Can you share code for inferencing Aria on videos using a single A100 80 GB GPU ?( docs/inference.md)

[PHOENIXFURY007]

https://github.com/rhymes-ai/Aria/blob/main/docs/inference.md#aria-model-inference-methods

Here, the code does not mention how to use it for videos, and only the image usage is specified. Can you share the code for testing on videos ?

2. Following the Commit 1f2e91a, I tried modifying the code for (inference/notebooks/04_video_understanding_vllm.ipynb)
   I ran into the following error:
   ValueError: Model architectures ['AriaForConditionalGeneration'] are not supported for now.

The existing 04_video_understanding_vllm .ipynb file seems to be running on the older version, so can you tell how to modify it so that I can run it using vllm?

[xffxff]

The existing 04_video_understanding_vllm .ipynb file seems to be running on the older version, so can you tell how to modify it so that I can run it using vllm?

Hi @PHOENIXFURY007, sry, my bad. We recently added the Aria model to the official vLLM repo, so I removed the local implementation of Aria’s vLLM version yesterday. However, I forgot to update the usage in these notebooks.

I’ve now fixed this in PR #82. Could you please pull the latest code and try again? Also, make sure you’ve installed the latest vLLM version with the following command:

pip install https://vllm-wheels.s3.us-west-2.amazonaws.com/nightly/vllm-1.0.0.dev-cp38-abi3-manylinux1_x86_64.whl

[PHOENIXFURY007]

Thank You. I was able to resolve it .