Update README.md

README.md

@@ -58,6 +58,8 @@ b. Open-source Models
 - mPLUG-Owl3
 - Ferret
 
+To ensure a smooth evaluation process, make sure to install the specific version of the Transformers library as specified in the [**repository**](https://github.com/open-compass/VLMEvalKit/tree/main#-datasets-models-and-evaluation-results) for the model you wish to evaluate.
+
 Some models are not supported by VLMEvalKit during our benchmark development. 
 
 We employed the models and the associated Transformers version for the benchmark as follows:
@@ -70,8 +72,6 @@ Model configuration file can be found at `MMDocBench/vlmeval/config.py`.
 
 Model paths should be specified in the configuration file. Additionally, we set the model's `max_new_tokens` parameter to 4096 in the config file as there are some tasks in our benchmark that requires long model prediction output. For some models such as llava and mplug_owl3, where the generation configuration is fixed, we modify the model files to support the usage of the `max_new_tokens` parameter.
 
-To ensure a smooth evaluation process, make sure to install the specific version of the Transformers library as specified in the [**repository**](https://github.com/open-compass/VLMEvalKit/tree/main#-datasets-models-and-evaluation-results) for the model you wish to evaluate.
-
 ### Run the evaluation 
 
 To run evaluation with either `python` or `torchrun` as follows to evaluate one model with MMDocBench.