ml-dash 0.6.1__py3-none-any.whl

ml_dash/cli_commands/upload.py

@@ -0,0 +1,1248 @@
+"""Upload command implementation for ML-Dash CLI."""
+
+import argparse
+import json
+from pathlib import Path
+from typing import List, Dict, Any, Optional
+from dataclasses import dataclass, field
+import threading
+from concurrent.futures import ThreadPoolExecutor, as_completed
+
+from rich.console import Console
+from rich.progress import Progress, SpinnerColumn, TextColumn, BarColumn, TaskProgressColumn
+from rich.table import Table
+from rich.panel import Panel
+
+from ..storage import LocalStorage
+from ..client import RemoteClient
+from ..config import Config
+
+# Initialize rich console
+console = Console()
+
+
+@dataclass
+class ExperimentInfo:
+    """Information about an experiment to upload."""
+    project: str
+    experiment: str
+    path: Path
+    folder: Optional[str] = None
+    has_logs: bool = False
+    has_params: bool = False
+    metric_names: List[str] = field(default_factory=list)
+    file_count: int = 0
+    estimated_size: int = 0  # in bytes
+
+
+@dataclass
+class ValidationResult:
+    """Result of experiment validation."""
+    is_valid: bool = True
+    warnings: List[str] = field(default_factory=list)
+    errors: List[str] = field(default_factory=list)
+    valid_data: Dict[str, Any] = field(default_factory=dict)
+
+
+@dataclass
+class UploadResult:
+    """Result of uploading an experiment."""
+    experiment: str
+    success: bool = False
+    uploaded: Dict[str, int] = field(default_factory=dict)  # {"logs": 100, "metrics": 3}
+    failed: Dict[str, List[str]] = field(default_factory=dict)  # {"files": ["error msg"]}
+    errors: List[str] = field(default_factory=list)
+    bytes_uploaded: int = 0  # Total bytes uploaded
+
+
+@dataclass
+class UploadState:
+    """Tracks upload state for resume functionality."""
+    local_path: str
+    remote_url: str
+    completed_experiments: List[str] = field(default_factory=list)  # ["project/experiment"]
+    failed_experiments: List[str] = field(default_factory=list)
+    in_progress_experiment: Optional[str] = None
+    timestamp: Optional[str] = None
+
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary for JSON serialization."""
+        return {
+            "local_path": self.local_path,
+            "remote_url": self.remote_url,
+            "completed_experiments": self.completed_experiments,
+            "failed_experiments": self.failed_experiments,
+            "in_progress_experiment": self.in_progress_experiment,
+            "timestamp": self.timestamp,
+        }
+
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> "UploadState":
+        """Create from dictionary."""
+        return cls(
+            local_path=data["local_path"],
+            remote_url=data["remote_url"],
+            completed_experiments=data.get("completed_experiments", []),
+            failed_experiments=data.get("failed_experiments", []),
+            in_progress_experiment=data.get("in_progress_experiment"),
+            timestamp=data.get("timestamp"),
+        )
+
+    def save(self, path: Path):
+        """Save state to file."""
+        import datetime
+        self.timestamp = datetime.datetime.now().isoformat()
+        with open(path, "w") as f:
+            json.dump(self.to_dict(), f, indent=2)
+
+    @classmethod
+    def load(cls, path: Path) -> Optional["UploadState"]:
+        """Load state from file."""
+        if not path.exists():
+            return None
+        try:
+            with open(path, "r") as f:
+                data = json.load(f)
+            return cls.from_dict(data)
+        except (json.JSONDecodeError, IOError, KeyError):
+            return None
+
+
+def add_parser(subparsers) -> argparse.ArgumentParser:
+    """Add upload command parser."""
+    parser = subparsers.add_parser(
+        "upload",
+        help="Upload local experiments to remote server",
+        description="Upload locally-stored ML-Dash experiment data to a remote server.",
+    )
+
+    # Positional argument
+    parser.add_argument(
+        "path",
+        nargs="?",
+        default="./.ml-dash",
+        help="Local storage directory to upload from (default: ./.ml-dash)",
+    )
+
+    # Remote configuration
+    parser.add_argument(
+        "--remote",
+        type=str,
+        help="Remote server URL (required unless set in config)",
+    )
+    parser.add_argument(
+        "--api-key",
+        type=str,
+        help="JWT token for authentication (optional - auto-loads from 'ml-dash login' if not provided)",
+    )
+
+    # Scope control
+    parser.add_argument(
+        "--project",
+        type=str,
+        help="Upload only experiments from this project",
+    )
+    parser.add_argument(
+        "--experiment",
+        type=str,
+        help="Upload only this specific experiment (requires --project)",
+    )
+
+    # Data filtering
+    parser.add_argument(
+        "--skip-logs",
+        action="store_true",
+        help="Don't upload logs",
+    )
+    parser.add_argument(
+        "--skip-metrics",
+        action="store_true",
+        help="Don't upload metrics",
+    )
+    parser.add_argument(
+        "--skip-files",
+        action="store_true",
+        help="Don't upload files",
+    )
+    parser.add_argument(
+        "--skip-params",
+        action="store_true",
+        help="Don't upload parameters",
+    )
+
+    # Behavior control
+    parser.add_argument(
+        "--dry-run",
+        action="store_true",
+        help="Show what would be uploaded without uploading",
+    )
+    parser.add_argument(
+        "--strict",
+        action="store_true",
+        help="Fail on any validation error (default: skip invalid data)",
+    )
+    parser.add_argument(
+        "-v", "--verbose",
+        action="store_true",
+        help="Show detailed progress",
+    )
+    parser.add_argument(
+        "--batch-size",
+        type=int,
+        default=100,
+        help="Batch size for logs/metrics (default: 100)",
+    )
+    parser.add_argument(
+        "--resume",
+        action="store_true",
+        help="Resume previous interrupted upload",
+    )
+    parser.add_argument(
+        "--state-file",
+        type=str,
+        default=".ml-dash-upload-state.json",
+        help="Path to state file for resume (default: .ml-dash-upload-state.json)",
+    )
+
+    return parser
+
+
+def discover_experiments(
+    local_path: Path,
+    project_filter: Optional[str] = None,
+    experiment_filter: Optional[str] = None,
+) -> List[ExperimentInfo]:
+    """
+    Discover experiments in local storage directory.
