[Fix] fix QtuneVL config name and model path

README.md

@@ -24,7 +24,7 @@ English | [简体中文](/docs/zh-CN/README_zh-CN.md) | [日本語](/docs/ja/REA
 **VLMEvalKit** (the python package name is **vlmeval**) is an **open-source evaluation toolkit** of **large vision-language models (LVLMs)**. It enables **one-command evaluation** of LVLMs on various benchmarks, without the heavy workload of data preparation under multiple repositories. In VLMEvalKit, we adopt **generation-based evaluation** for all LVLMs, and provide the evaluation results obtained with both **exact matching** and **LLM-based answer extraction**.
 
 ## Recent Codebase Changes
-- **[2025-08-04]** In [**PR 1175**](https://github.com/open-compass/VLMEvalKit/pull/1175), we refine the `can_infer_option` and `can_infer_text`, which increasingly route the evaluation to LLM choice extractors and empirically leads to slight performance improvement for MCQ benchmarks. 
+- **[2025-08-04]** In [**PR 1175**](https://github.com/open-compass/VLMEvalKit/pull/1175), we refine the `can_infer_option` and `can_infer_text`, which increasingly route the evaluation to LLM choice extractors and empirically leads to slight performance improvement for MCQ benchmarks.
 
 ## 🆕 News
 - **[2025-07-07]** Supported [**SeePhys**](https://seephys.github.io/), which is a ​full spectrum multimodal benchmark for evaluating physics reasoning across different knowledge levels. thanks to [**Quinn777**](https://github.com/Quinn777) 🔥🔥🔥

vlmeval/config.py

@@ -1517,13 +1517,13 @@
 
 # QTuneVL series
 qtunevl_series = {
-    "QTuneVL1.5-2B": partial(
-        QTuneVLChat, model_path="hanchaow/QTuneVL1.5-2B", version="V1.5"
+    "QTuneVL1_5-2B": partial(
+        QTuneVLChat, model_path="hanchaow/QTuneVL1_5-2B", version="V1.5"
     ),
 
-    "QTuneVL1-3B": partial(
+    "QTuneVL1_5-3B": partial(
         QTuneVL,
-        model_path="hanchaow/QTuneVL1.5-3B",
+        model_path="hanchaow/QTuneVL1_5-3B",
         min_pixels=1280 * 28 * 28,
         max_pixels=16384 * 28 * 28,
         use_custom_prompt=True,

vlmeval/vlm/qtunevl/qtune_vl.py

@@ -204,7 +204,7 @@ def __init__(
         self.model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
             model_path, torch_dtype='auto', device_map="auto", attn_implementation='flash_attention_2'
         )
-        self.model.eval()
+        self.model.eval().to('cuda')
 
         torch.cuda.empty_cache()
 

vlmeval/vlm/qtunevl/qtune_vl_chat.py

@@ -527,7 +527,7 @@ def __init__(self,
                 torch_dtype=torch.bfloat16,
                 load_in_8bit=load_in_8bit,
                 trust_remote_code=True,
-                low_cpu_mem_usage=True).eval().cuda()
+                low_cpu_mem_usage=True).eval().to('cuda')
             self.device = 'cuda'
 
         if best_of_n > 1:
@@ -711,7 +711,8 @@ def generate_v2(self, message, dataset=None):
         else:
             pixel_values = None
             num_patches_list = []
-
+        self.model.vision_model.to(self.device)
+        self.model.to(self.device)
         response_list = []
         for idx in range(self.best_of_n):
             kwargs_default = self.kwargs.copy()