dynamics.py",9308,0,"",python,selection_command
+14,219567,"train_dynamics.py",14859,0,"",python,selection_command
+15,221511,"train_dynamics.py",1129,0,"",python,selection_command
+16,221808,"train_dynamics.py",2276,0,"",python,selection_command
+17,222449,"train_dynamics.py",2338,0,"",python,selection_command
+18,222887,"train_dynamics.py",12432,0,"",python,selection_command
+19,224613,"train_dynamics.py",12433,0,"",python,selection_command
+20,224628,"train_dynamics.py",12433,1,"",python,content
+21,224759,"train_dynamics.py",14764,0,"",python,selection_command
+22,225788,"train_dynamics.py",14765,0,"",python,selection_command
+23,225792,"train_dynamics.py",14765,1,"",python,content

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-bc0084f0-c97c-4490-943b-c621798113041767623320849-2026_01_05-15.28.50.314/source.csv

@@ -0,0 +1,152 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,497,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"3:28:50 PM [info] Activating crowd-code\n3:28:50 PM [info] Recording started\n3:28:50 PM [info] Initializing git provider using file system watchers...\n3:28:50 PM [info] No workspace folder found\n",Log,tab
+3,2216,"Untitled-1",0,0,"",plaintext,tab
+4,4694,"Untitled-1",0,0,"d",plaintext,content
+5,4698,"Untitled-1",1,0,"",plaintext,selection_keyboard
+6,4828,"Untitled-1",1,0,"e",plaintext,content
+7,4830,"Untitled-1",2,0,"",plaintext,selection_keyboard
+8,4934,"Untitled-1",2,0,"f",plaintext,content
+9,4937,"Untitled-1",3,0,"",plaintext,selection_keyboard
+10,5019,"Untitled-1",3,0," ",plaintext,content
+11,5022,"Untitled-1",4,0,"",plaintext,selection_keyboard
+12,5038,"Untitled-1",4,0,"h",plaintext,content
+13,5040,"Untitled-1",5,0,"",plaintext,selection_keyboard
+14,5125,"Untitled-1",5,0,"e",plaintext,content
+15,5127,"Untitled-1",6,0,"",plaintext,selection_keyboard
+16,5290,"Untitled-1",6,0,"l",plaintext,content
+17,5292,"Untitled-1",7,0,"",plaintext,selection_keyboard
+18,5442,"Untitled-1",7,0,"l",plaintext,content
+19,5444,"Untitled-1",8,0,"",plaintext,selection_keyboard
+20,5559,"Untitled-1",8,0,"o",plaintext,content
+21,5561,"Untitled-1",9,0,"",plaintext,selection_keyboard
+22,7068,"Untitled-1",9,0,"_",plaintext,content
+23,7071,"Untitled-1",10,0,"",plaintext,selection_keyboard
+24,10710,"Untitled-1",0,10,"def hello_world\n",plaintext,content
+25,11243,"Untitled-1",15,1,"",plaintext,content
+26,11684,"Untitled-1",15,0,"()",plaintext,content
+27,11686,"Untitled-1",16,0,"",plaintext,selection_keyboard
+28,22387,"Untitled-1",17,0,"",plaintext,selection_command
+29,28157,"Untitled-1",17,0,":",plaintext,content
+30,28159,"Untitled-1",18,0,"",plaintext,selection_keyboard
+31,29850,"Untitled-1",18,0,"\n",plaintext,content
+32,32053,"Untitled-1",19,0," ",plaintext,content
+33,32055,"Untitled-1",20,0,"",plaintext,selection_keyboard
+34,32112,"Untitled-1",19,1,"",plaintext,content
+35,33793,"Untitled-1",18,1,"",plaintext,content
+36,33976,"Untitled-1",18,0,"\n",plaintext,content
+37,41115,"Untitled-1",18,1,"",plaintext,content
+38,41431,"Untitled-1",18,0,"\n",plaintext,content
+39,41912,"Untitled-1",0,19,"def hello_world():\n",plaintext,content
+40,42885,"Untitled-1",18,1,"",plaintext,content
+41,43756,"Untitled-1",18,0,"\n",plaintext,content
+42,44563,"Untitled-1",19,0,"    ",plaintext,content
+43,47090,"Untitled-1",23,0,"p",plaintext,content
+44,47092,"Untitled-1",24,0,"",plaintext,selection_keyboard
+45,50396,"Untitled-1",19,5,"    print(""Hello, World!"")\n",plaintext,content
+46,54237,"Untitled-1",19,0,"",plaintext,selection_command
