sglangv0.5.2 & support Qwen3-Next-80B-A3B-Instruct

benchmark/llava_bench/bench_hf_mme.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+
+python -m llava.eval.model_vqa_loader \
+    --model-path liuhaotian/llava-v1.5-7b \
+    --question-file ./mme_pack/llava_mme_bench_replace.jsonl \
+    --image-folder ./mme_pack/MME_Benchmark_release_version \
+    --answers-file ./answers_hf_mme.jsonl \
+    --temperature 0 \
+    --conv-mode vicuna_v1