Fix requirements file

[GilLevi]

The requirements file was not updated, casing a few small issues when running pip install -e . and trying to run the demo.

1. torchvision version is not specified, thus pip tries to install the latest version which is not compatible with torch 1.10.0, added torchvision==0.11.1 to the requirements file which is the compatible version.
2. The scriptlavis/models/img2prompt_models/img2prompt_vqa.py requires spacy, but it was not reflected in the requirements file resulting in an import error when running the demo, added spacy to the requirement file.
3. spacy requires a specific version of pydantic, see issue in spacy repo here, added pydantic==1.10.2 to the requirements file.

Overall small changes, but hopefully it will save a 10 minutes of debugging to the next person:)

[salesforce-cla[bot]]

Thanks for the contribution! Before we can merge this, we need @GilLevi to sign the Salesforce.com Contributor License Agreement.

[GilLevi]

This is weird, I signed the CLA, but the PR still has the cls:missing label.

[nick-konovalchuk]

This is weird, I signed the CLA, but the PR still has the cls:missing label.

Nah, it seems to work like this all the time

[GilLevi]

Thanks @bottledmind, hope it will get merged then.

[nick-konovalchuk]

The VQA pipeline still requires an en_core_web_sm model. I've created https://github.com/salesforce/LAVIS/pull/77 with a hacky way of solving this issue. If you have a better idea, merge it to this PR and inform me to remove mine