#mi300x

An engineering evaluation of speculative decoding methods in vLLM on AMD Instinct MI300X hardware, analyzing throughput, acceptance rates, and tradeoffs.

Kog AI's MI300X monokernel hits 3K+ t/s per request at batch 1. Try the KIE playground today or follow the HIP replication path — and know w

I loaded Qwen2-VL-72B-Instruct at full BF16 precision on a single GPU, served 64 concurrent DocVQA...