From 72688adb2f3ff6cc6302d95d7fced8d41d243270 Mon Sep 17 00:00:00 2001 From: Hongxin Liu Date: Mon, 22 May 2023 10:56:47 +0800 Subject: [PATCH] [doc] add booster docstring and fix autodoc (#3789) * [doc] add docstr for booster methods * [doc] fix autodoc --- colossalai/booster/booster.py | 14 ++++++++++++++ docs/source/en/basics/booster_api.md | 18 ------------------ docs/source/en/features/cluster_utils.md | 18 +----------------- docs/source/zh-Hans/basics/booster_api.md | 18 ------------------ docs/source/zh-Hans/features/cluster_utils.md | 18 +----------------- 5 files changed, 16 insertions(+), 70 deletions(-) diff --git a/colossalai/booster/booster.py b/colossalai/booster/booster.py index 6f2adaf03..4055e55df 100644 --- a/colossalai/booster/booster.py +++ b/colossalai/booster/booster.py @@ -130,6 +130,12 @@ class Booster: return model, optimizer, criterion, dataloader, lr_scheduler def backward(self, loss: torch.Tensor, optimizer: Optimizer) -> None: + """Backward pass. + + Args: + loss (torch.Tensor): The loss to be backpropagated. + optimizer (Optimizer): The optimizer to be updated. + """ # TODO: implement this method with plugin optimizer.backward(loss) @@ -146,6 +152,14 @@ class Booster: pass def no_sync(self, model: nn.Module) -> contextmanager: + """Context manager to disable gradient synchronization across DP process groups. + + Args: + model (nn.Module): The model to be disabled gradient synchronization. + + Returns: + contextmanager: Context to disable gradient synchronization. + """ assert self.plugin is not None, f'no_sync is only enabled when a plugin is provided and the plugin supports no_sync.' assert self.plugin.support_no_sync, f'The plugin {self.plugin.__class__.__name__} does not support no_sync.' return self.plugin.no_sync(model) diff --git a/docs/source/en/basics/booster_api.md b/docs/source/en/basics/booster_api.md index 18dec4500..cafcb6d43 100644 --- a/docs/source/en/basics/booster_api.md +++ b/docs/source/en/basics/booster_api.md @@ -25,24 +25,6 @@ Plugin is an important component that manages parallel configuration (eg: The ge {{ autodoc:colossalai.booster.Booster }} -{{ autodoc:colossalai.booster.Booster.boost }} - -{{ autodoc:colossalai.booster.Booster.backward }} - -{{ autodoc:colossalai.booster.Booster.no_sync }} - -{{ autodoc:colossalai.booster.Booster.save_model }} - -{{ autodoc:colossalai.booster.Booster.load_model }} - -{{ autodoc:colossalai.booster.Booster.save_optimizer }} - -{{ autodoc:colossalai.booster.Booster.load_optimizer }} - -{{ autodoc:colossalai.booster.Booster.save_lr_scheduler }} - -{{ autodoc:colossalai.booster.Booster.load_lr_scheduler }} - ## Usage In a typical workflow, you should launch distributed environment at the beginning of training script and create objects needed (such as models, optimizers, loss function, data loaders etc.) firstly, then call `colossalai.booster` to inject features into these objects, After that, you can use our booster APIs and these returned objects to continue the rest of your training processes. diff --git a/docs/source/en/features/cluster_utils.md b/docs/source/en/features/cluster_utils.md index 1903d64d2..7331d5e73 100644 --- a/docs/source/en/features/cluster_utils.md +++ b/docs/source/en/features/cluster_utils.md @@ -13,20 +13,4 @@ We provide a utility class `colossalai.cluster.DistCoordinator` to coordinate di {{ autodoc:colossalai.cluster.DistCoordinator }} -{{ autodoc:colossalai.cluster.DistCoordinator.is_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.is_node_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.is_last_process }} - -{{ autodoc:colossalai.cluster.DistCoordinator.print_on_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.print_on_node_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.priority_execution }} - -{{ autodoc:colossalai.cluster.DistCoordinator.destroy }} - -{{ autodoc:colossalai.cluster.DistCoordinator.block_all }} - -{{ autodoc:colossalai.cluster.DistCoordinator.on_master_only }} + diff --git a/docs/source/zh-Hans/basics/booster_api.md b/docs/source/zh-Hans/basics/booster_api.md index 5410cc213..1bb5fd69b 100644 --- a/docs/source/zh-Hans/basics/booster_api.md +++ b/docs/source/zh-Hans/basics/booster_api.md @@ -25,24 +25,6 @@ Booster插件是管理并行配置的重要组件（eg：gemini插件封装了ge {{ autodoc:colossalai.booster.Booster }} -{{ autodoc:colossalai.booster.Booster.boost }} - -{{ autodoc:colossalai.booster.Booster.backward }} - -{{ autodoc:colossalai.booster.Booster.no_sync }} - -{{ autodoc:colossalai.booster.Booster.save_model }} - -{{ autodoc:colossalai.booster.Booster.load_model }} - -{{ autodoc:colossalai.booster.Booster.save_optimizer }} - -{{ autodoc:colossalai.booster.Booster.load_optimizer }} - -{{ autodoc:colossalai.booster.Booster.save_lr_scheduler }} - -{{ autodoc:colossalai.booster.Booster.load_lr_scheduler }} - ## 使用方法及示例 在使用colossalai训练时，首先需要在训练脚本的开头启动分布式环境，并创建需要使用的模型、优化器、损失函数、数据加载器等对象。之后，调用`colossalai.booster` 将特征注入到这些对象中，您就可以使用我们的booster API去进行您接下来的训练流程。 diff --git a/docs/source/zh-Hans/features/cluster_utils.md b/docs/source/zh-Hans/features/cluster_utils.md index ca787a869..f54a72c63 100644 --- a/docs/source/zh-Hans/features/cluster_utils.md +++ b/docs/source/zh-Hans/features/cluster_utils.md @@ -13,20 +13,4 @@ {{ autodoc:colossalai.cluster.DistCoordinator }} -{{ autodoc:colossalai.cluster.DistCoordinator.is_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.is_node_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.is_last_process }} - -{{ autodoc:colossalai.cluster.DistCoordinator.print_on_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.print_on_node_master }} - -{{ autodoc:colossalai.cluster.DistCoordinator.priority_execution }} - -{{ autodoc:colossalai.cluster.DistCoordinator.destroy }} - -{{ autodoc:colossalai.cluster.DistCoordinator.block_all }} - -{{ autodoc:colossalai.cluster.DistCoordinator.on_master_only }} +