diff --git a/doc/en/install.md b/doc/en/install.md
index b3c42e4..0c721b9 100644
--- a/doc/en/install.md
+++ b/doc/en/install.md
@@ -5,10 +5,10 @@ The required packages and corresponding version are shown as follows:
 - Python == 3.10
 - GCC == 10.2.0
 - MPFR == 4.1.0
-- CUDA == 11.7
-- Pytorch == 1.13.1+cu117
+- CUDA >= 11.7
+- Pytorch >= 1.13.1
 - Transformers >= 4.28.0
-- Flash-Attention == v1.0.5
+- Flash-Attention >= v1.0.5
 - Apex == 23.05
 - GPU with Ampere or Hopper architecture (such as H100, A100)
 - Linux OS
@@ -57,3 +57,14 @@ cd ./third_party/apex
 pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
 cd ../../
 ```
+
+### Environment Image
+Users can obtain an image with the InternLM runtime environment installed from https://hub.docker.com/r/sunpengsdu/internlm. The commands for pulling the image and starting the container are as follows:
+
+```bash
+# pull image
+docker pull sunpengsdu/internlm:torch1.13-cuda11.7-flashatten1.0.5-centos
+# start container
+docker run --gpus all -d -it --shm-size=2gb --name myinternlm sunpengsdu/internlm:torch1.13-cuda11.7-flashatten1.0.5-centos
+docker exec -it myinternlm bash
+```
diff --git a/doc/install.md b/doc/install.md
index 5af3a95..d5e547e 100644
--- a/doc/install.md
+++ b/doc/install.md
@@ -5,10 +5,10 @@
 - Python == 3.10
 - GCC == 10.2.0
 - MPFR == 4.1.0
-- CUDA == 11.7
-- Pytorch == 1.13.1+cu117
+- CUDA >= 11.7
+- Pytorch >= 1.13.1
 - Transformers >= 4.28.0
-- Flash-Attention == v1.0.5
+- Flash-Attention >= v1.0.5
 - Apex == 23.05
 - Ampere或者Hopper架构的GPU (例如H100, A100)
 - Linux OS
@@ -57,3 +57,13 @@ cd ./third_party/apex
 pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
 cd ../../
 ```
+
+### 环境镜像
+用户可以从 https://hub.docker.com/r/sunpengsdu/internlm 获取安装了 InternLM 运行环境的镜像，拉取镜像及启动容器的命令如下：
+```bash
+# 拉取镜像
+docker pull sunpengsdu/internlm:torch1.13-cuda11.7-flashatten1.0.5-centos
+# 启动容器
+docker run --gpus all -d -it --shm-size=2gb --name myinternlm sunpengsdu/internlm:torch1.13-cuda11.7-flashatten1.0.5-centos
+docker exec -it myinternlm bash
+```
diff --git a/internlm/solver/lr_scheduler.py b/internlm/solver/lr_scheduler.py
index 2e228aa..bcbca88 100644
--- a/internlm/solver/lr_scheduler.py
+++ b/internlm/solver/lr_scheduler.py
@@ -27,7 +27,7 @@ class WarmupScheduler(_LRScheduler):
 
     def state_dict(self):
         state_dict = {key: value for key, value in self.__dict__.items() if key not in "optimizer"}
-        if isinstance(state_dict["after_scheduler"], _LRScheduler):
+        if isinstance(state_dict["after_scheduler"], (_LRScheduler, _CosineAnnealingLR)):
             state_dict["after_scheduler_type"] = type(state_dict["after_scheduler"]).__name__
             state_dict["after_scheduler_dict"] = state_dict["after_scheduler"].state_dict()
             del state_dict["after_scheduler"]
@@ -40,7 +40,7 @@ class WarmupScheduler(_LRScheduler):
         for key in list(self.__dict__.keys()):
             if key in state_dict:
                 self.__dict__[key] = state_dict[key]
-        if isinstance(self.after_scheduler, _LRScheduler):
+        if isinstance(self.after_scheduler, (_LRScheduler, _CosineAnnealingLR)):
             assert type(self.after_scheduler).__name__ == state_dict["after_scheduler_type"]
             # state_dict['after_scheduler_dict'] = state_dict['after_scheduler'].state_dict()
             self.after_scheduler.load_state_dict(state_dict["after_scheduler_dict"])
diff --git a/internlm/solver/optimizer/hybrid_zero_optim.py b/internlm/solver/optimizer/hybrid_zero_optim.py
index 7b19ecb..3ee2270 100644
--- a/internlm/solver/optimizer/hybrid_zero_optim.py
+++ b/internlm/solver/optimizer/hybrid_zero_optim.py
@@ -1,13 +1,13 @@
 #!/usr/bin/env python
 # -*- encoding: utf-8 -*-
 
+import math
 from functools import partial
 
 import amp_C
 import torch
 import torch.distributed as dist
 from apex.multi_tensor_apply import multi_tensor_applier
-from torch._six import inf
 from torch.optim import Optimizer
 
 from internlm.core.context import Config, ParallelMode
@@ -33,6 +33,7 @@ from internlm.utils.logger import get_logger
 from internlm.utils.megatron_timers import megatron_timer as timer
 from internlm.utils.parallel import is_model_parallel_parameter
 
+inf = math.inf
 logger = get_logger(__file__)