Update docs

README.md

@@ -27,6 +27,7 @@ Download prebuilt binaries from [Release](https://github.com/kendryte/nncase/rel
 
 - [TFLite ops](./docs/tflite_ops.md)
 - [Caffe ops](./docs/caffe_ops.md)
+- [ONNX ops](./docs/onnx_ops.md)
 
 ## Usage
 ## 使用方法

docs/USAGE_EN.md

@@ -21,7 +21,7 @@ OPTIONS
     compile
         <input file>        input file
         <output file>       output file
-        -i, --input-format  input file format: e.g. tflite, caffe
+        -i, --input-format  input file format: e.g. tflite, caffe, onnx
         -o, --output-format output file format: e.g. kmodel, default is kmodel
         -t, --target        target arch: e.g. cpu, k210, default is k210
         --dataset           calibration dataset, used in post quantization
@@ -55,10 +55,10 @@ OPTIONS
 
 `ncc` is the nncase command line tool. It has two commands: `compile` and `infer`.
 
-`compile` command compile your trained models (`.tflite`, `.caffemodel`) to `.kmodel`.
+`compile` command compile your trained models (`.tflite`, `.caffemodel`, `.onnx`) to `.kmodel`.
 - `<input file>` is your input model path.
 - `<output file>` is the output model path.
-- `-i, --input-format` option is used to specify the input model format. nncase supports `tflite` and `caffe` input model currently.
+- `-i, --input-format` option is used to specify the input model format. nncase supports `tflite`, `caffe` and `onnx` input model currently.
 - `-o, --output-format` option is used to sepecify the output model format. You have only one choice: `kmodel` currently.
 - `-t, --target` option is used to set your desired target device to run the model. `cpu` is the most general target that almost every platform should support. `k210` is the Kendryte K210 SoC platform. If you set this option to `k210`, this model can only run on K210 or be emulated on your PC.
 - `--inference-type` Set to `float` if you want precision, but you need more memory and lost K210 KPU acceleration. Set to `uint8` if you want KPU acceleration and fast speed and you need to provide a quantization calibration dataset to quantize your models later.

docs/USAGE_ZH.md

@@ -21,7 +21,7 @@ OPTIONS
     compile
         <input file>        input file
         <output file>       output file
-        -i, --input-format  input file format: e.g. tflite, caffe
+        -i, --input-format  input file format: e.g. tflite, caffe, onnx
         -o, --output-format output file format: e.g. kmodel, default is kmodel
         -t, --target        target arch: e.g. cpu, k210, default is k210
         --dataset           calibration dataset, used in post quantization
@@ -55,10 +55,10 @@ OPTIONS
 
 `ncc` 是 nncase 的命令行工具。它有两个命令： `compile` 和 `infer`。
 
-`compile` 命令将你训练好的模型 (`.tflite`, `.caffemodel`) 编译到 `.kmodel`。
+`compile` 命令将你训练好的模型 (`.tflite`, `.caffemodel`, `.onnx`) 编译到 `.kmodel`。
 - `<input file>` 是你输入模型的路径。
 - `<output file>` 是输出模型的路径。
-- `-i, --input-format` 用来指定输入模型的格式。nncase 现在支持 `tflite` 和 `caffe` 输入格式。
+- `-i, --input-format` 用来指定输入模型的格式。nncase 现在支持 `tflite`、`caffe` 和 `onnx` 输入格式。
 - `-o, --output-format` 用来指定输出模型的格式。你现在只有一个选项：`kmodel`。
 - `-t, --target` 用来指定你想要你的模型在哪种目标设备上运行。`cpu` 几乎所有平台都支持的通用目标。`k210` 是 Kendryte K210 SoC 平台。如果你指定了 `k210`，这个模型就只能在 K210 运行或在你的 PC 上模拟运行。
 - `--inference-type` 如果你需要精度，设置为 `float`，但你需要更多内存并且失去了 K210 KPU 的加速能力。如果你需要 KPU 加速和更快的执行速度，设置为 `uint8`，之后你需要提供量化校准集来量化你的模型。

docs/onnx_ops.md

@@ -0,0 +1,41 @@
+## Supported ONNX ops
+## 支持的 ONNX 算子
+
+| Operator | Is Supported |
+|-------|------------------ |
+| Abs |✅|
+| Add |✅|
+| AveragePool |✅|
+| Ceil |✅|
+| Concat |✅|
+| Constant |✅|
+| Conv |✅|
+| ConvTranspose |✅|
+| Cos |✅|
+| Div |✅|
+| Exp |✅|
+| Floor |✅|
+| LpNormalization |✅|
+| LeakyRelu |✅|
+| Log |✅|
+| MatMul |✅|
+| MaxPool |✅|
+| Max |✅|
+| Min |✅|
+| Mul |✅|
+| Neg |✅|
+| Pad |✅|
+| ReduceMax |✅|
+| ReduceMean |✅|
+| ReduceMin |✅|
+| ReduceSum |✅|
+| Relu |✅|
+| Reshape |✅|
+| Resize |✅|
+| Sin |✅|
+| Slice |✅|
+| Softmax |✅|
+| Sqrt |✅|
+| Squeeze |✅|
+| Sub |✅|
+| Transpose |✅|