fix llava runner

Author: shewu-quic

examples/models/llava/runner/llava_runner.cpp

@@ -54,6 +54,7 @@ Error LlavaRunner::load() {
   // Load the text prefiller
   text_prefiller_ = std::make_unique<llm::TextPrefiller>(
       text_decoder_runner_.get(),
+      /*use_int32_token=*/false,
       /*use_kv_cache=*/true,
       /*enable_parallel_prefill=*/true);
 
@@ -65,6 +66,7 @@ Error LlavaRunner::load() {
   text_token_generator_ = std::make_unique<llm::TextTokenGenerator>(
       tokenizer_.get(),
       text_decoder_runner_.get(),
+      /*use_int32_token=*/false,
       /*use_kv_cache=*/true,
       std::make_unique<std::unordered_set<uint64_t>>(
           std::unordered_set<uint64_t>{tokenizer_->eos_tok()}),