{ "epoch": 25.0, "eval_loss": 1.7208493947982788, "eval_mem_cpu_alloc_delta": 104295, "eval_mem_cpu_peaked_delta": 513049, "eval_mem_gpu_alloc_delta": 0, "eval_mem_gpu_peaked_delta": 5843399680, "eval_runtime": 2.9265, "eval_samples": 20, "eval_samples_per_second": 6.834, "init_mem_cpu_alloc_delta": 336056, "init_mem_cpu_peaked_delta": 18306, "init_mem_gpu_alloc_delta": 509558784, "init_mem_gpu_peaked_delta": 0, "perplexity": 5.58927394881408, "train_mem_cpu_alloc_delta": 846748, "train_mem_cpu_peaked_delta": 752929, "train_mem_gpu_alloc_delta": 1501306368, "train_mem_gpu_peaked_delta": 3372484096, "train_runtime": 1074.2277, "train_samples": 184, "train_samples_per_second": 4.282 }