Loading the SOTA2 catalog…
LLM Training Throughput on Llama3-8B (8k SeqLen, train) benchmark leaderboard · SOTA2 Research