Request for Meta-Llama-3-8B-Instruct when evaluating LLaVA-Video-7B with conv_template=qwen_1_5

[Ethan16162]

I am using the following script to evaluate llava-video-7B:

accelerate launch --num_processes=4 \
-m lmms_eval \
--model llava_vid \
--model_args pretrained=/data/gys/models/LLaVA-Video-7B-Qwen2,conv_template=qwen_1_5,max_frames_num=64,mm_spatial_pool_mode=average \
--tasks mme \
--batch_size 1 \
--log_samples \
--log_samples_suffix llava_vid \
--output_path ./logs/

But I encountered the following error:

'(MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /meta-llama/Meta-Llama-3-8B-Instruct/resolve/main/tokenizer_config.json (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x7f32141fb7f0>, 'Connection to huggingface.co timed out. (connect timeout=10)'))"), '(Request ID: 9d6f8820-9bec-4a91-ae0c-b189630c8e26)')' thrown while requesting HEAD https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct/resolve/main/tokenizer_config.json

However, I have already specified conv_template=qwen_1_5. Why is Meta-Llama-3-8B-Instruct still appearing here?

[rookiexiong7]

same issue