+47,55567,"Untitled-1",23,0,"",plaintext,selection_command
+48,55739,"Untitled-1",28,0,"",plaintext,selection_command
+49,55984,"Untitled-1",30,0,"",plaintext,selection_command
+50,56626,"Untitled-1",30,13,"",plaintext,content
+51,62279,"Untitled-1",19,14,"    print(""Hello, World!"")\n",plaintext,content
+52,63932,"Untitled-1",29,0,"",plaintext,selection_command
+53,64876,"Untitled-1",10,0,"",plaintext,selection_command
+54,65436,"Untitled-1",0,18,"def hello_world():",plaintext,selection_command
+55,65502,"Untitled-1",0,45,"def hello_world():\n    print(""Hello, World!"")",plaintext,selection_command
+56,65765,"Untitled-1",0,46,"def hello_world():\n    print(""Hello, World!"")\n",plaintext,selection_command
+57,66029,"Untitled-1",0,46,"",plaintext,content
+58,66630,"Untitled-1",0,0,"d",plaintext,content
+59,66632,"Untitled-1",1,0,"",plaintext,selection_keyboard
+60,66682,"Untitled-1",1,0,"e",plaintext,content
+61,66683,"Untitled-1",2,0,"",plaintext,selection_keyboard
+62,66872,"Untitled-1",2,0,"f",plaintext,content
+63,66874,"Untitled-1",3,0,"",plaintext,selection_keyboard
+64,66941,"Untitled-1",3,0," ",plaintext,content
+65,66943,"Untitled-1",4,0,"",plaintext,selection_keyboard
+66,67027,"Untitled-1",4,0,"f",plaintext,content
+67,67029,"Untitled-1",5,0,"",plaintext,selection_keyboard
+68,67036,"Untitled-1",5,0,"i",plaintext,content
+69,67037,"Untitled-1",6,0,"",plaintext,selection_keyboard
+70,67127,"Untitled-1",6,0,"b",plaintext,content
+71,67128,"Untitled-1",7,0,"",plaintext,selection_keyboard
+72,67297,"Untitled-1",7,0,"o",plaintext,content
+73,67299,"Untitled-1",8,0,"",plaintext,selection_keyboard
+74,68556,"Untitled-1",8,0,"n",plaintext,content
+75,68558,"Untitled-1",9,0,"",plaintext,selection_keyboard
+76,68699,"Untitled-1",9,0,"n",plaintext,content
+77,68701,"Untitled-1",10,0,"",plaintext,selection_keyboard
+78,68913,"Untitled-1",10,0,"a",plaintext,content
+79,68914,"Untitled-1",11,0,"",plaintext,selection_keyboard
+80,69135,"Untitled-1",11,0,"c",plaintext,content
+81,69137,"Untitled-1",12,0,"",plaintext,selection_keyboard
+82,70597,"Untitled-1",0,12,"def fibonnaci\n",plaintext,content
+83,71132,"Untitled-1",13,1,"",plaintext,content
+84,71638,"Untitled-1",13,0,"()",plaintext,content
+85,71640,"Untitled-1",14,0,"",plaintext,selection_keyboard
+86,155245,"Untitled-1",0,15,"def fibonnaci(n):\n",plaintext,content
+87,156484,"Untitled-1",15,0,"",plaintext,selection_command
+88,156581,"Untitled-1",16,0,"",plaintext,selection_command
+89,156764,"Untitled-1",17,0,"",plaintext,selection_command
+90,156996,"Untitled-1",17,0,"\n",plaintext,content
+91,158165,"Untitled-1",18,1,"    if n <= 1:\n",plaintext,content
+92,160225,"Untitled-1",33,0,"",plaintext,selection_command
+93,161697,"Untitled-1",33,0,"        return n\n",plaintext,content
+94,162835,"Untitled-1",50,0,"    return fibonnaci(n-1) + fibonnaci(n-2)\n",plaintext,content
+95,165097,"Untitled-1",92,1,"",plaintext,content
+96,165553,"Untitled-1",50,42,"",plaintext,content
+97,166147,"Untitled-1",33,17,"        return n\n",plaintext,content
+98,167346,"Untitled-1",50,0,"    else:\n",plaintext,content
+99,168289,"Untitled-1",60,0,"        return fibonnaci(n-1) + fibonnaci(n-2)\n",plaintext,content
+100,169550,"Untitled-1",60,0,"",plaintext,selection_command
+101,169694,"Untitled-1",50,0,"",plaintext,selection_command
+102,171090,"Untitled-1",54,0,"s",plaintext,content
+103,171092,"Untitled-1",55,0,"",plaintext,selection_keyboard
+104,171936,"Untitled-1",54,1,"",plaintext,content
+105,176069,"Untitled-1",58,0,"j",plaintext,content
