enable parallel prefill again

Differential Revision: D61751873

Pull Request resolved: #4893

Author: kimishpatel

examples/models/llama2/runner/runner.cpp

@@ -126,7 +126,7 @@ Error Runner::load() {
       tokenizer_.get(),
       text_decoder_runner_.get(),
       metadata_.at(kUseKVCache),
-      enable_parallel_prefill_);
+      metadata_.at(kEnableDynamicShape));
 
   text_token_generator_ = std::make_unique<TextTokenGenerator>(
       tokenizer_.get(),

examples/models/llama2/runner/runner.h

@@ -45,7 +45,6 @@ class Runner {
 
  private:
   float temperature_;
-  bool enable_parallel_prefill_;
   bool shouldStop_{false};
 
   // model