fix some bugs (vllm-project#2689)

Author: zspo

vllm/config.py

@@ -355,6 +355,9 @@ class ParallelConfig:
         worker_use_ray: Whether to use Ray for model workers. Will be set to
             True if either pipeline_parallel_size or tensor_parallel_size is
             greater than 1.
+        max_parallel_loading_workers: Maximum number of multiple batches
+            when load model sequentially. To avoid RAM OOM when using tensor
+            parallel and large models.
         disable_custom_all_reduce: Disable the custom all-reduce kernel and
             fall back to NCCL.
     """
@@ -470,7 +473,7 @@ def __post_init__(self):
         elif self.max_cpu_loras < self.max_loras:
             raise ValueError(
                 f"max_cpu_loras ({self.max_cpu_loras}) must be >= "
-                f"max_num_seqs ({self.max_loras})")
+                f"max_loras ({self.max_loras})")
 
     def verify_with_model_config(self, model_config: ModelConfig):
         if self.lora_dtype in (None, "auto"):

vllm/engine/async_llm_engine.py

@@ -299,6 +299,8 @@ class AsyncLLMEngine:
             async frontend will be executed in a separate process as the
             model workers.
         log_requests: Whether to log the requests.
+        max_log_len: Maximum number of prompt characters or prompt ID numbers
+            being printed in log.
         start_engine_loop: If True, the background task to run the engine
             will be automatically started in the generate call.
         *args: Arguments for LLMEngine.
@@ -434,8 +436,8 @@ async def add_request(
             logger.info(f"Received request {request_id}: "
                         f"prompt: {shortened_prompt!r}, "
                         f"prefix_pos: {prefix_pos},"
-                        f"sampling params: {sampling_params}, "
-                        f"prompt token ids: {shortened_token_ids}, "
+                        f"sampling_params: {sampling_params}, "
+                        f"prompt_token_ids: {shortened_token_ids}, "
                         f"lora_request: {lora_request}.")
 
         if not self.is_running: