Add the multi-cards support for image_to_text examples

[yuanwu2017]

What does this PR do?

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[yafshar]

@yuanwu2017 can you update the README? with information for multi cards run

[yafshar]

I have tried

QUANT_CONFIG=./quantization_config/maxabs_measure.json python ../gaudi_spawn.py --world_size 2 --use_deepspeed run_pipeline.py --model_name_or_path llava-hf/llava-v1.6-mistral-7b-hf --image_path "https://llava-vl.github.io/static/images/view.jpg" --use_hpu_graphs --bf16

and the command is working on your changes and without it. The same as below,

python3 ../gaudi_spawn.py --world_size 2 --use_deepspeed  run_pipeline.py \
    --model_name_or_path llava-hf/llava-v1.6-mistral-7b-hf \
    --use_hpu_graphs \
    --bf16

Can you comment on the PR?

[yuanwu2017]

I have tried

QUANT_CONFIG=./quantization_config/maxabs_measure.json python ../gaudi_spawn.py --world_size 2 --use_deepspeed run_pipeline.py --model_name_or_path llava-hf/llava-v1.6-mistral-7b-hf --image_path "https://llava-vl.github.io/static/images/view.jpg" --use_hpu_graphs --bf16

and the command is working on your changes and without it. The same as below,

python3 ../gaudi_spawn.py --world_size 2 --use_deepspeed  run_pipeline.py \
    --model_name_or_path llava-hf/llava-v1.6-mistral-7b-hf \
    --use_hpu_graphs \
    --bf16

Can you comment on the PR?
Yes, it can run without the patch, but it is not in TP mode. That is equivalent to running two separate instances that are not related.

[yuanwu2017]

@yuanwu2017 can you update the README? with information for multi cards run

Ok. I will update later.

[yafshar]

@yuanwu2017 would you please address the comment to wrap up this PR. We want to have it in 1.18 release. Thanks

[yuanwu2017]

Updated the README. The PR is ok. Please help to review. @regisss @libinta @yafshar

[libinta]

@yuanwu2017 it seems we already have multi-card examples now, please check and close the PR.thanks