{ "epoch": 3.0, "eval_loss": 1.8736294507980347, "eval_mem_cpu_alloc_delta": 73728, "eval_mem_cpu_peaked_delta": 69632, "eval_mem_gpu_alloc_delta": 0, "eval_mem_gpu_peaked_delta": 1368289280, "eval_runtime": 1.981, "eval_samples": 500, "eval_samples_per_second": 252.396, "init_mem_cpu_alloc_delta": 2391441408, "init_mem_cpu_peaked_delta": 380411904, "init_mem_gpu_alloc_delta": 500087808, "init_mem_gpu_peaked_delta": 0, "perplexity": 6.511888138461621, "train_mem_cpu_alloc_delta": 2262335488, "train_mem_cpu_peaked_delta": 53248, "train_mem_gpu_alloc_delta": 1501058560, "train_mem_gpu_peaked_delta": 1229582848, "train_runtime": 25.2846, "train_samples": 500, "train_samples_per_second": 1.898 }