+106,176071,"Untitled-1",59,0,"",plaintext,selection_keyboard
+107,176157,"Untitled-1",59,0,"f",plaintext,content
+108,176159,"Untitled-1",60,0,"",plaintext,selection_keyboard
+109,176181,"Untitled-1",60,0,"d",plaintext,content
+110,176182,"Untitled-1",61,0,"",plaintext,selection_keyboard
+111,176287,"Untitled-1",61,0,"k",plaintext,content
+112,176289,"Untitled-1",62,0,"",plaintext,selection_keyboard
+113,176390,"Untitled-1",62,0,"d",plaintext,content
+114,176391,"Untitled-1",63,0,"",plaintext,selection_keyboard
+115,176456,"Untitled-1",63,0,"k",plaintext,content
+116,176458,"Untitled-1",64,0,"",plaintext,selection_keyboard
+117,176563,"Untitled-1",64,0,"d",plaintext,content
+118,176565,"Untitled-1",65,0,"",plaintext,selection_keyboard
+119,177638,"Untitled-1",50,17,"    else:\n",plaintext,content
+120,181280,"Untitled-1",61,0,"",plaintext,selection_command
+121,181427,"Untitled-1",62,0,"",plaintext,selection_command
+122,181646,"Untitled-1",63,0,"",plaintext,selection_command
+123,182877,"Untitled-1",64,0,"",plaintext,selection_command
+124,183046,"Untitled-1",65,0,"",plaintext,selection_command
+125,183851,"Untitled-1",106,0,"\n    else:",plaintext,content
+126,183852,"Untitled-1",50,10,"",plaintext,content
+127,184213,"Untitled-1",96,0,"\n        return n",plaintext,content
+128,184214,"Untitled-1",33,17,"",plaintext,content
+129,185053,"Untitled-1",79,17,"",plaintext,content
+130,185054,"Untitled-1",33,0,"        return n\n",plaintext,content
+131,185188,"Untitled-1",96,10,"",plaintext,content
+132,185189,"Untitled-1",50,0,"    else:\n",plaintext,content
+133,185866,"Untitled-1",55,0,"",plaintext,selection_command
+134,185942,"Untitled-1",38,0,"",plaintext,selection_command
+135,186092,"Untitled-1",23,0,"",plaintext,selection_command
+136,190296,"Untitled-1",25,1,"",plaintext,content
+137,190559,"Untitled-1",25,0,"f",plaintext,content
+138,190560,"Untitled-1",26,0,"",plaintext,selection_keyboard
+139,190960,"Untitled-1",26,0,"f",plaintext,content
+140,190962,"Untitled-1",27,0,"",plaintext,selection_keyboard
+141,191033,"Untitled-1",27,0,"j",plaintext,content
+142,191034,"Untitled-1",28,0,"",plaintext,selection_keyboard
+143,191115,"Untitled-1",28,0,"s",plaintext,content
+144,191117,"Untitled-1",29,0,"",plaintext,selection_keyboard
+145,191143,"Untitled-1",29,0,"d",plaintext,content
+146,191144,"Untitled-1",30,0,"",plaintext,selection_keyboard
+147,191246,"Untitled-1",30,0,"f",plaintext,content
+148,191248,"Untitled-1",31,0,"",plaintext,selection_keyboard
+149,193530,"Untitled-1",18,20,"    if n <= 1:\n",plaintext,content
+150,196399,"Untitled-1",0,18,"def fibonnaci\n",plaintext,content
+151,198343,"Untitled-1",0,14,"def fibonnaci(n):\n",plaintext,content
+152,200616,"Untitled-1",30,0,"",plaintext,selection_command

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-e152ee4a-939b-4464-8ec7-3739b32281d61763456606729-2025_11_18-10.03.33.832/source.csv

@@ -0,0 +1,53 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,3,"slurm/dev/franz/berlin/crowd-pilot/nemo/test_sft.py",0,0,"import nemo_run as run\nfrom nemo.collections import llm\n\nNAME = ""qwen3_30b_a3b""\n\nrecipe = llm.qwen3_30b_a3b.finetune_recipe(\n    name=NAME,\n    dir=f""test_output"",\n    num_nodes=1,\n    num_gpus_per_node=8,\n    peft_scheme='lora',  # 'lora', 'none'\n    packed_sequence=True,\n)\n\ndef local_executor_torchrun(nodes: int, devices: int) -> run.LocalExecutor:\n    # Env vars for jobs are configured here\n    env_vars = {\n        ""TORCH_NCCL_AVOID_RECORD_STREAMS"": ""1"",\n        ""NCCL_NVLS_ENABLE"": ""0"",\n        ""NVTE_DP_AMAX_REDUCE_INTERVAL"": ""0"",\n        ""NVTE_ASYNC_AMAX_REDUCTION"": ""1"",\n        #""LD_PRELOAD"": ""/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nvidia/cuda_runtime/lib/libcudart.so.12"",\n    }\n\n    executor = run.LocalExecutor(ntasks_per_node=devices, launcher=""torchrun"", env_vars=env_vars)\n\n    return executor\n\ndef run_finetuning():\n    recipe.resume.restore_config.path = ""/fast/project/HFMI_SynergyUnit/tab_model/data/checkpoints/nemo_converted_weights_qwen3-coder-30b-a3b-instruct/""\n    recipe.data.delete_raw = False\n    executor = local_executor_torchrun(nodes=recipe.trainer.num_nodes, devices=recipe.trainer.devices)\n\n    run.run(recipe, executor=executor, name=NAME)\n\n# This condition is necessary for the script to be compatible with Python's multiprocessing module.\nif __name__ == ""__main__"":\n    run_finetuning()",python,tab
+2,81,"TERMINAL",0,0,"",,terminal_focus
+3,82,"TERMINAL",0,0,"",,terminal_focus
+4,306,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"10:03:33 AM [info] Activating crowd-code\n10:03:33 AM [info] Recording started\n10:03:33 AM [info] Initializing git provider using file system watchers...\n",Log,tab
+5,462,"extension-output-pdoom-org.crowd-code-#1-crowd-code",153,0,"10:03:34 AM [info] Git repository found\n10:03:34 AM [info] Git provider initialized successfully\n10:03:34 AM [info] Initial git state: [object Object]\n",Log,content
+6,544,"TERMINAL",0,0,"bash",,terminal_focus
+7,545,"slurm/dev/franz/berlin/crowd-pilot/nemo/test_sft.py",0,0,"",python,tab
+8,545,"TERMINAL",0,0,"bash",,terminal_focus
+9,623,"TERMINAL",0,0,"source /home/franz.srambical/crowd-pilot/.nemo/bin/activate",,terminal_command
+10,623,"TERMINAL",0,0,"source /home/franz.srambical/crowd-pilot/.nemo/bin/activate",,terminal_command
+11,624,"TERMINAL",0,0,"]633;C]0;franz.srambical@hai-login2:~/crowd-pilot",,terminal_output
+12,624,"TERMINAL",0,0,"]633;C]0;franz.srambical@hai-login2:~/crowd-pilot",,terminal_output
+13,207295,"TERMINAL",0,0,"squeue",,terminal_command
+14,207312,"TERMINAL",0,0,"]633;C  JOBID       USER PARTITION NODES  CPUS ST         SUBMIT_TIME          START_TIME     TIME TIME_LIMIT NODELIST(REASON)\r\n  34028   xiao.liu interacti     1   128  R 2025-11-17T23:53:51 2025-11-17T23:53:51 10:13:10   23:59:00 hai005\r\n  34027 franz.sram interacti     1    20  R 2025-11-17T22:06:31 2025-11-17T22:06:31 12:00:30 1-00:00:00 hai001\r\n  34026   xiao.liu interacti     1   128  R 2025-11-17T20:01:59 2025-11-17T20:01:59 14:05:02   23:59:00 hai006\r\n  34037 kalyan.nad  standard     1    64  R 2025-11-18T08:55:56 2025-11-18T08:55:56  1:11:05    8:00:00 hai002\r\n  34032   xiao.liu  standard     1   128  R 2025-11-18T03:50:45 2025-11-18T05:06:30  5:00:31   23:59:00 hai007\r\n  33959   xiao.liu  standard     1   128  R 2025-11-17T10:39:54 2025-11-17T10:40:37 23:26:24   23:59:00 hai004\r\n]0;franz.srambical@hai-login2:~/crowd-pilot",,terminal_output
+15,216561,"TERMINAL",0,0,"scancel 34027",,terminal_command
+16,216570,"TERMINAL",0,0,"]633;C]0;franz.srambical@hai-login2:~/crowd-pilot",,terminal_output
+17,230003,"TERMINAL",0,0,"salloc --gpus=8 --ntasks-per-node=1 --cpus-per-task=10 --mem=100G",,terminal_command
+18,230054,"TERMINAL",0,0,"]633;Csalloc: Pending job allocation 34038\r\nsalloc: job 34038 queued and waiting for resources\r\n",,terminal_output
+19,234099,"TERMINAL",0,0,"^Csalloc: Job aborted due to signal\r\nsalloc: Job allocation 34038 has been revoked.\r\n]0;franz.srambical@hai-login2:~/crowd-pilot",,terminal_output
+20,236703,"TERMINAL",0,0,"salloc --gpus=8 --ntasks-per-node=1 --cpus-per-task=10 --mem=100G --qos=low",,terminal_command
+21,236754,"TERMINAL",0,0,"]633;Csalloc: Granted job allocation 34039\r\n",,terminal_output
+22,236853,"TERMINAL",0,0,"salloc: Nodes hai003 are ready for job\r\n",,terminal_output
+23,237197,"TERMINAL",0,0,"Running inside SLURM, Job ID 34039.\r\n",,terminal_output
+24,237282,"TERMINAL",0,0,"]0;franz.srambical@hai-login2:~/crowd-pilot[?2004h[franz.srambical@hai003.haicore.berlin:~/crowd-pilot] $ ",,terminal_output
+25,240701,"TERMINAL",0,0,"\r(reverse-i-search)`': ",,terminal_output
+26,240882,"TERMINAL",0,0,"s': python /home/franz.srambical/slurm/dev/franz/berlin/crowd-pilot/nemo/test_sft.py",,terminal_output
+27,241023,"TERMINAL",0,0,"\ro': readelf -d $(python -c ""import functorch; import inspect, os; print(os.path.join(os.path.dirname(inspect.getfile(functorch)), '_C*.so'))"") | grep RUNPATH\ru': source /home/franz.srambical/crowd-pilot/.nemo/bin/activate\r",,terminal_output
+28,241225,"TERMINAL",0,0,"[1@r': sour",,terminal_output
+29,241960,"TERMINAL",0,0,"\r[30@[franz.srambical@hai003.haicore.berlin:~/crowd-pilot] $ sour\r\n[?2004l\r]0;franz.srambical@hai-login2:~/crowd-pilot[?2004h(.nemo) [franz.srambical@hai003.haicore.berlin:~/crowd-pilot] $ ",,terminal_output
+30,242463,"TERMINAL",0,0,"\r(reverse-i-search)`': ",,terminal_output
+31,242681,"TERMINAL",0,0,"p': source /home/franz.srambical/crowd-pilot/.nemo/bin/activate",,terminal_output
+32,242762,"TERMINAL",0,0,"\ry': python /home/franz.srambical/slurm/dev/franz/berlin/crowd-pilot/nemo/test_sft.py",,terminal_output
+33,242874,"TERMINAL",0,0,"\rt': python /home/franz.srambical/slurm/dev/franz/berlin/crowd-pilot/nemo/test_sft.py\r",,terminal_output
+34,243012,"TERMINAL",0,0,"[1@h': pyth",,terminal_output
+35,243077,"TERMINAL",0,0,"[1@o': pytho",,terminal_output
+36,243213,"TERMINAL",0,0,"[1@n': python",,terminal_output
+37,243662,"TERMINAL",0,0,"\r[36@.nemo) [franz.srambical@hai003.haicore.berlin:~/crowd-pilot] $ python",,terminal_output
+38,243943,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+39,252109,"TERMINAL",0,0,"/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\n  import pynvml  # type: ignore[import]\r\n",,terminal_output
+40,276475,"TERMINAL",0,0,"OneLogger: Setting error_handling_strategy to DISABLE_QUIETLY_AND_REPORT_METRIC_ERROR for rank (rank=0) with OneLogger disabled. To override: explicitly set error_handling_strategy parameter.\r\n",,terminal_output
+41,276578,"TERMINAL",0,0,"No exporters were provided. This means that no telemetry data will be collected.\r\n",,terminal_output
+42,284004,"TERMINAL",0,0,"───────────────────────────────────────────────────────────── Entering Experiment qwen3_30b_a3b with id: qwen3_30b_a3b_1763456897 ──────────────────────────────────────────────────────────────\r\n",,terminal_output
+43,284254,"TERMINAL",0,0,"[10:08:17] INFO     Log directory is: /home/franz.srambical/.nemo_run/experiments/qwen3_30b_a3b/qwen3_30b_a3b_1763456897/qwen3_30b_a3b                                    ]8;id=991907;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torchx/schedulers/local_scheduler.py\local_scheduler.py]8;;\:]8;id=464480;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torchx/schedulers/local_scheduler.py#777\777]8;;\\r\n[10:08:17] Launching job qwen3_30b_a3b for experiment qwen3_30b_a3b                                                                                                            ]8;id=915722;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nemo_run/run/experiment.py\experiment.py]8;;\:]8;id=538525;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nemo_run/run/experiment.py#795\795]8;;\\r\n           INFO     Launched app: local_persistent://nemo_run/qwen3_30b_a3b-k2spc72nk5x9lc                                                                                       ]8;id=246651;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nemo_run/run/torchx_backend/launcher.py\launcher.py]8;;\:]8;id=968411;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nemo_run/run/torchx_backend/launcher.py#116\116]8;;\\r\n──────────────────��─────────────────────────────────────────────── Waiting for Experiment qwen3_30b_a3b_1763456897 to finish ───────────────────────────────────────────────────────────────────\r\n\r\nExperiment Status for qwen3_30b_a3b_1763456897\r\n\r\nTask 0: qwen3_30b_a3b\r\n- Status: RUNNING\r\n- Executor: LocalExecutor\r\n- Job id: qwen3_30b_a3b-k2spc72nk5x9lc\r\n- Local Directory: /home/franz.srambical/.nemo_run/experiments/qwen3_30b_a3b/qwen3_30b_a3b_1763456897/qwen3_30b_a3b\r\n\r\n[10:08:18] INFO     Waiting for job qwen3_30b_a3b-k2spc72nk5x9lc to finish [log=True]...                                                                                         ]8;id=911442;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nemo_run/run/torchx_backend/launcher.py\launcher.py]8;;\:]8;id=91408;file:///fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/nemo_run/run/torchx_backend/launcher.py#136\136]8;;\\r\n",,terminal_output
+44,285105,"TERMINAL",0,0,"n3_30b_a3b/0 /fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0   import pynvml  # type: ignore[import]\r\n",,terminal_output
+45,286704,"TERMINAL",0,0,"n3_30b_a3b/0 I1118 10:08:20.315000 503305 torch/distributed/run.py:657] Using nproc_per_node=8.\r\nn3_30b_a3b/0 W1118 10:08:20.316000 503305 torch/distributed/run.py:774] \r\nn3_30b_a3b/0 W1118 10:08:20.316000 503305 torch/distributed/run.py:774] *****************************************\r\nn3_30b_a3b/0 W1118 10:08:20.316000 503305 torch/distributed/run.py:774] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. \r\nn3_30b_a3b/0 W1118 10:08:20.316000 503305 torch/distributed/run.py:774] *****************************************\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199] Starting elastic_operator with launch configs:\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   entrypoint         : nemo_run.core.runners.fdl_runner\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   min_nodes          : 1\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   max_nodes          : 1\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   nproc_per_node     : 8\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   run_id             : 2707\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   rdzv_backend       : c10d\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   rdzv_endpoint      : localhost:0\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   rdzv_configs       : {'timeout': 900}\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   max_restarts       : 0\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   monitor_interval   : 0.1\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   log_dir            : /home/franz.srambical/.nemo_run/experiments/qwen3_30b_a3b/qwen3_30b_a3b_1763456897/qwen3_30b_a3b/nemo_run/qwen3_30b_a3b-k2spc72nk5x9lc/torchelastic/qwen3_30b_a3b\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   metrics_cfg        : {}\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199]   event_log_handler  : null\r\nn3_30b_a3b/0 I1118 10:08:20.319000 503305 torch/distributed/launcher/api.py:199] \r\nn3_30b_a3b/0 I1118 10:08:20.327000 503305 torch/distributed/elastic/agent/server/api.py:869] [default] starting workers for entrypoint: python\r\nn3_30b_a3b/0 I1118 10:08:20.328000 503305 torch/distributed/elastic/agent/server/api.py:677] [default] Rendezvous'ing worker group\r\n",,terminal_output
