Update src/worker.js

src/worker.js

@@ -80,9 +80,9 @@ await transcriber(new Float32Array(INPUT_SAMPLE_RATE)); // Compile shaders
 const llm_model_id = "onnx-community/Qwen3-1.7B-ONNX";
 const tokenizer = await AutoTokenizer.from_pretrained("Qwen/Qwen3-1.7B"); // Load tokenizer from original repo
 const llm = await AutoModelForCausalLM.from_pretrained(llm_model_id, {
-  dtype: "q4f16",
+  dtype: "q4f16", // This alone should make it pick model_q4f16.onnx
   device: "webgpu",
-  model_file_name: "model_q4f16.onnx" // Specify exact file to avoid external data format
+  // Don't specify model_file_name - let Transformers.js construct the path
 });
 
 const SYSTEM_MESSAGE = {