add qwen3

vllm-v0.6.2/examples/cambricon_custom_func/vllm/mlu_hijack/entrypoints/__init__.py

@@ -0,0 +1 @@
+from . import llm

vllm-v0.6.2/examples/cambricon_custom_func/vllm/mlu_hijack/entrypoints/llm.py

@@ -0,0 +1,98 @@
+from typing import Optional, Dict, Any
+from vllm.entrypoints.llm import LLM
+from vllm_mlu.mlu_hijack_utils import MluHijackObject
+from vllm.logger import init_logger
+from vllm.engine.arg_utils import (EngineArgs, HfOverrides, PoolerConfig,
+                                   TaskOption)
+
+
+logger = init_logger(__name__)
+
+
+vllm__entrypoints__llm__LLM____init__org = LLM.__init__
+
+def vllm__entrypoints__llm__LLM____init__(
+    self,
+    model: str,
+    tokenizer: Optional[str] = None,
+    tokenizer_mode: str = "auto",
+    skip_tokenizer_init: bool = False,
+    trust_remote_code: bool = False,
+    allowed_local_media_path: str = "",
+    tensor_parallel_size: int = 1,
+    dtype: str = "auto",
+    quantization: Optional[str] = None,
+    revision: Optional[str] = None,
+    tokenizer_revision: Optional[str] = None,
+    seed: int = 0,
+    gpu_memory_utilization: float = 0.9,
+    swap_space: float = 4,
+    cpu_offload_gb: float = 0,
+    enforce_eager: Optional[bool] = None,
+    max_seq_len_to_capture: int = 8192,
+    disable_custom_all_reduce: bool = False,
+    disable_async_output_proc: bool = False,
+    hf_overrides: Optional[HfOverrides] = None,
+    mm_processor_kwargs: Optional[Dict[str, Any]] = None,
+    # After positional args are removed, move this right below `model`
+    task: TaskOption = "auto",
+    override_pooler_config: Optional[PoolerConfig] = None,
+    **kwargs,
+) -> None:
+    '''
+    LLM constructor.
+
+    Note: if enforce_eager is unset (enforce_eager is None)
+    it defaults to False.
+    '''
+
+    '''
+    =============================
+    Modify by vllm_mlu
+    =============================
+    @brief: add cp and ep parameter
+    '''
+    # pop context_parallel_size
+    EngineArgs.context_parallel_size = kwargs.pop("context_parallel_size", 1)
+    # pop moe_tp_size and moe_ep_size
+    EngineArgs.moe_tp_size = kwargs.pop("moe_tp_size", -1)
+    # pop moe_ep_size and moe_ep_size
+    EngineArgs.moe_ep_size = kwargs.pop("moe_ep_size", -1)
+    '''
+    ==================
+    End of MLU Hijack
+    ==================
+    '''
+    vllm__entrypoints__llm__LLM____init__org(
+        self=self,
+        model=model,
+        tokenizer=tokenizer,
+        tokenizer_mode=tokenizer_mode,
+        skip_tokenizer_init=skip_tokenizer_init,
+        trust_remote_code=trust_remote_code,
+        allowed_local_media_path=allowed_local_media_path,
+        tensor_parallel_size=tensor_parallel_size,
+        dtype=dtype,
+        quantization=quantization,
+        revision=revision,
+        tokenizer_revision=tokenizer_revision,
+        seed=seed,
+        gpu_memory_utilization=gpu_memory_utilization,
+        swap_space=swap_space,
+        cpu_offload_gb=cpu_offload_gb,
+        enforce_eager=enforce_eager,
+        max_seq_len_to_capture=max_seq_len_to_capture,
+        disable_custom_all_reduce=disable_custom_all_reduce,
+        disable_async_output_proc=disable_async_output_proc,
+        hf_overrides=hf_overrides,
+        mm_processor_kwargs=mm_processor_kwargs,
+        # After positional args are removed, move this right below `model`
+        task=task,
+        override_pooler_config=override_pooler_config,
+        **kwargs
+    )
+
+
+MluHijackObject.apply_hijack(LLM,
+                             LLM.__init__,
+                             vllm__entrypoints__llm__LLM____init__)