Added readmes to examples (#2835)

* added chatGLM readme

* changed wording in readme

* added readme for chinese-clip

* added readme for convmixer

* added readme for custom ops

* added readme for efficientnet

* added readme for llama

* added readme to mnist-training

* added readme to musicgen

* added readme to quantized-phi

* added readme to starcoder2

* added readme to whisper-microphone

* added readme to yi

* added readme to yolo-v3

* added readme to whisper-microphone

* added space to example in glm4 readme

* fixed mamba example readme to run mamba instead of mamba-minimal

* removed slash escape character

* changed moondream image to yolo-v8 example image

* added procedure for making the reinforcement-learning example work with a virtual environment on my machine

* added simple one line summaries to the example readmes without

* changed non-existant image to yolo example's bike.jpg

* added backslash to sam command

* removed trailing - from siglip

* added SoX to silero-vad example readme

* replaced procedure for uv on mac with warning that uv isn't currently compatible with pyo3

* added example to falcon readme

* added --which arg to stella-en-v5 readme

* fixed image path in vgg readme

* fixed the image path in the vit readme

* Update README.md

* Update README.md

* Update README.md

---------

Co-authored-by: Laurent Mazare <laurent.mazare@gmail.com>

candle-examples/examples/chatglm/README.md

@@ -0,0 +1,13 @@
+# candle-chatglm
+
+Uses `THUDM/chatglm3-6b` to generate chinese text. Will not generate text for english (usually).
+ 
+## Text Generation
+
+```bash
+cargo run --example chatglm --release  -- --prompt "部署门槛较低等众多优秀特 "
+
+> 部署门槛较低等众多优秀特 点，使得其成为了一款备受欢迎的AI助手。
+> 
+> 作为一款人工智能助手，ChatGLM3-6B
+```

candle-examples/examples/chinese_clip/README.md

@@ -0,0 +1,42 @@
+# candle-chinese-clip
+
+Contrastive Language-Image Pre-Training (CLIP) is an architecture trained on
+pairs of images with related texts. This one is trained using in chinese instead of english.
+
+## Running on cpu
+
+```bash
+$ cargo run --example chinese_clip --release -- --images "candle-examples/examples/stable-diffusion/assets/stable-diffusion-xl.jpg","candle-examples/examples/yolo-v8/assets/bike.jpg" --cpu --sequences "一场自行车比赛","两只猫的照片","一个机器人拿着蜡烛"
+
+> Results for image: candle-examples/examples/stable-diffusion/assets/stable-diffusion-xl.jpg
+>
+> 2025-03-25T19:22:01.325177Z  INFO chinese_clip: Probability: 0.0000% Text: 一场自行车比赛 
+> 2025-03-25T19:22:01.325179Z  INFO chinese_clip: Probability: 0.0000% Text: 两只猫的照片 
+> 2025-03-25T19:22:01.325181Z  INFO chinese_clip: Probability: 100.0000% Text: 一个机器人拿着蜡烛 
+> 2025-03-25T19:22:01.325183Z  INFO chinese_clip: 
+> 
+> Results for image: candle-examples/examples/yolo-v8/assets/bike.jpg
+> 
+> 2025-03-25T19:22:01.325184Z  INFO chinese_clip: Probability: 100.0000% Text: 一场自行车比赛 
+> 2025-03-25T19:22:01.325186Z  INFO chinese_clip: Probability: 0.0000% Text: 两只猫的照片 
+> 2025-03-25T19:22:01.325187Z  INFO chinese_clip: Probability: 0.0000% Text: 一个机器人拿着蜡烛 
+```
+
+## Running on metal
+
+```bash 
+$ cargo run --features metal --example chinese_clip --release -- --images "candle-examples/examples/stable-diffusion/assets/stable-diffusion-xl.jpg","candle-examples/examples/yolo-v8/assets/bike.jpg" --cpu --sequences "一场自行车比赛","两只猫的照片","一个机器人拿着蜡烛"
+
+> Results for image: candle-examples/examples/stable-diffusion/assets/stable-diffusion-xl.jpg
+>
+> 2025-03-25T19:22:01.325177Z  INFO chinese_clip: Probability: 0.0000% Text: 一场自行车比赛 
+> 2025-03-25T19:22:01.325179Z  INFO chinese_clip: Probability: 0.0000% Text: 两只猫的照片 
+> 2025-03-25T19:22:01.325181Z  INFO chinese_clip: Probability: 100.0000% Text: 一个机器人拿着蜡烛 
+> 2025-03-25T19:22:01.325183Z  INFO chinese_clip: 
+> 
+> Results for image: candle-examples/examples/yolo-v8/assets/bike.jpg
+> 
+> 2025-03-25T19:22:01.325184Z  INFO chinese_clip: Probability: 100.0000% Text: 一场自行车比赛 
+> 2025-03-25T19:22:01.325186Z  INFO chinese_clip: Probability: 0.0000% Text: 两只猫的照片 
+> 2025-03-25T19:22:01.325187Z  INFO chinese_clip: Probability: 0.0000% Text: 一个机器人拿着蜡烛 
+```

candle-examples/examples/convmixer/README.md

@@ -0,0 +1,17 @@
+# candle-convmixer
+
+A lightweight CNN architecture that processes image patches similar to a vision transformer, with separate spatial and channel convolutions.
+
+ConvMixer from [Patches Are All You Need?](https://arxiv.org/pdf/2201.09792) and [ConvMixer](https://github.com/locuslab/convmixer). 
+
+## Running an example
+
+```bash
+$ cargo run --example convmixer --release -- --image candle-examples/examples/yolo-v8/assets/bike.jpg
+
+> mountain bike, all-terrain bike, off-roader: 61.75%
+> unicycle, monocycle     : 5.73%
+> moped                   : 3.66%
+> bicycle-built-for-two, tandem bicycle, tandem: 3.51%
+> crash helmet            : 0.85%
+```

candle-examples/examples/custom-ops/README.md

@@ -0,0 +1,17 @@
+# candle-custom-ops
+
+ This example illustrates how to implement forward and backward passes for custom operations on the CPU and GPU.
+ The custom op in this example implements RMS normalization for the CPU and CUDA.
+ 
+## Running an example
+
+```bash
+$ cargo run --example custom-ops
+
+> [[ 0.,  1.,  2.,  3.,  4.,  5.,  6.],
+>  [ 7.,  8.,  9., 10., 11., 12., 13.]]
+> Tensor[[2, 7], f32]
+> [[0.0000, 0.2773, 0.5547, 0.8320, 1.1094, 1.3867, 1.6641],
+>  [0.6864, 0.7845, 0.8825, 0.9806, 1.0786, 1.1767, 1.2748]]
+> Tensor[[2, 7], f32]
+```

candle-examples/examples/efficientnet/README.md

@@ -0,0 +1,15 @@
+# candle-efficientnet
+
+Demonstrates a Candle implementation of EfficientNet for image classification based on ImageNet classes.
+
+## Running an example
+
+```bash
+$ cargo run --example efficientnet --release -- --image candle-examples/examples/yolo-v8/assets/bike.jpg --which b1
+
+> bicycle-built-for-two, tandem bicycle, tandem: 45.85%
+> mountain bike, all-terrain bike, off-roader: 30.45%
+> crash helmet            : 2.58%
+> unicycle, monocycle     : 2.21%
+> tricycle, trike, velocipede: 1.53%
+```

candle-examples/examples/falcon/README.md

@@ -1,3 +1,10 @@
 # candle-falcon
 
 Falcon is a general large language model.
+
+## Running an example
+
+Make sure to include the `--use-f32` flag if using CPU, because there isn't a BFloat16 implementation yet.
+```
+cargo run --example falcon --release -- --prompt "Flying monkeys are" --use-f32
+```

candle-examples/examples/llama/README.md

@@ -0,0 +1,11 @@
+# candle-llama
+
+Candle implementations of various Llama based architectures.
+
+## Running an example
+
+```bash
+$ cargo run --example llama -- --prompt "Machine learning is " --which v32-3b-instruct
+
+> Machine learning is  the part of computer science which deals with the development of algorithms and
+```

candle-examples/examples/mamba/README.md

@@ -12,6 +12,6 @@ would only work for inference.
 ## Running the example
 
 ```bash
-$ cargo run --example mamba-minimal --release -- --prompt "Mamba is the"
+$ cargo run --example mamba --release -- --prompt "Mamba is the"
 ```
 

candle-examples/examples/metavoice/README.md

@@ -13,6 +13,6 @@ Note that the current candle implementation suffers from some limitations as of
 ## Run an example
 
 ```bash
-cargo run --example metavoice --release -- \\
+cargo run --example metavoice --release -- \
   --prompt "This is a demo of text to speech by MetaVoice-1B, an open-source foundational audio model."
 ```

candle-examples/examples/mnist-training/README.md

@@ -0,0 +1,16 @@
+# candle-mnist-training
+
+Training a 2 layer MLP on mnist in Candle.
+
+## Running an example
+
+```bash
+$ cargo run --example mnist-training --features candle-datasets
+
+> train-images: [60000, 784]
+> train-labels: [60000]
+> test-images: [10000, 784]
+> test-labels: [10000]
+>    1 train loss:  2.30265 test acc: 68.08%
+>    2 train loss:  1.50815 test acc: 60.77%
+```

candle-examples/examples/moondream/README.md

@@ -12,7 +12,7 @@ $ wget https://raw.githubusercontent.com/vikhyat/moondream/main/assets/demo-1.jp
 
 Now you can run Moondream from the `candle-examples` crate:
 ```bash
-$ cargo run --example moondream --release -- --prompt "What is the girl eating?" --image "./demo-1.jpg"
+$ cargo run --example moondream --release -- --prompt "Describe the people behind the bikers?" --image "candle-examples/examples/yolo-v8/assets/bike.jpg"
 
 avavx: false, neon: true, simd128: false, f16c: false
 temp: 0.00 repeat-penalty: 1.00 repeat-last-n: 64

candle-examples/examples/musicgen/README.md

@@ -0,0 +1,20 @@
+# candle-musicgen
+
+Candle implementation of musicgen from [Simple and Controllable Music Generation](https://arxiv.org/pdf/2306.05284).
+
+## Running an example
+
+```bash
+$ cargo run --example musicgen -- --prompt "90s rock song with loud guitars and heavy drums"
+
+> tokens: [2777, 7, 2480, 2324, 28, 8002, 5507, 7, 11, 2437, 5253, 7, 1]
+> Tensor[dims 1, 13; u32]
+> [[[ 0.0902,  0.1256, -0.0585, ...,  0.1057, -0.5141, -0.4675],
+>   [ 0.1972, -0.0268, -0.3368, ..., -0.0495, -0.3597, -0.3940],
+>   [-0.0855, -0.0007,  0.2225, ..., -0.2804, -0.5360, -0.2436],
+>   ...
+>   [ 0.0515,  0.0235, -0.3855, ..., -0.4728, -0.6858, -0.2923],
+>   [-0.3728, -0.1442, -0.1179, ..., -0.4388, -0.0287, -0.3242],
+>   [ 0.0163,  0.0012, -0.0020, ...,  0.0142,  0.0173, -0.0103]]]
+> Tensor[[1, 13, 768], f32]
+```

candle-examples/examples/quantized-phi/README.md

@@ -0,0 +1,20 @@
+# candle-quantized-phi
+
+Candle implementation of various quantized Phi models.
+
+## Running an example
+
+```bash
+$ cargo run --example quantized-phi --release -- --prompt "The best thing about coding in rust is "
+
+> - it's memory safe (without you having to worry too much) 
+> - the borrow checker is really smart and will catch your mistakes for free, making them show up as compile errors instead of segfaulting in runtime.
+> 
+> This alone make me prefer using rust over c++ or go, python/Cython etc.
+> 
+> The major downside I can see now: 
+> - it's slower than other languages (viz: C++) and most importantly lack of libraries to leverage existing work done by community in that language. There are so many useful machine learning libraries available for c++, go, python etc but none for Rust as far as I am aware of on the first glance. 
+> - there aren't a lot of production ready projects which also makes it very hard to start new one (given my background)
+> 
+> Another downside:
+```

candle-examples/examples/quantized-t5/README.md

@@ -1,5 +1,7 @@
 # candle-quantized-t5
 
+Candle implementation for quantizing and running T5 translation models.
+
 ## Seq2Seq example
 
 This example uses a quantized version of the t5 model.

candle-examples/examples/reinforcement-learning/README.md

@@ -2,6 +2,11 @@
 
 Reinforcement Learning examples for candle.
 
+> [!WARNING]  
+> uv is not currently compatible with pyo3 as of 2025/3/28.
+
+## System wide python
+
 This has been tested with `gymnasium` version `0.29.1`. You can install the
 Python package with:
 ```bash

candle-examples/examples/resnet/README.md

@@ -7,7 +7,7 @@ probabilities for the top-5 classes.
 ## Running an example
 
 ```
-$ cargo run --example resnet --release -- --image tiger.jpg
+$ cargo run --example resnet --release -- --image candle-examples/examples/yolo-v8/assets/bike.jpg
 
 loaded image Tensor[dims 3, 224, 224; f32]
 model built

candle-examples/examples/segformer/README.md

@@ -10,9 +10,11 @@ If you want you can use the example images from this [pull request][pr], downloa
 
 ```bash
 # run the image classification task
-cargo run --example segformer classify <path-to-image>
+cargo run --example segformer classify candle-examples/examples/yolo-v8/assets/bike.jpg
+
 # run the segmentation task
-cargo run --example segformer segment <path-to-image>
+cargo run --example segformer segment candle-examples/examples/yolo-v8/assets/bike.jpg
+
 ```
 
 Example output for classification:

candle-examples/examples/segment-anything/README.md

@@ -14,8 +14,8 @@ based on [MobileSAM](https://github.com/ChaoningZhang/MobileSAM).
 
 ```bash
 cargo run --example segment-anything --release -- \
-    --image candle-examples/examples/yolo-v8/assets/bike.jpg
-    --use-tiny
+    --image candle-examples/examples/yolo-v8/assets/bike.jpg \
+    --use-tiny \
     --point 0.6,0.6 --point 0.6,0.55
 ```
 

candle-examples/examples/siglip/README.md

@@ -5,7 +5,7 @@ SigLIP is multi-modal text-vision model that improves over CLIP by using a sigmo
 
 ### Running an example
 ```
-$ cargo run --features cuda -r --example siglip -
+$ cargo run --features cuda -r --example siglip
 softmax_image_vec: [2.1912122e-14, 2.3624872e-14, 1.0, 1.0, 2.4787932e-8, 3.2784535e-12]
 
 

candle-examples/examples/silero-vad/README.md

@@ -6,7 +6,14 @@ This example uses the models available in the hugging face [onnx-community/siler
 
 ## Running the example
 
+### using arecord
+
 ```bash
 $ arecord -t raw -f S16_LE -r 16000 -c 1 -d 5 - | cargo run --example silero-vad --release --features onnx -- --sample-rate 16000
 ```
 
+### using SoX
+
+```bash
+$ rec -t raw -r 48000 -b 16 -c 1 -e signed-integer - trim 0 5 | sox -t raw -r 48000 -b 16 -c 1 -e signed-integer - -t raw -r 16000 -b 16 -c 1 -e signed-integer - | cargo run --example silero-vad --release --features onnx -- --sample-rate 16000
+```

candle-examples/examples/starcoder2/README.md

@@ -0,0 +1,15 @@
+# candle-starcoder2
+
+Candle implementation of Star Coder 2 family of code generation model from [StarCoder 2 and The Stack v2: The Next Generation](https://arxiv.org/pdf/2402.19173).
+
+## Running an example
+
+```bash
+$ cargo run --example starcoder2 -- --prompt "write a recursive fibonacci function in python "
+
+> # that returns the nth number in the sequence.
+> 
+> def fib(n):
+>     if n
+
+```

candle-examples/examples/stella-en-v5/README.md

@@ -10,7 +10,7 @@ Stella_en_1.5B_v5 is used to generate text embeddings embeddings for a prompt. T
 are downloaded from the hub on the first run.
 
 ```bash
-$ cargo run --example stella-en-v5 --release  -- --query "What are safetensors?"
+$ cargo run --example stella-en-v5 --release  -- --query "What are safetensors?" --which 1.5b
 
 > [[ 0.3905, -0.0130,  0.2072, ..., -0.1100, -0.0086,  0.6002]]
 >  Tensor[[1, 1024], f32]

candle-examples/examples/t5/README.md

@@ -1,5 +1,7 @@
 # candle-t5
 
+Candle implementations of the T5 family of translation models.
+
 ## Encoder-decoder example:
 
 ```bash

candle-examples/examples/vgg/README.md

@@ -7,7 +7,7 @@ The VGG models are defined in `candle-transformers/src/models/vgg.rs`. The main
 You can run the example with the following command:
 
 ```bash
-cargo run --example vgg --release -- --image ../yolo-v8/assets/bike.jpg --which vgg13
+cargo run --example vgg --release -- --image candle-examples/examples/yolo-v8/assets/bike.jpg --which vgg13
 ```
 
 In the command above, `--image` specifies the path to the image file and `--which` specifies the VGG model to use (vgg13, vgg16, or vgg19).

candle-examples/examples/vit/README.md

@@ -7,8 +7,8 @@ probabilities for the top-5 classes.
 
 ## Running an example
 
-```
-$ cargo run --example vit --release -- --image tiger.jpg
+```bash
+$ cargo run --example vit --release -- --image candle-examples/examples/yolo-v8/assets/bike.jpg
 
 loaded image Tensor[dims 3, 224, 224; f32]
 model built

candle-examples/examples/whisper-microphone/README.md

@@ -0,0 +1,15 @@
+# candle-whisper-microphone
+
+Whisper implementation using microphone as input.
+
+## Running an example
+
+```bash
+$ cargo run --example whisper-microphone --features microphone
+
+> transcribing audio...
+> 480256 160083
+> language_token: None
+> 0.0s -- 30.0s:  Hello, hello, I don't know if this is working, but You know, how long did I make this?
+> 480256 160085
+```

candle-examples/examples/yi/README.md

@@ -0,0 +1,13 @@
+# candle-yi
+
+Candle implentations of the Yi family of bilingual (English, Chinese) LLMs.
+
+## Running an example
+
+```bash
+$ cargo run --example yi -- --prompt "Here is a test sentence"
+
+> python
+> print("Hello World")
+> 
+```

candle-examples/examples/yolo-v3/README.md

@@ -0,0 +1,32 @@
+# candle-yolo-v3:
+
+Candle implementation of Yolo-V3 for object detection.
+
+## Running an example
+
+```bash
+$ cargo run --example yolo-v3 --release -- candle-examples/examples/yolo-v8/assets/bike.jpg
+
+> generated predictions Tensor[dims 10647, 85; f32]
+> person: Bbox { xmin: 46.362198, ymin: 72.177, xmax: 135.92522, ymax: 339.8356, confidence: 0.99705493, data: () }
+> person: Bbox { xmin: 137.25645, ymin: 67.58148, xmax: 216.90437, ymax: 333.80756, confidence: 0.9898516, data: () }
+> person: Bbox { xmin: 245.7842, ymin: 82.76726, xmax: 316.79053, ymax: 337.21613, confidence: 0.9884322, data: () }
+> person: Bbox { xmin: 207.52783, ymin: 61.815224, xmax: 266.77884, ymax: 307.92606, confidence: 0.9860648, data: () }
+> person: Bbox { xmin: 11.457404, ymin: 60.335564, xmax: 34.39357, ymax: 187.7714, confidence: 0.9545012, data: () }
+> person: Bbox { xmin: 251.88353, ymin: 11.235481, xmax: 286.56607, ymax: 92.54697, confidence: 0.8439807, data: () }
+> person: Bbox { xmin: -0.44309902, ymin: 55.486923, xmax: 13.160354, ymax: 184.09705, confidence: 0.8266243, data: () }
+> person: Bbox { xmin: 317.40826, ymin: 55.39501, xmax: 370.6704, ymax: 153.74887, confidence: 0.7327442, data: () }
+> person: Bbox { xmin: 370.02835, ymin: 66.120224, xmax: 404.22824, ymax: 142.09691, confidence: 0.7265741, data: () }
+> person: Bbox { xmin: 250.36511, ymin: 57.349842, xmax: 280.06335, ymax: 116.29384, confidence: 0.709422, data: () }
+> person: Bbox { xmin: 32.573215, ymin: 66.66239, xmax: 50.49056, ymax: 173.42068, confidence: 0.6998766, data: () }
+> person: Bbox { xmin: 131.72215, ymin: 63.946213, xmax: 166.66151, ymax: 241.52773, confidence: 0.64457536, data: () }
+> person: Bbox { xmin: 407.42416, ymin: 49.106407, xmax: 415.24307, ymax: 84.7134, confidence: 0.5955802, data: () }
+> person: Bbox { xmin: 51.650482, ymin: 64.4985, xmax: 67.40904, ymax: 106.952385, confidence: 0.5196007, data: () }
+> bicycle: Bbox { xmin: 160.10031, ymin: 183.90837, xmax: 200.86832, ymax: 398.609, confidence: 0.9623588, data: () }
+> bicycle: Bbox { xmin: 66.570915, ymin: 192.56966, xmax: 112.06765, ymax: 369.28497, confidence: 0.9174347, data: () }
+> bicycle: Bbox { xmin: 258.2856, ymin: 197.04532, xmax: 298.43106, ymax: 364.8627, confidence: 0.6851388, data: () }
+> bicycle: Bbox { xmin: 214.0034, ymin: 175.76498, xmax: 252.45158, ymax: 356.53818, confidence: 0.67071193, data: () }
+> motorbike: Bbox { xmin: 318.23938, ymin: 95.22487, xmax: 369.9743, ymax: 213.46263, confidence: 0.96691036, data: () }
+> motorbike: Bbox { xmin: 367.46417, ymin: 100.07982, xmax: 394.9981, ymax: 174.6545, confidence: 0.9185384, data: () }
+> writing "candle-examples/examples/yolo-v8/assets/bike.pp.jpg"
+```