🧠 两小时从 0 训出你的 64M 大模型。Train your own 64M LLM from scratch in 2 hours.
修复了wandb的bug,避免了多次产生项目
Yu Chengzhangef9a592d1431
diff --git a/1-pretrain.py b/1-pretrain.pyindex 8560126..56c937d 100644--- a/1-pretrain.py+++ b/1-pretrain.py@@ -146,11 +146,6 @@ if __name__ == "__main__":use_wandb = True #是否使用wandbwandb_project = "MiniMind-Pretrain"wandb_run_name = f"MiniMind-Pretrain-Epoch-{epochs}-BatchSize-{batch_size}-LearningRate-{learning_rate}"- if use_wandb:- import wandb- wandb.init(project=wandb_project, name=wandb_run_name)- else:- wandb = Nonectx = (@@ -163,6 +158,12 @@ if __name__ == "__main__":if ddp:init_distributed_mode()device = torch.device(DEVICE)++ if use_wandb and (not ddp or ddp_local_rank == 0):+ import wandb+ wandb.init(project=wandb_project, name=wandb_run_name)+ else:+ wandb = None# -----------------------------------------------------------------------------# -----init dataloader------