{ "vocab_size": 50304, "context_length": 1024, "n_embed": 1024, "n_head": 16, "n_blocks": 24, "device": "cuda", "amp_dtype": "bf16", "seed": 1337, "compile": false, "ckpt_dir": "/ephemeral/ckpts", "log_dir": "/ephemeral/logs", "use_wandb": false, "wandb_project": "train-llm-from-scratch-posttrain" }