Did not detect the .env file when using torchrun

[Blissy-32]

when using python run.py --data ChartQA_TEST--model 7B --verbose，it goes fine but too much slow when using torchrun command like，torchrun --nproc-per-node=4 run.py --data ChartQA_TEST--model 7B --verbose， it comes with the following issue

W0423 17:11:07.226000 3089305 site-packages/torch/distributed/run.py:793] ***************************************** [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load. [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load. [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load. [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load. [2025-04-23 17:11:11,812] WARNING - RUN - run.py: main - 174: --reuse is not set, will not reuse previous (before one day) temporary files [2025-04-23 17:11:11] WARNING - run.py: main - 174: --reuse is not set, will not reuse previous (before one day) temporary files [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load. [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load. [2025-04-23 17:11:11] ERROR - misc.py: load_env - 212: Did not detect the .env file at /data3/xxf/VLMEvalKit/.env, failed to load.

could u help me to solve it，many thanks

[MaoSong2022]

I test with torchrun --nproc-per-node=1 run.py --data ChartQA_TEST --model Eagle-X5-7B --verbose and it works fine. Please check if the env file /data3/xxf/VLMEvalKit/.env exists

[Blissy-32]

I test with torchrun --nproc-per-node=1 run.py --data ChartQA_TEST --model Eagle-X5-7B --verbose and it works fine. Please check if the env file /data3/xxf/VLMEvalKit/.env exists

many thanks, and I have another question: Is the [VLM-EvalKit] suitable for evaluating models trained with LoRA? Specifically, I fine-tuned qwenvl using LoRA and want to test its performance. Would the toolkit support this scenario, or are there any compatibility considerations I should be aware of?

[MaoSong2022]

If you fine-tune a supported model using VLMEvalKit, evaluating it should be straightforward. You'll need to define your model, inherit from the base model architecture, and specify the path to your fine-tuned weights in vlmeval/config.py as follows:

qwen_series = {
    "qwen_lora": partial(QwenVL, model_path="path/to/model_path"),
    "qwen_base": partial(QwenVL, model_path="Qwen/Qwen-VL"),
    ...,
}

see config.py/qwenvl_series for details