Skip to content

P
Paddle

afeixing77 / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 0
Fork 0
P
Paddle

体验新版 GitCode，发现更多精彩内容 >>

未验证 提交 7f17f9a0 编写于 作者: J JZ-LIANG 提交者： GitHub
浏览文件
操作

bugfix for mp accuracy (#35326)

上级 b24f84c8
隐藏空白更改
内联 并排
Showing with 1 addition and 1 deletion
python/paddle/distributed/fleet/meta_optimizers/tensor_parallel_optimizer.py
浏览文件 @ 7f17f9a0
...@@ -124,7 +124,7 @@ class TensorParallelOptimizer(MetaOptimizerBase): ...@@ -124,7 +124,7 @@ class TensorParallelOptimizer(MetaOptimizerBase):
collective_helper._init_communicator( collective_helper._init_communicator(
self.startup_program, self.current_endpoint, self.mp_endpoints, self.startup_program, self.current_endpoint, self.mp_endpoints,
self.mp_rank, self.mp_ring_id, True, self.global_ring_id, True) self.mp_rank, self.mp_ring_id, True, self.global_ring_id, True)
#self._broadcast_params(self.mp_ring_id, mp_mode=True) self._broadcast_params(self.mp_ring_id, mp_mode=True)
# Create dp rings # Create dp rings
if self.nranks > self.mp_degree: if self.nranks > self.mp_degree:
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或