[Feature Request] Integrate FlashInfer for Prefill and Decode

[lausannel]

Prerequisites

• I have searched existing issues and reviewed documentation.

Problem Description

We'd like to integrate FlashInfer into the project to improve decoding and prefill efficiency.

• For decode, we need to port the logic described here:
  https://docs.flashinfer.ai/generated/flashinfer.decode.single_decode_with_kv_cache.html

• For prefill, a full integration of the necessary components from FlashInfer is required.

This integration aims to leverage FlashInfer's optimized kernel for faster inference.

Proposed Solution

Alternatives Considered

No response

Additional Context

No response

Importance

Important

Usage Statistics (Optional)

No response