[Book] Add small error management + start training (with generic dataset

inclusion).

candle-book/src/SUMMARY.md

@@ -12,7 +12,11 @@
 
 - [Running a model](inference/README.md)
     - [Using the hub](inference/hub.md)
-- [Error management]()
+- [Error management](error_manage.md)
+- [Training](training/README.md)
+    - [MNIST]()
+    - [Fine-tuning]()
+    - [Serialization]()
 - [Advanced Cuda usage]()
     - [Writing a custom kernel]()
     - [Porting a custom kernel]()
@@ -21,7 +25,3 @@
     - [Creating a WASM app]()
     - [Creating a REST api webserver]()
     - [Creating a desktop Tauri app]()
-- [Training]()
-    - [MNIST]()
-    - [Fine-tuning]()
-    - [Serialization]()

candle-book/src/training/README.md

@@ -1 +1,17 @@
 # Training
+
+
+Training starts with data. We're going to use the huggingface hub and 
+start with the Hello world dataset of machine learning, MNIST.
+
+Let's start with downloading `MNIST` from [huggingface](https://huggingface.co/datasets/mnist).
+
+
+```rust
+use candle_datasets::from_hub;
+
+
+let dataset = from_hub("mnist")?;
+```
+
+This uses the standardized `parquet` files from the `refs/convert/parquet` branch on every dataset.