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] moe unused

jingyaogong5dd4df7e182e
diff --git a/trainer/train_distill_reason.py b/trainer/train_distill_reason.py
index be69738..6ff7ea3 100644
--- a/trainer/train_distill_reason.py
+++ b/trainer/train_distill_reason.py
@@ -161,7 +161,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_distillation.py b/trainer/train_distillation.py
index 712bac6..c99bbd7 100644
--- a/trainer/train_distillation.py
+++ b/trainer/train_distillation.py
@@ -210,7 +210,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config_student, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_dpo.py b/trainer/train_dpo.py
index b20b53f..d3c3405 100644
--- a/trainer/train_dpo.py
+++ b/trainer/train_dpo.py
@@ -195,7 +195,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_full_sft.py b/trainer/train_full_sft.py
index cea2ffb..6b3f426 100644
--- a/trainer/train_full_sft.py
+++ b/trainer/train_full_sft.py
@@ -96,7 +96,7 @@ if __name__ == "__main__":
parser.add_argument("--log_interval", type=int, default=100, help="日志打印间隔")
parser.add_argument("--save_interval", type=int, default=1000, help="模型保存间隔")
parser.add_argument('--hidden_size', default=512, type=int, help="隐藏层维度")
- parser.add_argument('--num_hidden_layers', default=8, type=int, help="隐藏层数量")
+ parser.add_argument('--num_hidden_layers', default=16, type=int, help="隐藏层数量")
parser.add_argument('--max_seq_len', default=340, type=int, help="训练的最大截断长度(中文1token≈1.5~1.7字符)")
parser.add_argument('--use_moe', default=0, type=int, choices=[0, 1], help="是否使用MoE架构(0=否,1=是)")
parser.add_argument("--data_path", type=str, default="../dataset/sft_mini_512.jsonl", help="训练数据路径")
@@ -149,7 +149,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_grpo.py b/trainer/train_grpo.py
index 897d9a8..580768a 100755
--- a/trainer/train_grpo.py
+++ b/trainer/train_grpo.py
@@ -272,7 +272,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_lora.py b/trainer/train_lora.py
index 89cb7a9..1fba33d 100644
--- a/trainer/train_lora.py
+++ b/trainer/train_lora.py
@@ -162,7 +162,7 @@ if __name__ == "__main__":
# ========== 8. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 9. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_ppo.py b/trainer/train_ppo.py
index cb0ec38..67cb301 100644
--- a/trainer/train_ppo.py
+++ b/trainer/train_ppo.py
@@ -344,8 +344,8 @@ if __name__ == "__main__":
if dist.is_initialized():
actor_model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
critic_model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- actor_model = DistributedDataParallel(actor_model, device_ids=[local_rank])
- critic_model = DistributedDataParallel(critic_model, device_ids=[local_rank])
+ actor_model = DistributedDataParallel(actor_model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
+ critic_model = DistributedDataParallel(critic_model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
old_actor_model.to(args.device)
# ========== 8. 开始训练 ==========
diff --git a/trainer/train_pretrain.py b/trainer/train_pretrain.py
index 5f05341..09eb284 100644
--- a/trainer/train_pretrain.py
+++ b/trainer/train_pretrain.py
@@ -95,7 +95,7 @@ if __name__ == "__main__":
parser.add_argument("--log_interval", type=int, default=100, help="日志打印间隔")
parser.add_argument("--save_interval", type=int, default=1000, help="模型保存间隔")
parser.add_argument('--hidden_size', default=512, type=int, help="隐藏层维度")
- parser.add_argument('--num_hidden_layers', default=8, type=int, help="隐藏层数量")
+ parser.add_argument('--num_hidden_layers', default=16, type=int, help="隐藏层数量")
parser.add_argument('--max_seq_len', default=340, type=int, help="训练的最大截断长度(中文1token≈1.5~1.7字符)")
parser.add_argument('--use_moe', default=0, type=int, choices=[0, 1], help="是否使用MoE架构(0=否,1=是)")
parser.add_argument("--data_path", type=str, default="../dataset/pretrain_hq.jsonl", help="预训练数据路径")
@@ -148,7 +148,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):
diff --git a/trainer/train_spo.py b/trainer/train_spo.py
index 37493e4..700b14e 100755
--- a/trainer/train_spo.py
+++ b/trainer/train_spo.py
@@ -322,7 +322,7 @@ if __name__ == "__main__":
# ========== 7. DDP包模型 ==========
if dist.is_initialized():
model._ddp_params_and_buffers_to_ignore = {"freqs_cos", "freqs_sin"}
- model = DistributedDataParallel(model, device_ids=[local_rank])
+ model = DistributedDataParallel(model, device_ids=[local_rank], find_unused_parameters=getattr(lm_config, 'use_moe', False))
# ========== 8. 开始训练 ==========
for epoch in range(start_epoch, args.epochs):