Skip to content

🧠 两小时从 0 训出你的 64M 大模型。Train your own 64M LLM from scratch in 2 hours.

[fix] dtype & lr

jingyaogong5129f0e2a28f
diff --git a/model/model_minimind.py b/model/model_minimind.py
index ad62a68..f7e49d1 100755
--- a/model/model_minimind.py
+++ b/model/model_minimind.py
@@ -313,7 +313,7 @@ class MOEFeedForward(nn.Module):
flat_topk_idx = topk_idx.view(-1)
if self.training:
x = x.repeat_interleave(self.config.num_experts_per_tok, dim=0)
- y = torch.empty_like(x, dtype=torch.float16)
+ y = torch.empty_like(x, dtype=x.dtype)
for i, expert in enumerate(self.experts):
y[flat_topk_idx == i] = expert(x[flat_topk_idx == i]).to(y.dtype) # 确保类型一致
y = (y.view(*topk_weight.shape, -1) * topk_weight.unsqueeze(-1)).sum(dim=1)
diff --git a/trainer/trainer_utils.py b/trainer/trainer_utils.py
index 0ad1b00..c1a4ca2 100644
--- a/trainer/trainer_utils.py
+++ b/trainer/trainer_utils.py
@@ -23,8 +23,7 @@ def Logger(content):
def get_lr(current_step, total_steps, lr):
- min_lr = lr / 10
- return min_lr + 0.5 * (lr - min_lr) * (1 + math.cos(math.pi * current_step / total_steps))
+ return lr*(0.1 + 0.45*(1 + math.cos(math.pi * current_step / total_steps)))
def init_distributed_mode():