Improve DeepSeek Generation Speed with FP8 Quantization

[guyueh1]

Is your feature request related to a problem? Please describe.
Support FP8 (block-quant recipe) for DeepSeek models to accelerate generation.

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.