stanfordnlp/CoreNLP · error · RuntimeException

format error

Error message

format error

What it means

loadTextClassifier parses a text-serialized CRF line by line and requires a strict header line "labelIndices.length=\t<size>". If the first tab-separated token doesn't equal "labelIndices.length=", it throws this RuntimeException, meaning the stream is not in the expected text format (wrong loader, wrong file, or version mismatch).

Solutions

1. Use the matching loader: binary files go through the binary load path (getClassifier/serializeTo); loadTextClassifier only reads files written by serializeTextClassifier.
2. Verify the file's first line is exactly "labelIndices.length=\t<N>" (key, TAB, integer).
3. Re-serialize the classifier with the same library version that reads it.
4. Confirm the path points at the intended file and that it isn't truncated.

Example fix

// before
// file saved with serializeTo (binary) but read as text -> "format error"
classifier.loadTextClassifier(new BufferedReader(new FileReader("/models/ner.ser.gz")));
// after
// save text: classifier.serializeTextClassifier("/models/ner.txt");
classifier.loadTextClassifier(new BufferedReader(new FileReader("/models/ner.txt")));

try (BufferedReader br = new BufferedReader(new FileReader(modelFile))) {
  String first = br.readLine();
  if (first == null || !first.split("\t")[0].equals("labelIndices.length=")) {
    throw new IllegalArgumentException(modelFile + " is not a text-serialized CRF; use the binary loader");
  }
}

try {
  classifier.loadTextClassifier(br);
} catch (Exception e) {
  if (String.valueOf(e.getMessage()).contains("format error")) {
    // wrong loader or corrupt file: fall back to the binary loader
    classifier = CRFClassifier.getClassifier(modelFile);
  } else throw e;
}

When it happens

Trigger: Calling loadTextClassifier on a binary serialized classifier, an empty/truncated/corrupt file, a manually edited file, or a model saved by a different Stanford NLP version with a changed text format.

Common situations: Mixing serializeTo (binary) output with the text loader; passing the wrong file path; loading models across incompatible library versions.

Understand the failure class

Background: "Invalid ... format", "must be in format X", "does not look like a ..." — invalid argument format errors across CLI tools and libraries — this error's family across 17 libraries.

Related errors

• weights format error
• invalid format: ||
• is not a legal LogPrior.
• Bad data format:
• RuntimeIOException wrapping IOException

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/08659063ceadb653. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/ie/crf/CRFClassifier.java:2105

  protected static List<List<CRFDatum<Collection<String>, String>>> loadProcessedData(String filename) {
    List<List<CRFDatum<Collection<String>, String>>> result;
    try {
      result = IOUtils.readObjectFromURLOrClasspathOrFileSystem(filename);
    } catch (Exception e) {
      log.warn(e);
      result = Collections.emptyList();
    }
    log.info("Loading processed data from serialized file ... done. Got " + result.size() + " datums.");
    return result;
  }

  protected void loadTextClassifier(BufferedReader br) throws Exception {
    String line = br.readLine();
    // first line should be this format:
    // labelIndices.size()=\t%d
    String[] toks = line.split("\\t");
    if (!toks[0].equals("labelIndices.length=")) {
      throw new RuntimeException("format error");
    }
    int size = Integer.parseInt(toks[1]);
    labelIndices = new ArrayList<>(size);
    for (int labelIndicesIdx = 0; labelIndicesIdx < size; labelIndicesIdx++) {
      line = br.readLine();
      // first line should be this format:
      // labelIndices.length=\t%d
      // labelIndices[0].size()=\t%d
      toks = line.split("\\t");
      if (!(toks[0].startsWith("labelIndices[") && toks[0].endsWith("].size()="))) {
        throw new RuntimeException("format error");
      }
      int labelIndexSize = Integer.parseInt(toks[1]);
      labelIndices.add(new HashIndex<>());
      int count = 0;
      while (count < labelIndexSize) {
        line = br.readLine();
        toks = line.split("\\t");

View on GitHub (pinned to 1b7edd19c4)