Support torchrun vllm DP

[luccafong]

Support torchrun external launch mode vllm DP from PR vllm-project/vllm#24899

torchrun --nproc-per-node=8 --no-python  lm_eval     --model vllm     --model_args "pretrained=/data/local/models/oss/DeepSeek-R1-0528,max_model_len=20000,gpu_memory_utilization=0.9,tensor_parallel_size=1,data_parallel_size=8,enable_expert_parallel=true,max_num_seqs=256,distributed_executor_backend=external_launcher"     --batch_size 256     --task gsm8k

|Tasks|Version|     Filter     |n-shot|  Metric   |   |Value|   |Stderr|
|-----|------:|----------------|-----:|-----------|---|----:|---|-----:|
|gsm8k|      3|flexible-extract|     5|exact_match|↑  |0.956|±  |0.0056|
|     |       |strict-match    |     5|exact_match|↑  |0.953|±  |0.0058|

[CLAassistant]

All committers have signed the CLA.