+
+    Supports both flat (local_path/project/experiment) and folder-based
+    (local_path/folder/project/experiment) hierarchies.
+
+    Args:
+        local_path: Root path of local storage
+        project_filter: Only discover experiments in this project
+        experiment_filter: Only discover this experiment (requires project_filter)
+
+    Returns:
+        List of ExperimentInfo objects
+    """
+    local_path = Path(local_path)
+
+    if not local_path.exists():
+        return []
+
+    experiments = []
+
+    # Find all experiment.json files recursively
+    for exp_json in local_path.rglob("*/experiment.json"):
+        exp_dir = exp_json.parent
+
+        # Extract project and experiment names from path
+        # Path structure: local_path / [folder] / project / experiment
+        try:
+            relative_path = exp_dir.relative_to(local_path)
+            parts = relative_path.parts
+
+            if len(parts) < 2:
+                continue  # Need at least project/experiment
+
+            # Last two parts are project/experiment
+            exp_name = parts[-1]
+            project_name = parts[-2]
+
+            # Apply filters
+            if project_filter and project_name != project_filter:
+                continue
+            if experiment_filter and exp_name != experiment_filter:
+                continue
+
+            # Read folder from experiment.json
+            folder = None
+            try:
+                with open(exp_json, 'r') as f:
+                    metadata = json.load(f)
+                    folder = metadata.get('folder')
+            except:
+                pass
+
+            # Create experiment info
+            exp_info = ExperimentInfo(
+                project=project_name,
+                experiment=exp_name,
+                path=exp_dir,
+                folder=folder,
+            )
+        except (ValueError, IndexError):
+            continue
+
+        # Check for parameters
+        params_file = exp_dir / "parameters.json"
+        exp_info.has_params = params_file.exists()
+
+        # Check for logs
+        logs_file = exp_dir / "logs" / "logs.jsonl"
+        exp_info.has_logs = logs_file.exists()
+
+        # Check for metrics
+        metrics_dir = exp_dir / "metrics"
+        if metrics_dir.exists():
+            for metric_dir in metrics_dir.iterdir():
+                if metric_dir.is_dir():
+                    data_file = metric_dir / "data.jsonl"
+                    if data_file.exists():
+                        exp_info.metric_names.append(metric_dir.name)
+
+        # Check for files
+        files_dir = exp_dir / "files"
+        if files_dir.exists():
+            try:
+                # Count files recursively
+                exp_info.file_count = sum(1 for _ in files_dir.rglob("*") if _.is_file())
+
+                # Estimate size
+                exp_info.estimated_size = sum(
+                    f.stat().st_size for f in files_dir.rglob("*") if f.is_file()
+                )
+            except (OSError, PermissionError):
+                pass
+
+        experiments.append(exp_info)
+
+    return experiments
+
+
+class ExperimentValidator:
+    """Validates local experiment data before upload."""
+
+    def __init__(self, strict: bool = False):
+        """
+        Initialize validator.
+
+        Args:
+            strict: If True, fail on any validation error
+        """
+        self.strict = strict
+
+    def validate_experiment(self, exp_info: ExperimentInfo) -> ValidationResult:
+        """
+        Validate experiment directory structure and data.
+
+        Args:
+            exp_info: Experiment information
+
+        Returns:
+            ValidationResult with validation status and messages
+        """
+        result = ValidationResult()
+        result.valid_data = {}
+
+        # 1. Validate experiment metadata (required)
+        if not self._validate_experiment_metadata(exp_info, result):
+            result.is_valid = False
+            return result
+
+        # 2. Validate parameters (optional)
+        self._validate_parameters(exp_info, result)
+
+        # 3. Validate logs (optional)
+        self._validate_logs(exp_info, result)
+
+        # 4. Validate metrics (optional)
+        self._validate_metrics(exp_info, result)
+
+        # 5. Validate files (optional)
+        self._validate_files(exp_info, result)
+
+        # In strict mode, any warning becomes an error
+        if self.strict and result.warnings:
+            result.errors.extend(result.warnings)
+            result.warnings = []
+            result.is_valid = False
+
+        return result
+
+    def _validate_experiment_metadata(self, exp_info: ExperimentInfo, result: ValidationResult) -> bool:
+        """Validate experiment.json exists and is valid."""
