dotnet/machinelearning · critical · ArgumentException

The provided model file {filePath} doesn't exist.

Error message

The provided model file {filePath} doesn't exist.

What it means

FileModelLoader.Start loads an ML.NET model from a file path and throws ArgumentException when the file does not exist at the given path. The loader needs a real file to watch and load (including for reload polling), so it fails fast before setting up watchers. The message includes the offending path to make the misconfiguration obvious.

Source

Thrown at src/Microsoft.Extensions.ML/ModelLoaders/FileModelLoader.cs:42

        private readonly MLContext _context;

        private readonly object _lock;

        public FileModelLoader(IOptions<MLOptions> contextOptions, ILogger<FileModelLoader> logger)
        {
            _logger = logger ?? throw new ArgumentNullException(nameof(logger));
            _context = contextOptions.Value?.MLContext ?? throw new ArgumentNullException(nameof(contextOptions));
            _lock = new object();
        }

        public void Start(string filePath, bool watchFile)
        {
            _filePath = filePath;
            _reloadToken = new ModelReloadToken();

            if (!File.Exists(filePath))
            {
                throw new ArgumentException($"The provided model file {filePath} doesn't exist.");
            }

            var directory = Path.GetDirectoryName(filePath);

            if (string.IsNullOrEmpty(directory))
            {
                directory = Directory.GetCurrentDirectory();
            }

            var file = Path.GetFileName(filePath);

            LoadModel();

            if (watchFile)
            {
                _watcher = new FileSystemWatcher(directory, file);
                _watcher.EnableRaisingEvents = true;
                _watcher.Changed += WatcherChanged;

View on GitHub (pinned to 7b76e69cf9)

Solutions

1. Verify the file exists at the exact path passed to Start: use an absolute path or check File.Exists first.
2. Fix relative-path resolution — confirm the process's current working directory and anchor the path with Path.Combine(AppContext.BaseDirectory, ...) or an absolute config value.
3. Ensure the model file is deployed/copied into the container or output directory (set Copy to Output Directory, or bake it into the image).
4. If the model is produced by a training job, verify that job succeeded and wrote to the expected location before starting the app.

Example fix

// before
var loader = new FileModelLoader(_logger, "models/model.zip");
loader.Start(); // ArgumentException if CWD differs or file missing
// after
string modelPath = Path.Combine(AppContext.BaseDirectory, "models", "model.zip");
if (!File.Exists(modelPath))
    throw new FileNotFoundException("ML.NET model file not found.", modelPath);
var loader = new FileModelLoader(_logger, modelPath);
loader.Start();

string modelPath = config["MlModel:Path"];
if (string.IsNullOrWhiteSpace(modelPath) || !File.Exists(modelPath))
    throw new FileNotFoundException($"Model file not found: {modelPath}");

static bool ModelFileExists(string path) => !string.IsNullOrWhiteSpace(path) && File.Exists(path);

try
{
    loader.Start();
}
catch (ArgumentException ex) when (ex.Message.Contains("doesn't exist"))
{
    logger.LogCritical(ex, "Configured ML.NET model file does not exist.");
    throw; // fail fast at startup
}

When it happens

Trigger: Calling predictionEnginePool/model loading pipeline wiring with a FileModelLoader whose path string points to a nonexistent file; a typo'd relative path where the working directory differs from expected; the model file deleted or never produced by training; container image missing the model artifact.

Common situations: ASP.NET Core app registering AddPredictionEnginePool with a model path that exists on the dev machine but not in the container; CI working directory differences; model path configured via appsettings pointing at a staging path; training job failed so the .zip model was never written.

Understand the failure class

Background: "File not found" and ENOENT errors: why libraries can't find a file that should exist — this error's family across 50 libraries.

Related errors

• Start must be called on a ModelLoader before it can be used.
• File '{path}' does not exist
• If provided, {nameof(samplingKeyColumnName)} must be the sam
• Training data cannot be null
• Training data has 0 rows

AI-assisted analysis of dotnet/machinelearning@7b76e69cf9 (2026-09-11). Data as JSON: /api/errors/3990f77042fb127f. Report an issue: GitHub.