stanfordnlp/CoreNLP · error · IllegalArgumentException

The dimension of embedding file does not match…

Error message

The dimension of embedding file does not match config.embeddingSize (<dim> vs <config.embeddingSize>).  Perhaps set the -embeddingSize flag

What it means

When loading pre-trained word embeddings, DependencyParser.readEmbedFile compares the vector dimension parsed from the embedding file (number of whitespace-separated columns minus one) with config.embeddingSize. A mismatch means the neural model dimensions won't line up, so IllegalArgumentException is thrown, advising to set the -embeddingSize flag.

Solutions

1. Add -embeddingSize <dim> matching your embedding file's dimension (e.g. 100, 300)
2. Regenerate or download embeddings with the dimension the config expects
3. Count the columns in the first line of the embedding file to confirm the true dim before rerunning

Example fix

// before
java -cp stanford-corenlp.jar edu.stanford.nlp.parser.nndep.DependencyParser -trainFile ... -embeddingFile w2v.txt # dim 300 vs 50
// after
java -cp stanford-corenlp.jar edu.stanford.nlp.parser.nndep.DependencyParser -trainFile ... -embeddingFile w2v.txt -embeddingSize 300

// verify embedding dim before training
String firstLine = Files.readAllLines(Paths.get(embedFile)).get(0);
int dim = firstLine.trim().split("\\s+").length - 1;
props.setProperty("embeddingSize", String.valueOf(dim));

try {
  DependencyParser.train(config);
} catch (IllegalArgumentException e) {
  // adjust -embeddingSize to the file's real dimension and retry
}

When it happens

Trigger: Passing -embeddingFile whose vectors have a dimension different from config.embeddingSize (default 50) during training or model building.

Common situations: Using word2vec/GloVe/fastText files trained with 100/300 dimensions while the parser config still has embeddingSize=50; reusing an old config with new embeddings; forgetting to pass -embeddingSize alongside -embeddingFile.

Related errors

• Language does not support parsing!
• Parser requires words with part-of-speech tag annotations
• Unknown language <props.containsKey("language")>
• Cannot put a child trie with no keys
• CoreMap must have either a Calendar or DocDate annotation

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/c487d04baee17b5a. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/parser/nndep/DependencyParser.java:656

  private double[][] readEmbedFile(String embedFile, Map<String, Integer> embedID) {

    double[][] embeddings = null;
    if (embedFile != null) {
      try (BufferedReader input = IOUtils.readerFromString(embedFile)) {
        List<String> lines = new ArrayList<>();
        for (String s; (s = input.readLine()) != null; ) {
          lines.add(s);
        }

        int nWords = lines.size();
        String[] splits = lines.get(0).split("\\s+");

        int dim = splits.length - 1;
        embeddings = new double[nWords][dim];
        log.info("Embedding File " + embedFile + ": #Words = " + nWords + ", dim = " + dim);

        if (dim != config.embeddingSize)
            throw new IllegalArgumentException("The dimension of embedding file does not match config.embeddingSize (" + dim + " vs " + config.embeddingSize + ").  Perhaps set the -embeddingSize flag");

        for (int i = 0; i < lines.size(); ++i) {
          splits = lines.get(i).split("\\s+");
          embedID.put(splits[0], i);
          for (int j = 0; j < dim; ++j)
            embeddings[i][j] = Double.parseDouble(splits[j + 1]);
        }
      } catch (IOException e) {
        throw new RuntimeIOException(e);
      }
      embeddings = Util.scaling(embeddings, 0, 1.0);
    }
    return embeddings;
  }

  /**
   * Train a new dependency parser model.
   *

View on GitHub (pinned to 1b7edd19c4)