+        exp_json = exp_info.path / "experiment.json"
+
+        if not exp_json.exists():
+            result.errors.append("Missing experiment.json")
+            return False
+
+        try:
+            with open(exp_json, "r") as f:
+                metadata = json.load(f)
+
+            # Check required fields
+            if "name" not in metadata or "project" not in metadata:
+                result.errors.append("experiment.json missing required fields (name, project)")
+                return False
+
+            result.valid_data["metadata"] = metadata
+            return True
+
+        except json.JSONDecodeError as e:
+            result.errors.append(f"Invalid JSON in experiment.json: {e}")
+            return False
+        except IOError as e:
+            result.errors.append(f"Cannot read experiment.json: {e}")
+            return False
+
+    def _validate_parameters(self, exp_info: ExperimentInfo, result: ValidationResult):
+        """Validate parameters.json format."""
+        if not exp_info.has_params:
+            return
+
+        params_file = exp_info.path / "parameters.json"
+        try:
+            with open(params_file, "r") as f:
+                params = json.load(f)
+
+            # Check if it's a dict
+            if not isinstance(params, dict):
+                result.warnings.append("parameters.json is not a dict (will skip)")
+                return
+
+            # Check for valid data key if using versioned format
+            if "data" in params:
+                if not isinstance(params["data"], dict):
+                    result.warnings.append("parameters.json data is not a dict (will skip)")
+                    return
+                result.valid_data["parameters"] = params["data"]
+            else:
+                result.valid_data["parameters"] = params
+
+        except json.JSONDecodeError as e:
+            result.warnings.append(f"Invalid JSON in parameters.json: {e} (will skip)")
+        except IOError as e:
+            result.warnings.append(f"Cannot read parameters.json: {e} (will skip)")
+
+    def _validate_logs(self, exp_info: ExperimentInfo, result: ValidationResult):
+        """Validate logs.jsonl format."""
+        if not exp_info.has_logs:
+            return
+
+        logs_file = exp_info.path / "logs" / "logs.jsonl"
+        invalid_lines = []
+
+        try:
+            with open(logs_file, "r") as f:
+                for line_num, line in enumerate(f, start=1):
+                    try:
+                        log_entry = json.loads(line)
+                        # Check required fields
+                        if "message" not in log_entry:
+                            invalid_lines.append(line_num)
+                    except json.JSONDecodeError:
+                        invalid_lines.append(line_num)
+
+            if invalid_lines:
+                count = len(invalid_lines)
+                preview = invalid_lines[:5]
+                result.warnings.append(
+                    f"logs.jsonl has {count} invalid lines (e.g., {preview}...) - will skip these"
+                )
+
+        except IOError as e:
+            result.warnings.append(f"Cannot read logs.jsonl: {e} (will skip logs)")
+
+    def _validate_metrics(self, exp_info: ExperimentInfo, result: ValidationResult):
+        """Validate metrics data."""
+        if not exp_info.metric_names:
+            return
+
+        for metric_name in exp_info.metric_names:
+            metric_dir = exp_info.path / "metrics" / metric_name
+            data_file = metric_dir / "data.jsonl"
+
+            invalid_lines = []
+            try:
+                with open(data_file, "r") as f:
+                    for line_num, line in enumerate(f, start=1):
+                        try:
+                            data_point = json.loads(line)
+                            # Check for data field
+                            if "data" not in data_point:
+                                invalid_lines.append(line_num)
+                        except json.JSONDecodeError:
+                            invalid_lines.append(line_num)
+
+                if invalid_lines:
+                    count = len(invalid_lines)
+                    preview = invalid_lines[:5]
+                    result.warnings.append(
+                        f"metric '{metric_name}' has {count} invalid lines (e.g., {preview}...) - will skip these"
+                    )
+
+            except IOError as e:
+                result.warnings.append(f"Cannot read metric '{metric_name}': {e} (will skip)")
+
+    def _validate_files(self, exp_info: ExperimentInfo, result: ValidationResult):
+        """Validate files existence."""
+        files_dir = exp_info.path / "files"
+        if not files_dir.exists():
+            return
+
+        metadata_file = files_dir / ".files_metadata.json"
+        if not metadata_file.exists():
+            return
+
+        try:
+            with open(metadata_file, "r") as f:
+                files_metadata = json.load(f)
+
+            missing_files = []
+            for file_id, file_info in files_metadata.items():
+                if isinstance(file_info, dict) and file_info.get("deletedAt") is None:
+                    # Check if file exists
+                    file_path = files_dir / file_info.get("prefix", "") / file_id / file_info.get("filename", "")
+                    if not file_path.exists():
+                        missing_files.append(file_info.get("filename", file_id))
+
+            if missing_files:
+                count = len(missing_files)
+                preview = missing_files[:3]
+                result.warnings.append(
+                    f"{count} files referenced in metadata but missing on disk (e.g., {preview}...) - will skip these"
+                )
+
+        except (json.JSONDecodeError, IOError):
+            pass  # If we can't read metadata, just skip file validation
+
+
+class ExperimentUploader:
+    """Handles uploading a single experiment."""
+
+    def __init__(
+        self,
+        local_storage: LocalStorage,
+        remote_client: RemoteClient,
+        batch_size: int = 100,
+        skip_logs: bool = False,
+        skip_metrics: bool = False,
+        skip_files: bool = False,
+        skip_params: bool = False,
+        verbose: bool = False,
+        progress: Optional[Progress] = None,
+        max_concurrent_metrics: int = 5,
+    ):
+        """
+        Initialize uploader.
+
+        Args:
+            local_storage: Local storage instance
+            remote_client: Remote client instance
+            batch_size: Batch size for logs/metrics
+            skip_logs: Skip uploading logs
+            skip_metrics: Skip uploading metrics
+            skip_files: Skip uploading files
+            skip_params: Skip uploading parameters
+            verbose: Show verbose output
+            progress: Optional rich Progress instance for tracking
+            max_concurrent_metrics: Maximum concurrent metric uploads (default: 5)
+        """
+        self.local = local_storage
+        self.remote = remote_client
+        self.batch_size = batch_size
+        self.skip_logs = skip_logs
+        self.skip_metrics = skip_metrics
+        self.skip_files = skip_files
+        self.skip_params = skip_params
+        self.verbose = verbose
+        self.progress = progress
+        self.max_concurrent_metrics = max_concurrent_metrics
+        # Thread-safe lock for shared state updates
+        self._lock = threading.Lock()
+        # Thread-local storage for remote clients (for thread-safe HTTP requests)
+        self._thread_local = threading.local()
+
+    def _get_remote_client(self) -> RemoteClient:
+        """Get thread-local remote client for safe concurrent access."""
+        if not hasattr(self._thread_local, 'client'):
+            # Create a new client for this thread
+            self._thread_local.client = RemoteClient(
+                base_url=self.remote.base_url,
+                api_key=self.remote.api_key
+            )
+        return self._thread_local.client
+
+    def upload_experiment(
+        self, exp_info: ExperimentInfo, validation_result: ValidationResult, task_id=None
+    ) -> UploadResult:
+        """
+        Upload a single experiment with all its data.
+
+        Args:
+            exp_info: Experiment information
+            validation_result: Validation results
+            task_id: Optional progress task ID
+
+        Returns:
+            UploadResult with upload status
+        """
+        result = UploadResult(experiment=f"{exp_info.project}/{exp_info.experiment}")
+
+        # Calculate total steps for progress tracking
+        total_steps = 1  # metadata
+        if not self.skip_params and "parameters" in validation_result.valid_data:
+            total_steps += 1
+        if not self.skip_logs and exp_info.has_logs:
+            total_steps += 1
+        if not self.skip_metrics and exp_info.metric_names:
+            total_steps += len(exp_info.metric_names)
+        if not self.skip_files and exp_info.file_count > 0:
+            total_steps += exp_info.file_count
+
+        current_step = 0
+
+        def update_progress(description: str):
+            nonlocal current_step
+            current_step += 1
+            if self.progress and task_id is not None:
+                self.progress.update(task_id, completed=current_step, total=total_steps, description=description)
+
+        try:
+            # 1. Create/update experiment metadata
+            update_progress("Creating experiment...")
+            if self.verbose:
+                console.print(f"  [dim]Creating experiment...[/dim]")
+
+            exp_data = validation_result.valid_data
+
+            # Store folder path in metadata (not as folderId which expects Snowflake ID)
+            custom_metadata = exp_data.get("metadata") or {}
+            if exp_data.get("folder"):
+                custom_metadata["folder"] = exp_data["folder"]
+
+            response = self.remote.create_or_update_experiment(
+                project=exp_info.project,
+                name=exp_info.experiment,
+                description=exp_data.get("description"),
+                tags=exp_data.get("tags"),
+                bindrs=exp_data.get("bindrs"),
+                folder=None,  # Don't send folder path as folderId (expects Snowflake ID)
+                write_protected=exp_data.get("write_protected", False),
+                metadata=custom_metadata if custom_metadata else None,
+            )
+
+            # Extract experiment ID from nested response
+            experiment_id = response.get("experiment", {}).get("id") or response.get("id")
+            if self.verbose:
+                console.print(f"  [green]✓[/green] Created experiment (id: {experiment_id})")
+
+            # 2. Upload parameters
+            if not self.skip_params and "parameters" in validation_result.valid_data:
+                update_progress("Uploading parameters...")
+                if self.verbose:
+                    console.print(f"  [dim]Uploading parameters...[/dim]")
+
+                params = validation_result.valid_data["parameters"]
+                self.remote.set_parameters(experiment_id, params)
+                result.uploaded["params"] = len(params)
+                # Track bytes (approximate JSON size)
+                result.bytes_uploaded += len(json.dumps(params).encode('utf-8'))
+
+                if self.verbose:
+                    console.print(f"  [green]✓[/green] Uploaded {len(params)} parameters")
+
+            # 3. Upload logs
+            if not self.skip_logs and exp_info.has_logs:
+                count = self._upload_logs(experiment_id, exp_info, result, task_id, update_progress)
+                result.uploaded["logs"] = count
+
+            # 4. Upload metrics
+            if not self.skip_metrics and exp_info.metric_names:
+                count = self._upload_metrics(experiment_id, exp_info, result, task_id, update_progress)
+                result.uploaded["metrics"] = count
+
+            # 5. Upload files
+            if not self.skip_files and exp_info.file_count > 0:
+                count = self._upload_files(experiment_id, exp_info, result, task_id, update_progress)
+                result.uploaded["files"] = count
+
+            result.success = True
+
+        except Exception as e:
+            result.success = False
+            result.errors.append(str(e))
+            if self.verbose:
+                console.print(f"  [red]✗ Error: {e}[/red]")
+
+        return result
+
+    def _upload_logs(self, experiment_id: str, exp_info: ExperimentInfo, result: UploadResult,
+                     task_id=None, update_progress=None) -> int:
+        """Upload logs in batches."""
