update dtk to 24.04.1 and modify README

vllm/vllm/engine/llm_engine.py

@@ -181,7 +181,7 @@ class LLMEngine:
         # Create the scheduler.
         # NOTE: the cache_config here have been updated with the numbers of
         # GPU and CPU blocks, which are profiled in the distributed executor.
-        self.scheduler = Scheduler(scheduler_config, cache_config, lora_config)
+        self.scheduler = Scheduler(model_config, parallel_config, scheduler_config, cache_config, lora_config)
 
         # Metric Logging.
         if self.log_stats:
@@ -204,6 +204,13 @@ class LLMEngine:
                     self.get_tokenizer_for_seq,
                 ),
             ))
+        self.num_layers = model_config.get_num_layers(parallel_config)
+        self.total_num_heads = model_config.hf_config.num_attention_heads
+        self.head_size = model_config.get_head_size()
+        self.hidden_size = self.head_size * self.total_num_heads
+        self.device_config = (device_config if device_config is not None else DeviceConfig())
+        self.device = self.device_config.device
+        self.dtype = self.model_config.dtype
 
     def _initialize_kv_caches(self) -> None:
         """Initialize the KV cache in the worker(s).
@@ -385,7 +392,8 @@ class LLMEngine:
             lora_request).eos_token_id
         seq = Sequence(seq_id, prompt, prompt_token_ids, block_size,
                        eos_token_id, lora_request)
-
+        if self.model_config.hf_config.model_type == 'yuan':
+            seq.create_lf_caches(self.hidden_size, self.num_layers, self.device, self.dtype)
         # Defensive copy of SamplingParams, which are used by the sampler,
         # this doesn't deep-copy LogitsProcessor objects
         sampling_params = sampling_params.clone()