diff --git a/docs/blender/blender_tutorial.md b/docs/blender/blender_tutorial.md
index d30f089f3..f85b87903 100644
--- a/docs/blender/blender_tutorial.md
+++ b/docs/blender/blender_tutorial.md
@@ -4,15 +4,15 @@
 
 First set transparent rendering:
 
-![image-20230817121854933](D:\aa\Notebooks\docs\model\blender_tutorial.assets\image-20230817121854933.png)
+![image-20230817121854933](blender_tutorial.assets\image-20230817121854933.png)
 
 Then Change color view transform to standard (the same tab, just below `Film`):
 
-![image-20230817121902473](D:\aa\Notebooks\docs\model\blender_tutorial.assets\image-20230817121902473.png)
+![image-20230817121902473](blender_tutorial.assets\image-20230817121902473.png)
 
 Then open compositing window, and set use nodes, then add an "alpha over" layer like this:
 
-![image-20230817121909364](D:\aa\Notebooks\docs\model\blender_tutorial.assets\image-20230817121909364.png)
+![image-20230817121909364](blender_tutorial.assets\image-20230817121909364.png)
 
 
 ### Add-on
@@ -45,7 +45,6 @@ in the same folder, and a name keyword is necessary.
 
   ![image-20230914131227283](blender_tutorial.assets/image-20230914131227283.png)
 
-  
 * Adjust the bone's size, rotation, make it as the main spine of your model.
 
 * Go to edit mode:
@@ -141,19 +140,16 @@ In the world properties:
   * `5`: toggle perspective / orthogonal
   * `0`: toggle camera view
   
-  
 * Shift + A: add new objects
   
 * Shift + Z: toggle wireframe
   
 * Ctrl + Space: toggle maximize area with the current application window.
   
-   
 * Ctrl + Dragging: discrete steps
 
 * Shift + Dragging: precise steps
   
-  
 * Ctrl + Z: undo
 
 * Ctrl + Shift + Z: redo
diff --git a/docs/deeplearning/autocast.md b/docs/deeplearning/autocast.md
deleted file mode 100644
index f3f28f385..000000000
--- a/docs/deeplearning/autocast.md
+++ /dev/null
@@ -1,18 +0,0 @@
-### autocast
-
-Native functions in an `autocast` context will incur the `fp16` version.
-
-But if you want a function to be always high precision `fp32`, you should wrap it with:
-
-```python
-@torch.cuda.amp.autocast(enabled=False)
-def high_prec_func(x):
-    ...
-    return x
-
-# call 
-with torch.cuda.amp.autocast(enabled=True):
-    x = high_prec_func(x)
-    x = low_prec_func(x)
-```
-
diff --git a/docs/deeplearning/flax_tutorial.md b/docs/deeplearning/flax_tutorial.md
deleted file mode 100644
index 314362529..000000000
--- a/docs/deeplearning/flax_tutorial.md
+++ /dev/null
@@ -1,24 +0,0 @@
-## flax tutorial
-
-
-### install
-
-```bash
-pip install flax
-```
-
-
-### basic
-
-```python
-import jax
-from jax import lax, random, numpy as jnp
-
-import flax
-from flax.core import freeze, unfreeze
-from flax import linen as nn
-
-from typing import Any, Callable, Sequence, Optional
-```
-
-
diff --git a/docs/onnx/basics.md b/docs/deeplearning/onnx/basics.md
similarity index 100%
rename from docs/onnx/basics.md
rename to docs/deeplearning/onnx/basics.md
diff --git a/docs/onnx/model_visualization.md b/docs/deeplearning/onnx/model_visualization.md
similarity index 100%
rename from docs/onnx/model_visualization.md
rename to docs/deeplearning/onnx/model_visualization.md
diff --git a/docs/onnx/torch2onnx.md b/docs/deeplearning/onnx/torch2onnx.md
similarity index 100%
rename from docs/onnx/torch2onnx.md
rename to docs/deeplearning/onnx/torch2onnx.md
diff --git a/docs/onnx/web.md b/docs/deeplearning/onnx/web.md
similarity index 100%
rename from docs/onnx/web.md
rename to docs/deeplearning/onnx/web.md
diff --git a/docs/deeplearning/speech_recognition.md b/docs/deeplearning/speech_recognition.md
deleted file mode 100644
index 1dc4874b9..000000000
--- a/docs/deeplearning/speech_recognition.md
+++ /dev/null
@@ -1,7 +0,0 @@
-# Speech recognition
-
-### [CTC (Connectionist Temporal Classification)](https://distill.pub/2017/ctc/)
-
-A method to handle **sequence classification problems with unaligned label** (e.g., speech recognition, handwriting recognition).
-
-It's an alignment-free algorithm, by computing the probability of ALL possible alignments with dynamic programming.
\ No newline at end of file
diff --git a/docs/linux/ubuntu_install_list.md b/docs/linux/ubuntu_install_list.md
index 7b1099894..5e2b338ee 100644
--- a/docs/linux/ubuntu_install_list.md
+++ b/docs/linux/ubuntu_install_list.md
@@ -62,7 +62,6 @@ Then another long wait...
 
   TUNA source: https://mirrors.tuna.tsinghua.edu.cn/help/ubuntu/
 
-  
 * VPN
 
   ```bash
@@ -271,7 +270,8 @@ Then another long wait...
 
   ```bash
   # download the correct runfile and run it to install
-  # use sh, not bash, and do not run with sudo.
+  # use sh, not bash
+  # use root to run, do not run with sudo (will silently quit)
   sh cuda.xxxx.sh
   # just make sure the cuda version is compatible to the driver... (your driver is newer to the default one)
   # create soft link in /user/local/cuda
diff --git a/docs/math/ddpm.md b/docs/math/ddpm.md
deleted file mode 100644
index 4124c32a2..000000000
--- a/docs/math/ddpm.md
+++ /dev/null
@@ -1,12 +0,0 @@
-## Denoising Diffusion Probabilistic Model
-
-### forward
-
-
-$$
-\displaylines{
-x_t = \sqrt \alpha_tx_{t-1}
-}
-$$
-
-
diff --git a/docs/math/probability/distributions.md b/docs/math/probability/distributions.md
deleted file mode 100644
index 7eeda4323..000000000
--- a/docs/math/probability/distributions.md
+++ /dev/null
@@ -1,3 +0,0 @@
-* Gamma
-* Beta
-* Dirichlet
\ No newline at end of file
diff --git a/docs/math/real_analysis/01.md b/docs/math/real_analysis/01.md
deleted file mode 100644
index 5626e9d0b..000000000
--- a/docs/math/real_analysis/01.md
+++ /dev/null
@@ -1,2 +0,0 @@
-# Peano Axiom
-
diff --git a/docs/minecraft/server.md b/docs/minecraft/server.md
deleted file mode 100644
index 1ba796b18..000000000
--- a/docs/minecraft/server.md
+++ /dev/null
@@ -1,5 +0,0 @@
-## Server
-
-如何组建一个基本的服务器：
-
-* HMCL启动器：https://github.com/HMCL-dev/HMCL
\ No newline at end of file