+        if update_progress:
+            update_progress("Uploading logs...")
+        if self.verbose:
+            console.print(f"  [dim]Uploading logs...[/dim]")
+
+        logs_file = exp_info.path / "logs" / "logs.jsonl"
+        logs_batch = []
+        total_uploaded = 0
+        skipped = 0
+
+        try:
+            with open(logs_file, "r") as f:
+                for line in f:
+                    try:
+                        log_entry = json.loads(line)
+
+                        # Validate required fields
+                        if "message" not in log_entry:
+                            skipped += 1
+                            continue
+
+                        # Prepare log entry for API
+                        api_log = {
+                            "timestamp": log_entry.get("timestamp"),
+                            "level": log_entry.get("level", "info"),
+                            "message": log_entry["message"],
+                        }
+                        if "metadata" in log_entry:
+                            api_log["metadata"] = log_entry["metadata"]
+
+                        logs_batch.append(api_log)
+                        # Track bytes
+                        result.bytes_uploaded += len(line.encode('utf-8'))
+
+                        # Upload batch
+                        if len(logs_batch) >= self.batch_size:
+                            self.remote.create_log_entries(experiment_id, logs_batch)
+                            total_uploaded += len(logs_batch)
+                            logs_batch = []
+
+                    except json.JSONDecodeError:
+                        skipped += 1
+                        continue
+
+            # Upload remaining logs
+            if logs_batch:
+                self.remote.create_log_entries(experiment_id, logs_batch)
+                total_uploaded += len(logs_batch)
+
+            if self.verbose:
+                msg = f"  [green]✓[/green] Uploaded {total_uploaded} log entries"
+                if skipped > 0:
+                    msg += f" (skipped {skipped} invalid)"
+                console.print(msg)
+
+        except IOError as e:
+            result.failed.setdefault("logs", []).append(str(e))
+
+        return total_uploaded
+
+    def _upload_single_metric(
+        self,
+        experiment_id: str,
+        metric_name: str,
+        metric_dir: Path,
+        result: UploadResult
+    ) -> Dict[str, Any]:
+        """
+        Upload a single metric (thread-safe helper).
+
+        Returns:
+            Dict with 'success', 'uploaded', 'skipped', 'bytes', and 'error' keys
+        """
+        data_file = metric_dir / "data.jsonl"
+        data_batch = []
+        total_uploaded = 0
+        skipped = 0
+        bytes_uploaded = 0
+
+        # Get thread-local client for safe concurrent HTTP requests
+        remote_client = self._get_remote_client()
+
+        try:
+            with open(data_file, "r") as f:
+                for line in f:
+                    try:
+                        data_point = json.loads(line)
+
+                        # Validate required fields
+                        if "data" not in data_point:
+                            skipped += 1
+                            continue
+
+                        data_batch.append(data_point["data"])
+                        bytes_uploaded += len(line.encode('utf-8'))
+
+                        # Upload batch using thread-local client
+                        if len(data_batch) >= self.batch_size:
+                            remote_client.append_batch_to_metric(
+                                experiment_id, metric_name, data_batch
+                            )
+                            total_uploaded += len(data_batch)
+                            data_batch = []
+
+                    except json.JSONDecodeError:
+                        skipped += 1
+                        continue
+
+            # Upload remaining data points using thread-local client
+            if data_batch:
+                remote_client.append_batch_to_metric(experiment_id, metric_name, data_batch)
+                total_uploaded += len(data_batch)
+
+            return {
+                'success': True,
+                'uploaded': total_uploaded,
+                'skipped': skipped,
+                'bytes': bytes_uploaded,
+                'error': None
+            }
+
+        except Exception as e:
+            return {
+                'success': False,
+                'uploaded': 0,
+                'skipped': 0,
+                'bytes': 0,
+                'error': str(e)
+            }
+
+    def _upload_metrics(self, experiment_id: str, exp_info: ExperimentInfo, result: UploadResult,
+                        task_id=None, update_progress=None) -> int:
+        """Upload metrics in parallel with concurrency limit."""
+        if not exp_info.metric_names:
+            return 0
+
+        total_metrics = 0
+
+        # Use ThreadPoolExecutor for parallel uploads
+        with ThreadPoolExecutor(max_workers=self.max_concurrent_metrics) as executor:
+            # Submit all metric upload tasks
+            future_to_metric = {}
+            for metric_name in exp_info.metric_names:
+                metric_dir = exp_info.path / "metrics" / metric_name
+                future = executor.submit(
+                    self._upload_single_metric,
+                    experiment_id,
+                    metric_name,
+                    metric_dir,
+                    result
+                )
+                future_to_metric[future] = metric_name
+
+            # Process completed uploads as they finish
+            for future in as_completed(future_to_metric):
+                metric_name = future_to_metric[future]
+
+                # Update progress
+                if update_progress:
+                    update_progress(f"Uploading metric '{metric_name}'...")
