fix release notes for 2.3.0rc2
This commit is contained in:
parent
c6a1400a90
commit
058a3f4b8f
|
|
@ -59,4 +59,6 @@ https://mindspore.cn/search/en?inputValue=Index%20values
|
|||
|
||||
https://mindspore.cn*/r2.3.q1/*
|
||||
https://www.mindspore.cn*/r2.3.q1/*
|
||||
https://mindspore.cn*/r2.3.0rc2/*
|
||||
https://www.mindspore.cn*/r2.3.0rc2/*
|
||||
http://sox.sourceforge.net/sox.html
|
||||
18
RELEASE.md
18
RELEASE.md
|
|
@ -9,29 +9,29 @@
|
|||
#### AutoParallel
|
||||
|
||||
- [STABLE] Transpose/Sub/Add/Mul/Div/ReLU/Softmax/Sigmoid supports layout configuration.
|
||||
- [STABLE] The collective communication precision will affect network convergence. The configuration item force_fp32_communication is provided in the interface mindspore.set_auto_parallel_context. When set to True, the communication type of the reduce communication operator can be forced to be converted to float32.
|
||||
- [STABLE] The collective communication precision will affect network convergence. The configuration item [force_fp32_communication](https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/mindspore/mindspore.set_auto_parallel_context.html) is provided in the interface mindspore.set_auto_parallel_context. When set to True, the communication type of the reduce communication operator can be forced to be converted to float32.
|
||||
- [BETA] Pipeline parallel support Interleave. Optimize the performance when micro batch is limited.
|
||||
- [BETA] Optimize checkpoint transformation speed when using pipeline parallel, support single stage transform.
|
||||
- [BETA] Pynative mode supports long sequence parallel of RingAttention. Optimizes long sequence training performance.
|
||||
|
||||
#### PyNative
|
||||
|
||||
- [STABLE] Support recompute on PyNative mode
|
||||
- [STABLE] Support register_hook on PyNative mode
|
||||
- [BETA] Support [recompute](https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/mindspore/mindspore.recompute.html) on PyNative mode.
|
||||
- [STABLE] Support [register_hook](https://www.mindspore.cn/docs/en/r2.3.0rc2/api_python/mindspore/Tensor/mindspore.Tensor.register_hook.html#mindspore.Tensor.register_hook) on PyNative mode.
|
||||
|
||||
### API Change
|
||||
|
||||
Add timeout environment variables in dynamic networking scenarios:
|
||||
Add timeout environment variables in [dynamic networking](https://www.mindspore.cn/tutorials/experts/en/r2.3.0rc2/parallel/dynamic_cluster.html) scenarios:
|
||||
|
||||
- MS_TOPO_TIMEOUT: Cluster networking phase timeout time in seconds.
|
||||
- MS_NODE_TIMEOUT: Node heartbeat timeout in seconds.
|
||||
- MS_RECEIVE_MSG_TIMEOUT: Node timeout for receiving messages in seconds.
|
||||
- `MS_TOPO_TIMEOUT`: Cluster networking phase timeout time in seconds.
|
||||
- `MS_NODE_TIMEOUT`: Node heartbeat timeout in seconds.
|
||||
- `MS_RECEIVE_MSG_TIMEOUT`: Node timeout for receiving messages in seconds.
|
||||
|
||||
Added new environment variable MS_ENABLE_LCCL to support the use of LCCL communication library.
|
||||
Added new environment variable `MS_ENABLE_LCCL` to support the use of LCCL communication library.
|
||||
|
||||
### Bug Fixes
|
||||
|
||||
- [#I9CR96] Fix the issue of insufficient timeout time causing failure for dynamic networking startup in large-scale clusters.
|
||||
- [#I94AQQ] Fixed the problem of incorrect output shape of ops.Addcdiv operator in graph mode.
|
||||
|
||||
### Contributors
|
||||
|
||||
|
|
|
|||
|
|
@ -9,29 +9,29 @@
|
|||
#### AutoParallel
|
||||
|
||||
- [STABLE] Transpose/Sub/Add/Mul/Div/ReLU/Softmax/Sigmoid算子支持配置Layout。
|
||||
- [STABLE] 集合通信精度会影响网络收敛，在接口mindspore.set_auto_parallel_context提供配置项force_fp32_communication，设为True时可以强制将reduce类通信算子的通信类型转为float32。
|
||||
- [BETA] 流水并行支持Interleave调度，优化micro batch大小受限场景下的模型性能。
|
||||
- [BETA] 优化流水线并行场景下提高模型转换速度，支持单个stage单独转换。
|
||||
- [BETA] 动态图长序列并行支持RingAttention，优化长序列训练性能。
|
||||
- [STABLE] 集合通信精度会影响网络收敛，在接口mindspore.set_auto_parallel_context提供配置项[force_fp32_communication](https://www.mindspore.cn/docs/zh-CN/r2.3.0rc2/api_python/mindspore/mindspore.set_auto_parallel_context.html)，设为True时可以强制将reduce类通信算子的通信类型转为float32。
|
||||
- [BETA] pipeline并行支持Interleave调度，优化micro batch大小受限场景下的模型性能。
|
||||
- [BETA] 优化pipeline并行场景下提高模型转换速度，支持单个stage单独转换。
|
||||
|
||||
#### PyNative
|
||||
|
||||
- [STABLE] 动态图下支持重计算功能。
|
||||
- [STABLE] 动态图下支持register_hook功能。
|
||||
- [BETA] 动态图下支持[重计算](https://www.mindspore.cn/docs/zh-CN/r2.3.0rc2/api_python/mindspore/mindspore.recompute.html)功能。
|
||||
- [STABLE] 动态图下支持[register_hook](https://www.mindspore.cn/docs/zh-CN/r2.3.0rc2/api_python/mindspore/Tensor/mindspore.Tensor.register_hook.html#mindspore.Tensor.register_hook)功能。
|
||||
|
||||
### API变更
|
||||
|
||||
增加动态组网场景下各类超时时间环境变量配置：
|
||||
增加[动态组网](https://www.mindspore.cn/tutorials/experts/zh-CN/r2.3.0rc2/parallel/dynamic_cluster.html)场景下各类超时时间环境变量配置：
|
||||
|
||||
- MS_TOPO_TIMEOUT： 集群组网阶段超时时间，单位：秒。
|
||||
- MS_NODE_TIMEOUT：节点心跳超时时间，单位：秒。
|
||||
- MS_RECEIVE_MSG_TIMEOUT：节点接收消息超时时间，单位：秒。
|
||||
- `MS_TOPO_TIMEOUT`： 集群组网阶段超时时间，单位：秒。
|
||||
- `MS_NODE_TIMEOUT`：节点心跳超时时间，单位：秒。
|
||||
- `MS_RECEIVE_MSG_TIMEOUT`：节点接收消息超时时间，单位：秒。
|
||||
|
||||
新增环境变量 MS_ENABLE_LCCL，支持昇腾后端单机多卡场景下使用LCCL通信库。
|
||||
新增环境变量 `MS_ENABLE_LCCL`，支持昇腾后端单机多卡场景下使用LCCL通信库。
|
||||
|
||||
### 问题修复
|
||||
|
||||
- [#I9CR96] 修复在大规模集群下，动态组网启动方式的超时时间不足导致集群启动失败的问题。
|
||||
- [#I94AQQ] 修复ops.Addcdiv算子在图模式下输出shape有误问题。
|
||||
|
||||
### 贡献者
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue