Expanding code post-training beyond unit tests to competitive, long-horizon programming tasks with CodeClash’s new training arenas
Technical exploration of LLM inference metrics, batching strategies, and GPU optimization with TensorRT-LLM - from latency metrics to in-flight batching