+
+                try:
+                    upload_result = future.result()
+
+                    # Thread-safe update of shared state
+                    with self._lock:
+                        result.bytes_uploaded += upload_result['bytes']
+
+                    if upload_result['success']:
+                        total_metrics += 1
+
+                        # Thread-safe console output
+                        if self.verbose:
+                            msg = f"  [green]✓[/green] Uploaded {upload_result['uploaded']} data points for '{metric_name}'"
+                            if upload_result['skipped'] > 0:
+                                msg += f" (skipped {upload_result['skipped']} invalid)"
+                            with self._lock:
+                                console.print(msg)
+                    else:
+                        # Record failure
+                        error_msg = f"{metric_name}: {upload_result['error']}"
+                        with self._lock:
+                            result.failed.setdefault("metrics", []).append(error_msg)
+                            if self.verbose:
+                                console.print(f"  [red]✗[/red] Failed to upload '{metric_name}': {upload_result['error']}")
+
+                except Exception as e:
+                    # Handle unexpected errors
+                    error_msg = f"{metric_name}: {str(e)}"
+                    with self._lock:
+                        result.failed.setdefault("metrics", []).append(error_msg)
+                        if self.verbose:
+                            console.print(f"  [red]✗[/red] Failed to upload '{metric_name}': {e}")
+
+        return total_metrics
+
+    def _upload_files(self, experiment_id: str, exp_info: ExperimentInfo, result: UploadResult,
+                      task_id=None, update_progress=None) -> int:
+        """Upload files one by one."""
+        files_dir = exp_info.path / "files"
+        total_uploaded = 0
+
+        # Use LocalStorage to list files
+        try:
+            files_list = self.local.list_files(exp_info.project, exp_info.experiment)
+
+            for file_info in files_list:
+                # Skip deleted files
+                if file_info.get("deletedAt") is not None:
+                    continue
+
+                try:
+                    if update_progress:
+                        update_progress(f"Uploading {file_info['filename']}...")
+
+                    # Get file path directly from storage without copying
+                    file_id = file_info["id"]
+                    experiment_dir = self.local._get_experiment_dir(exp_info.project, exp_info.experiment)
+                    files_dir = experiment_dir / "files"
+
+                    # Construct file path
+                    file_prefix = file_info["path"].lstrip("/") if file_info["path"] else ""
+                    if file_prefix:
+                        file_path = files_dir / file_prefix / file_id / file_info["filename"]
+                    else:
+                        file_path = files_dir / file_id / file_info["filename"]
+
+                    # Upload to remote with correct parameters
+                    self.remote.upload_file(
+                        experiment_id=experiment_id,
+                        file_path=str(file_path),
+                        prefix=file_info.get("path", ""),
+                        filename=file_info["filename"],
+                        description=file_info.get("description"),
+                        tags=file_info.get("tags", []),
+                        metadata=file_info.get("metadata"),
+                        checksum=file_info["checksum"],
+                        content_type=file_info["contentType"],
+                        size_bytes=file_info["sizeBytes"],
+                    )
+
+                    total_uploaded += 1
+                    # Track bytes
+                    result.bytes_uploaded += file_info.get("sizeBytes", 0)
+
+                    if self.verbose:
+                        size_mb = file_info.get("sizeBytes", 0) / (1024 * 1024)
+                        console.print(f"    [green]✓[/green] {file_info['filename']} ({size_mb:.1f}MB)")
+
+                except Exception as e:
+                    result.failed.setdefault("files", []).append(f"{file_info['filename']}: {e}")
+
+        except Exception as e:
+            result.failed.setdefault("files", []).append(str(e))
+
+        if self.verbose and not result.failed.get("files"):
+            console.print(f"  [green]✓[/green] Uploaded {total_uploaded} files")
+
+        return total_uploaded
+
+
+def cmd_upload(args: argparse.Namespace) -> int:
+    """
+    Execute upload command.
+
+    Args:
+        args: Parsed command-line arguments
+
+    Returns:
+        Exit code (0 for success, 1 for error)
+    """
+    # Load config
+    config = Config()
+
+    # Get remote URL (command line > config)
+    remote_url = args.remote or config.remote_url
+    if not remote_url:
+        console.print("[red]Error:[/red] --remote URL is required (or set in config)")
+        return 1
+
+    # Get API key (command line > config > auto-load from storage)
+    # RemoteClient will auto-load from storage if api_key is None
+    api_key = args.api_key or config.api_key
+
+    # Validate experiment filter requires project
+    if args.experiment and not args.project:
+        console.print("[red]Error:[/red] --experiment requires --project")
+        return 1
+
+    # Discover experiments
+    local_path = Path(args.path)
+    if not local_path.exists():
+        console.print(f"[red]Error:[/red] Local storage path does not exist: {local_path}")
+        return 1
+
+    # Handle state file for resume functionality
+    state_file = Path(args.state_file)
+    upload_state = None
+
+    if args.resume:
+        upload_state = UploadState.load(state_file)
+        if upload_state:
+            # Validate state matches current upload
+            if upload_state.local_path != str(local_path.absolute()):
+                console.print("[yellow]Warning:[/yellow] State file local path doesn't match. Starting fresh upload.")
