Fix the deepspeed init_inference reserves device memory as model size

[yeonsily]

For text-generation, set 'keep_module_on_host' to True to save device memory.

For quantization run, https://docs.habana.ai/en/latest/PyTorch/Inference_on_PyTorch/Inference_Using_FP8.html#enabling-and-running-inc-in-pytorch-models, says "If DeepSpeed is used, INC should be called after deepspeed.init_inference. "
But when deepspeed.init_inference() is called, it reserves the device memory as much as bf16 model size because it's initialized as bf16 model dtype first. This reserves extra device memory.

This change is fix for that.

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?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.