Speeding up Speculative Decoding via Approximate Verification

Open in new window