+46,286999,"TERMINAL",0,0,"n3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523] [default] Rendezvous complete for workers. Result:\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   restart_count=0\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   master_addr=hai003.haicore.berlin\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   master_port=45071\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   group_rank=0\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   group_world_size=1\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   local_ranks=[0, 1, 2, 3, 4, 5, 6, 7]\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   role_ranks=[0, 1, 2, 3, 4, 5, 6, 7]\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   global_ranks=[0, 1, 2, 3, 4, 5, 6, 7]\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   role_world_sizes=[8, 8, 8, 8, 8, 8, 8, 8]\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   global_world_sizes=[8, 8, 8, 8, 8, 8, 8, 8]\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523]   event_log_handler=null\r\nn3_30b_a3b/0 I1118 10:08:20.587000 503305 torch/distributed/elastic/agent/server/api.py:523] \r\nn3_30b_a3b/0 I1118 10:08:20.588000 503305 torch/distributed/elastic/agent/server/api.py:685] [default] Starting worker group\r\nn3_30b_a3b/0 I1118 10:08:20.589000 503305 torch/distributed/elastic/agent/server/local_elastic_agent.py:296] use_agent_store: True\r\nn3_30b_a3b/0 I1118 10:08:20.589000 503305 torch/distributed/elastic/agent/server/local_elastic_agent.py:192] Environment variable 'TORCHELASTIC_ENABLE_FILE_TIMER' not found. Do not start FileTimerServer.\r\nn3_30b_a3b/0 I1118 10:08:20.590000 503305 torch/distributed/elastic/agent/server/local_elastic_agent.py:236] Environment variable 'TORCHELASTIC_HEALTH_CHECK_PORT' not found. Do not start health check.\r\n",,terminal_output
+47,290775,"TERMINAL",0,0,"n3_30b_a3b/0 [default6]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default6]:  import pynvml  # type: ignore[import]\r\n",,terminal_output
+48,296737,"TERMINAL",0,0,"n3_30b_a3b/0 [default1]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default1]:  import pynvml  # type: ignore[import]\r\n",,terminal_output
+49,297159,"TERMINAL",0,0,"n3_30b_a3b/0 [default3]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default3]:  import pynvml  # type: ignore[import]\r\nn3_30b_a3b/0 [default0]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default0]:  import pynvml  # type: ignore[import]\r\n",,terminal_output
+50,297577,"TERMINAL",0,0,"n3_30b_a3b/0 [default4]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default4]:  import pynvml  # type: ignore[import]\r\n",,terminal_output
+51,297891,"TERMINAL",0,0,"n3_30b_a3b/0 [default5]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default5]:  import pynvml  # type: ignore[import]\r\n",,terminal_output
+52,298212,"TERMINAL",0,0,"n3_30b_a3b/0 [default2]:/fast/home/franz.srambical/crowd-pilot/.nemo/lib/python3.10/site-packages/torch/cuda/__init__.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.\r\nn3_30b_a3b/0 [default2]:  import pynvml  # type: ignore[import]\r\n",,terminal_output