dotnet/machinelearning · error · ArgumentException

unknown schema type: {schema.Type}

Error message

unknown schema type: {schema.Type}

What it means

NumericOptionConverter.Read deserializes a JSON schema object into a concrete numeric option. The schema's 'type' string must be one of 'int', 'float', or 'double'; any other value throws ArgumentException 'unknown schema type: {schema.Type}'.

Source

Thrown at src/Microsoft.ML.SearchSpace/Converter/NumericOptionConverter.cs:46

            public object Min { get; set; }

            [JsonPropertyName("max")]
            public object Max { get; set; }

            [JsonPropertyName("log_base")]
            public bool LogBase { get; set; }
        }

        public override UniformNumericOption Read(ref Utf8JsonReader reader, Type typeToConvert, JsonSerializerOptions options)
        {
            var schema = JsonSerializer.Deserialize<Schema>(ref reader, options);

            return schema.Type switch
            {
                "int" => new UniformIntOption(Convert.ToInt32(schema.Min), Convert.ToInt32(schema.Max), schema.LogBase, Convert.ToInt32(schema.Default)),
                "float" => new UniformSingleOption(Convert.ToSingle(schema.Min), Convert.ToSingle(schema.Max), schema.LogBase, Convert.ToSingle(schema.Default)),
                "double" => new UniformDoubleOption(Convert.ToDouble(schema.Min), Convert.ToDouble(schema.Max), schema.LogBase, Convert.ToDouble(schema.Default)),
                _ => throw new ArgumentException($"unknown schema type: {schema.Type}"),
            };
        }

        public override void Write(Utf8JsonWriter writer, UniformNumericOption value, JsonSerializerOptions options)
        {
            var schema = value switch
            {
                UniformIntOption intOption => new Schema
                {
                    Type = "int",
                    Default = intOption.SampleFromFeatureSpace(intOption.Default).AsType<int>(),
                    Min = Convert.ToInt32(intOption.Min),
                    Max = Convert.ToInt32(intOption.Max),
                    LogBase = intOption.LogBase,
                },
                UniformDoubleOption doubleOption => new Schema
                {
                    Type = "double",

View on GitHub (pinned to 7b76e69cf9)

Solutions

1. Fix the JSON so type is exactly "int", "float", or "double" (lowercase).
2. Validate the schema file against the expected shape before deserializing.
3. If the source schema uses other names, map them (integer→int, number→double) before conversion.
4. Ensure you're deserializing with the correct target type; a numeric option JSON should not carry other type tags.

Example fix

// before (json)
{ "type": "integer", "min": 1, "max": 10 }
// after (json)
{ "type": "int", "min": 1, "max": 10 }

// C#: validate schema type before deserializing
static readonly HashSet<string> Allowed = new() { "int", "float", "double" };
using var doc = JsonDocument.Parse(json);
var type = doc.RootElement.GetProperty("type").GetString();
if (!Allowed.Contains(type))
    throw new ArgumentException($"Bad schema type '{type}'; expected int|float|double");

static bool IsKnownNumericSchemaType(string t) => t is "int" or "float" or "double";

try { opt = JsonSerializer.Deserialize<OptionBase>(json, options); }
catch (ArgumentException ex) when (ex.Message.StartsWith("unknown schema type"))
{
    // normalize type field (integer->int, number->double) and retry
}

When it happens

Trigger: Deserializing JSON into OptionBase/UniformNumericOption via OptionConverter/NumericOptionConverter when the JSON contains type values other than int/float/double (e.g. 'integer', 'long', 'number', or a hand-written/foreign schema with type 'string').

Common situations: Hand-authoring or generating search-space JSON with non-ML.NET type names; schema files produced by a different tool/version; typos in the type field; JSON edited manually and type field corrupted.

Related errors

• unknown type
• unknown option type
• Unsupported reader type {reader.TokenType}
• Problems met when parsing JSON vocabulary object.{Environmen
• Problems met when parsing JSON vocabulary object.{Environmen

AI-assisted analysis of dotnet/machinelearning@7b76e69cf9 (2026-09-11). Data as JSON: /api/errors/7337f1950d60bbd7. Report an issue: GitHub.