+                upload_state = None
+            elif upload_state.remote_url != remote_url:
+                console.print("[yellow]Warning:[/yellow] State file remote URL doesn't match. Starting fresh upload.")
+                upload_state = None
+            else:
+                console.print(f"[green]Resuming previous upload from {upload_state.timestamp}[/green]")
+                console.print(f"  Already completed: {len(upload_state.completed_experiments)} experiments")
+                console.print(f"  Failed: {len(upload_state.failed_experiments)} experiments")
+        else:
+            console.print("[yellow]No previous upload state found. Starting fresh upload.[/yellow]")
+
+    # Create new state if not resuming
+    if not upload_state:
+        upload_state = UploadState(
+            local_path=str(local_path.absolute()),
+            remote_url=remote_url,
+        )
+
+    console.print(f"[bold]Scanning local storage:[/bold] {local_path.absolute()}")
+    experiments = discover_experiments(
+        local_path,
+        project_filter=args.project,
+        experiment_filter=args.experiment,
+    )
+
+    if not experiments:
+        if args.project and args.experiment:
+            console.print(f"[yellow]No experiment found:[/yellow] {args.project}/{args.experiment}")
+        elif args.project:
+            console.print(f"[yellow]No experiments found in project:[/yellow] {args.project}")
+        else:
+            console.print("[yellow]No experiments found in local storage[/yellow]")
+        return 1
+
+    # Filter out already completed experiments when resuming
+    if args.resume and upload_state.completed_experiments:
+        original_count = len(experiments)
+        experiments = [
+            exp for exp in experiments
+            if f"{exp.project}/{exp.experiment}" not in upload_state.completed_experiments
+        ]
+        skipped_count = original_count - len(experiments)
+        if skipped_count > 0:
+            console.print(f"[dim]Skipping {skipped_count} already completed experiment(s)[/dim]")
+
+    console.print(f"[green]Found {len(experiments)} experiment(s) to upload[/green]")
+
+    # Display discovered experiments
+    if args.verbose or args.dry_run:
+        console.print("\n[bold]Discovered experiments:[/bold]")
+        for exp in experiments:
+            parts = []
+            if exp.has_logs:
+                parts.append("logs")
+            if exp.has_params:
+                parts.append("params")
+            if exp.metric_names:
+                parts.append(f"{len(exp.metric_names)} metrics")
+            if exp.file_count:
+                size_mb = exp.estimated_size / (1024 * 1024)
+                parts.append(f"{exp.file_count} files ({size_mb:.1f}MB)")
+
+            details = ", ".join(parts) if parts else "metadata only"
+            console.print(f"  [cyan]•[/cyan] {exp.project}/{exp.experiment} [dim]({details})[/dim]")
+
+    # Dry-run mode: stop here
+    if args.dry_run:
+        console.print("\n[yellow bold]DRY RUN[/yellow bold] - No data will be uploaded")
+        console.print("Run without --dry-run to proceed with upload.")
+        return 0
+
+    # Validate experiments
+    console.print("\n[bold]Validating experiments...[/bold]")
+    validator = ExperimentValidator(strict=args.strict)
+    validation_results = {}
+    valid_experiments = []
+    invalid_experiments = []
+
+    for exp in experiments:
+        validation = validator.validate_experiment(exp)
+        validation_results[f"{exp.project}/{exp.experiment}"] = validation
+
+        if validation.is_valid:
+            valid_experiments.append(exp)
+        else:
+            invalid_experiments.append(exp)
+
+        # Show warnings and errors
+        if args.verbose or validation.errors:
+            exp_key = f"{exp.project}/{exp.experiment}"
+            if validation.errors:
+                console.print(f"  [red]✗[/red] {exp_key}:")
+                for error in validation.errors:
+                    console.print(f"      [red]{error}[/red]")
+            elif validation.warnings:
+                console.print(f"  [yellow]⚠[/yellow] {exp_key}:")
+                for warning in validation.warnings:
+                    console.print(f"      [yellow]{warning}[/yellow]")
+
+    if invalid_experiments:
+        console.print(f"\n[yellow]{len(invalid_experiments)} experiment(s) failed validation and will be skipped[/yellow]")
+        if args.strict:
+            console.print("[red]Error: Validation failed in --strict mode[/red]")
+            return 1
+
+    if not valid_experiments:
+        console.print("[red]Error: No valid experiments to upload[/red]")
+        return 1
+
+    console.print(f"[green]{len(valid_experiments)} experiment(s) ready to upload[/green]")
+
+    # Initialize remote client and local storage
+    remote_client = RemoteClient(base_url=remote_url, api_key=api_key)
+    local_storage = LocalStorage(root_path=local_path)
+
+    # Upload experiments with progress tracking
+    console.print(f"\n[bold]Uploading to:[/bold] {remote_url}")
+    results = []
+
+    # Track upload timing
+    import time
+    start_time = time.time()
+
+    # Create progress bar for overall upload
+    with Progress(
+        SpinnerColumn(),
+        TextColumn("[progress.description]{task.description}"),
+        BarColumn(),
+        TaskProgressColumn(),
+        console=console,
+        transient=not args.verbose,  # Keep progress visible in verbose mode
+    ) as progress:
+        # Create uploader with progress tracking
+        uploader = ExperimentUploader(
