Protobuf parsing failed when trying to replicate README example.

[Lyreck]

Hi all. Thanks for this very cool work, it's very useful to be able to use these models in the browser.

I am not sure I understand how the library is supposed to be used. I tried replicating the minimal example provided in the README but can't get to run any GLINER model in Chrome on my Mac.

1. Either I point to the remote repo on HuggingFace (tokenizerPath : "onnx-community/gliner_small-v2" and modelPath : "https://huggingface.co/onnx-community/gliner_small-v2/blob/main/onnx/model.onnx")
2. Either I download the model locally and point to it (tokenizerPath : "onnx-community/gliner_small-v2" and modelPath : "https://huggingface.co/onnx-community/gliner_small-v2/blob/main/onnx/model.onnx").

In situation 1. I get an error (unable to fetch). In the second case, I get a protobuf error: ERROR: Can't create a session. ERROR_CODE: 7, ERROR_MESSAGE: Failed to load model because protobuf parsing failed. I checked online, I did download the full model (approx. 611Mb).

Is there something I am doing wrong ? Or should this work ? I tried a smaller model (the quantized version), I tried with both wasm and webgpu.

Thanks again.

EDIT: I also tried the example provided in the repo files, but end up with the exact same error.

import { useState } from "react";
import { Gliner } from "gliner";

function GlinerSmallTab() {
  const [status, setStatus] = useState("idle");
  const [logs, setLogs] = useState([]);
  const [result, setResult] = useState(null);
  const [error, setError] = useState(null);

  const addLog = (msg) => {
    setLogs((prev) => [...prev, `[${new Date().toLocaleTimeString()}] ${msg}`]);
  };

  const handleTest = async () => {
    setStatus("loading");
    setLogs([]);
    setResult(null);
    setError(null);

    try {
      addLog("Creating Gliner instance...");
      const gliner = new Gliner({
        tokenizerPath: "onnx-community/gliner_small-v2",
        onnxSettings: {
          modelPath: "public/model_quantized.onnx",//"https://huggingface.co/onnx-community/gliner_small-v2/blob/main/onnx/model.onnx",
          executionProvider: "webgpu"
        }
      });

      addLog("Calling initialize()...");
      await gliner.initialize();
      addLog("Model initialized successfully!");

      addLog("Running inference on test text...");
      const texts = ["I'm Francis I come from Jakarta, Indonesia"];
      const entities = ["city", "country", "person"];
      const options = {
        flatNer: false,
        threshold: 0.1,
        multiLabel: false,
      };

      const results = await gliner.inference({
        texts,
        entities,
        ...options,
      });

      addLog("Inference complete!");
      setResult(results);
      setStatus("done");
    } catch (err) {
      addLog(`ERROR: ${err.message}`);
      setError(err.message);
      setStatus("error");
    }
  };

  return (
    <div className="tab-content">
      <h2>Gliner Small Test</h2>
      <p>Testing GLiner.js with gliner_small-v2 model</p>
      
      <button
        onClick={handleTest}
        disabled={status === "loading"}
        className="benchmark-button"
      >
        {status === "loading" ? "Running Test..." : "Run Test"}
      </button>

      <div className="logs-container" style={{ marginTop: "20px" }}>
        <h3>Logs</h3>
        <pre className="logs">
          {logs.length === 0 ? "No logs yet..." : logs.join("\n")}
        </pre>
      </div>

      {result && (
        <div className="results" style={{ marginTop: "20px" }}>
          <h3>Results</h3>
          <pre>{JSON.stringify(result, null, 2)}</pre>
        </div>
      )}

      {error && (
        <div className="error-details" style={{ marginTop: "20px" }}>
          <h3>Error</h3>
          <p style={{ color: "red" }}>{error}</p>
        </div>
      )}
    </div>
  );
}

export default GlinerSmallTab;```