Starting the book.

.github/workflows/book.yml

@@ -24,6 +24,6 @@ jobs:
         curl -sSL $url | tar -xz --directory=bin
         echo "$(pwd)/bin" >> $GITHUB_PATH
     - name: Run tests
-      run: cd candle-book && mdbook test
+      run: cd candle-book && cargo build && mdbook test -L ../target/debug/deps/
 
 

README.md

@@ -48,6 +48,8 @@ trunk serve --release --public-url /candle-llama2/ --port 8081
 And then browse to
 [http://localhost:8081/candle-llama2](http://localhost:8081/candle-llama2).
 
+<!--- ANCHOR: features --->
+
 ## Features
 
 - Simple syntax, looks and like PyTorch.
@@ -60,8 +62,11 @@ And then browse to
 - Embed user-defined ops/kernels, such as [flash-attention
   v2](https://github.com/LaurentMazare/candle/blob/89ba005962495f2bfbda286e185e9c3c7f5300a3/candle-flash-attn/src/lib.rs#L152).
 
+<!--- ANCHOR_END: features --->
+
 ## How to use ?
 
+<!--- ANCHOR: cheatsheet --->
 Cheatsheet:
 
 |            | Using PyTorch                            | Using Candle                                                     |
@@ -76,6 +81,8 @@ Cheatsheet:
 | Saving     | `torch.save({"A": A}, "model.bin")`      | `tensor.save_safetensors("A", "model.safetensors")?`             |
 | Loading    | `weights = torch.load("model.bin")`      | TODO (see the examples for now)                                  |
 
+<!--- ANCHOR_END: cheatsheet --->
+
 
 ## Structure
 

candle-book/src/README.md

@@ -1 +1,6 @@
 # Introduction
+
+{{#include ../../README.md:features}}
+
+
+This book will introduce step by step how to use `candle`.

candle-book/src/SUMMARY.md

@@ -6,13 +6,13 @@
 
 - [Installation](guide/installation.md)
 - [Hello World - MNIST](guide/hello_world.md)
-- [PyTorch cheatsheet](guide/hello_world.md)
+- [PyTorch cheatsheet](guide/cheatsheet.md)
 
 # Reference Guide
 
 - [Running a model](inference/README.md)
-    - [Serialization](inference/serialization.md)
     - [Using the hub](inference/hub.md)
+    - [Serialization](inference/serialization.md)
     - [Advanced Cuda usage](inference/cuda/README.md)
         - [Writing a custom kernel](inference/cuda/writing.md)
         - [Porting a custom kernel](inference/cuda/porting.md)
@@ -24,3 +24,4 @@
 - [Training](training/README.md)
     - [MNIST](training/mnist.md)
     - [Fine-tuning](training/finetuning.md)
+- [Using MKL](advanced/mkl.md)

candle-book/src/advanced/mkl.md

@@ -0,0 +1 @@
+# Using MKL

candle-book/src/guide/cheatsheet.md

@@ -0,0 +1,3 @@
+# Pytorch cheatsheet
+
+{{#include ../../../README.md:cheatsheet}}

candle-book/src/guide/hello_world.md

@@ -1 +1,53 @@
-# PyTorch cheatsheet
+# Hello world !
+
+We will now create the hello world of the ML world, building a model capable of solving MNIST dataset.
+
+Open `src/main.rs` and fill in with these contents:
+
+```rust
+# extern crate candle;
+use candle::{DType, Device, Result, Tensor};
+
+struct Model {
+    first: Tensor,
+    second: Tensor,
+}
+
+impl Model {
+    fn forward(&self, image: &Tensor) -> Result<Tensor> {
+        let x = image.matmul(&self.first)?;
+        let x = x.relu()?;
+        x.matmul(&self.second)
+    }
+}
+
+fn main() -> Result<()> {
+    // Use Device::new_cuda(0)?; to use the GPU.
+    let device = Device::Cpu;
+
+    let first = Tensor::zeros((784, 100), DType::F32, &device)?;
+    let second = Tensor::zeros((100, 10), DType::F32, &device)?;
+    let model = Model { first, second };
+
+    let dummy_image = Tensor::zeros((1, 784), DType::F32, &device)?;
+
+    let digit = model.forward(&dummy_image)?;
+    println!("Digit {digit:?} digit");
+    Ok(())
+}
+```
+
+Everything should now run with:
+
+```bash
+cargo run --release
+```
+
+Now that we have the running dummy code we can get to more advanced topics:
+
+
+- [For PyTorch users](./guide/cheatsheet.md)
+- [Running existing models](./inference/README.md)
+- [Training models](./training/README.md)
+
+

candle-book/src/guide/installation.md

@@ -1 +1,24 @@
 # Installation
+
+Start by creating a new app:
+
+```bash
+cargo new myapp
+cd myapp
+cargo add --git https://github.com/LaurentMazare/candle.git candle
+```
+
+At this point, candle will be built **without** CUDA support.
+To get CUDA support use the feature `cuda`
+```bash
+cargo add --git https://github.com/LaurentMazare/candle.git candle --features cuda
+```
+
+You can check everything works properly:
+
+```bash
+cargo build
+```
+
+
+You can also see feature `mkl` which could be interesting to get faster inference on CPU. [Using mkl](./advanced/mkl.md)