+            local_storage=local_storage,
+            remote_client=remote_client,
+            batch_size=args.batch_size,
+            skip_logs=args.skip_logs,
+            skip_metrics=args.skip_metrics,
+            skip_files=args.skip_files,
+            skip_params=args.skip_params,
+            verbose=args.verbose,
+            progress=progress,
+        )
+
+        for i, exp in enumerate(valid_experiments, start=1):
+            exp_key = f"{exp.project}/{exp.experiment}"
+
+            # Create task for this experiment
+            task_id = progress.add_task(
+                f"[{i}/{len(valid_experiments)}] {exp_key}",
+                total=100,  # Will be updated with actual steps
+            )
+
+            # Update state - mark as in progress
+            upload_state.in_progress_experiment = exp_key
+            if not args.dry_run:
+                upload_state.save(state_file)
+
+            validation = validation_results[exp_key]
+            result = uploader.upload_experiment(exp, validation, task_id=task_id)
+            results.append(result)
+
+            # Update state - mark as completed or failed
+            upload_state.in_progress_experiment = None
+            if result.success:
+                upload_state.completed_experiments.append(exp_key)
+            else:
+                upload_state.failed_experiments.append(exp_key)
+
+            if not args.dry_run:
+                upload_state.save(state_file)
+
+            # Update task to completed
+            progress.update(task_id, completed=100, total=100)
+
+            if not args.verbose:
+                # Show brief status
+                if result.success:
+                    parts = []
+                    if result.uploaded.get("params"):
+                        parts.append(f"{result.uploaded['params']} params")
+                    if result.uploaded.get("logs"):
+                        parts.append(f"{result.uploaded['logs']} logs")
+                    if result.uploaded.get("metrics"):
+                        parts.append(f"{result.uploaded['metrics']} metrics")
+                    if result.uploaded.get("files"):
+                        parts.append(f"{result.uploaded['files']} files")
+                    status = ", ".join(parts) if parts else "metadata only"
+                    console.print(f"  [green]✓[/green] Uploaded ({status})")
+                else:
+                    console.print(f"  [red]✗[/red] Failed")
+                    if result.errors:
+                        for error in result.errors[:3]:  # Show first 3 errors
+                            console.print(f"      [red]{error}[/red]")
+
+    # Calculate timing
+    end_time = time.time()
+    elapsed_time = end_time - start_time
+    total_bytes = sum(r.bytes_uploaded for r in results)
+
+    # Print summary with rich Table
+    console.print()
+
+    successful = [r for r in results if r.success]
+    failed = [r for r in results if not r.success]
+
+    # Create summary table
+    summary_table = Table(title="Upload Summary", show_header=True, header_style="bold")
+    summary_table.add_column("Status", style="cyan")
+    summary_table.add_column("Count", justify="right")
+
+    summary_table.add_row("Successful", f"[green]{len(successful)}/{len(results)}[/green]")
+    if failed:
+        summary_table.add_row("Failed", f"[red]{len(failed)}/{len(results)}[/red]")
+
+    # Add timing information
+    summary_table.add_row("Total Time", f"{elapsed_time:.2f}s")
+
+    # Calculate and display upload speed
+    if total_bytes > 0 and elapsed_time > 0:
+        # Convert to appropriate unit
+        if total_bytes < 1024 * 1024:  # Less than 1 MB
+            speed_kb = (total_bytes / 1024) / elapsed_time
+            summary_table.add_row("Avg Speed", f"{speed_kb:.2f} KB/s")
+        else:  # 1 MB or more
+            speed_mb = (total_bytes / (1024 * 1024)) / elapsed_time
+            summary_table.add_row("Avg Speed", f"{speed_mb:.2f} MB/s")
+
+    console.print(summary_table)
+
+    # Show failed experiments
+    if failed:
+        console.print("\n[bold red]Failed Experiments:[/bold red]")
+        for result in failed:
+            console.print(f"  [red]✗[/red] {result.experiment}")
+            for error in result.errors:
+                console.print(f"      [dim]{error}[/dim]")
+
+    # Data statistics
+    total_logs = sum(r.uploaded.get("logs", 0) for r in results)
+    total_metrics = sum(r.uploaded.get("metrics", 0) for r in results)
+    total_files = sum(r.uploaded.get("files", 0) for r in results)
+
+    if total_logs or total_metrics or total_files:
+        data_table = Table(title="Data Uploaded", show_header=True, header_style="bold")
+        data_table.add_column("Type", style="cyan")
+        data_table.add_column("Count", justify="right", style="green")
+
+        if total_logs:
+            data_table.add_row("Logs", f"{total_logs} entries")
+        if total_metrics:
+            data_table.add_row("Metrics", f"{total_metrics} metrics")
+        if total_files:
+            data_table.add_row("Files", f"{total_files} files")
+
+        console.print()
+        console.print(data_table)
+
+    # Clean up state file if all uploads succeeded
+    if not args.dry_run and len(failed) == 0 and state_file.exists():
+        state_file.unlink()
+        console.print("\n[dim]Upload complete. State file removed.[/dim]")
+    elif not args.dry_run and failed:
+        console.print(f"\n[yellow]State saved to {state_file}. Use --resume to retry failed uploads.[/yellow]")
+
+    # Return exit code
+    return 0 if len(failed) == 0 else 1