QNN override

Author: kirklandsign

examples/qualcomm/oss_scripts/llama/runner/runner.cpp

@@ -464,6 +464,13 @@ Error Runner<T>::generate_from_prompt_or_file(
   return Error::Ok;
 }
 
+template <typename T>
+::executorch::runtime::Error prefill(
+    const std::string& prompt,
+    const GenerationConfig& config = {}) {
+  return ::Error::NotImplemented;
+}
+
 template <typename T>
 Result<DecoderModelVersion> Runner<T>::get_decoder_model_version() {
   if (!is_loaded()) {

examples/qualcomm/oss_scripts/llama/runner/runner.h

@@ -79,6 +79,11 @@ class Runner : public executorch::extension::llm::IRunner {
       const executorch::extension::llm::GenerationConfig& config,
       std::function<void(const std::string&)> token_callback = {},
       std::function<void(const executorch::llm::Stats&)> stats_callback = {});
+
+  executorch::runtime::Error prefill(
+      const std::string& prompt,
+      const GenerationConfig& config = {}) override;
+  void stop() override {};
   void stop() override {};
   void reset() override {};
   executorch::runtime::Result<DecoderModelVersion> get_decoder_model_version();