wonnx-preprocessing 0.5.0

Preprocessing utility crate for WONNX. WONNX is an ONNX runtime based on wgpu aimed at being a universal GPU runtime, written in Rust.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
<center><img src="logo.svg" alt="WONNX" width="700"/></center>

![GitHub Workflow Status](https://img.shields.io/github/actions/workflow/status/webonnx/wonnx/rust.yml?branch=master)
[![docs.rs](https://img.shields.io/docsrs/wonnx)](https://docs.rs/wonnx)
![Crates.io (latest)](https://img.shields.io/crates/dv/wonnx)
![Crates.io](https://img.shields.io/crates/l/wonnx)


Wonnx is a GPU-accelerated ONNX inference run-time written 100% in Rust, ready for the web.

## Supported Platforms (enabled by `wgpu`)

   API   |    Windows                    |  Linux & Android   |    macOS & iOS     |
  -----  | ----------------------------- | ------------------ | ------------------ |
  Vulkan | ✅                            | ✅                 |                    |
  Metal  |                               |                    | ✅                 |
  DX12   | ✅                 (W10 only) |                    |                    |
  DX11   | :construction:                |                    |                    |
  GLES3  |                               | :ok:               |                    |

:white_check_mark: = First Class Support — :ok: = Best Effort Support — :construction: = Unsupported, but support in progress

## Getting started

### From the command line

Ensure your system supports either Vulkan, Metal or DX12 for access to the GPU. Then either download a binary release,
or install Rust and run `cargo install --git https://github.com/webonnx/wonnx.git wonnx-cli` to install the CLI.

The CLI tool (`nnx`) provides a convenient interface for tinkering with models (see the [README](./wonnx-cli/README.md) for more information):

````bash
nnx info ./data/models/opt-squeeze.onnx
nnx infer ./data/models/opt-squeeze.onnx -i data=./data/images/pelican.jpeg --labels ./data/models/squeeze-labels.txt --top 3
````

### From Rust

Add the `wonnx` crate as dependency (`cargo add wonnx` if you have cargo-add). Then, see the [examples](./wonnx/examples)
for usage examples, or [browse the API docs](https://docs.rs/wonnx).

### From Python

```bash
pip install wonnx
```

And then, to use:

```python
from wonnx import Session
session = Session.from_path(
    "../data/models/single_relu.onnx"
)
inputs = {"x": [-1.0, 2.0]}
assert session.run(inputs) == {"y": [0.0, 2.0]}
```

Then run `python3` with the above Python code!

For more details on the Python package including build instructions, see [wonnx-py](./wonnx-py/README.md).

### In the browser, using WebGPU + WebAssembly

````bash
npm install @webonnx/wonnx-wasm
````

And then, on the client side:

````js
import init, { Session, Input } from "@webonnx/wonnx-wasm";

// Check for WebGPU availability first: if(navigator.gpu) { .. }
await init();
const session = await Session.fromBytes(modelBytes /* Uint8Array containing the ONNX file */);
const input = new Input();
input.insert("x", [13.0, -37.0]);
const result = await session.run(input); // This will be an object where the keys are the names of the model outputs and the values are arrays of numbers.
session.free();
input.free();
````

The package [@webonnx/wonnx-wasm](https://www.npmjs.com/package/@webonnx/wonnx-wasm) provides an interface to WONNX, 
which is included as WebAssembly module and will use the browser's WebGPU implementation. See [wonnx-wasm-example](https://github.com/webonnx/wonnx-wasm-example)
for a more complete usage example involving a bundler.

For more details on the JS/WASM package including build instructions, see [wonnx-wasm](./wonnx-wasm/README.md).

### For development

To work on wonnx itself, follow the following steps:

- Install Rust
- Install Vulkan, Metal, or DX12 for the GPU API.
- git clone this repo.

```bash 
git clone https://github.com/webonnx/wonnx.git
```

Then, you're all set! You can run one of the included examples through cargo:

```bash
cargo run --example squeeze --release
```

## Running other models

- To run an onnx model, first simplify it with `nnx prepare` (substitute with `cargo run -- prepare` when inside this repo):

````bash
nnx prepare -i ./some-model.onnx ./some-model-prepared.onnx
````

To specify dynamic dimension parameters, add e.g. `--set batch_size=1`.

You can also use an external tool, such as [onnx-simplifier](https://github.com/daquexian/onnx-simplifier), with the command:

```bash
# pip install -U pip && pip install onnx-simplifier
python -m onnxsim mnist-8.onnx opt-mnist.onnx
```

- Then you can run it using the CLI (see [README](./wonnx-cli/README.md) or programmatically, following the 
[examples in the examples folder](./wonnx/examples/). To run an example:

```bash
cargo run --example mnist --release
```

## Tested models

- Squeezenet
- MNIST
- BERT

## GPU selection

Except when running in WebAssembly, you may set the following environment variables to influence GPU selection by WGPU:

* `WGPU_ADAPTER_NAME` with a substring of the name of the adapter you want to use (e.g. `1080` will match `NVIDIA GeForce 1080ti`).
* `WGPU_BACKEND` with a comma separated list of the backends you want to use (`vulkan`, `metal`, `dx12`, `dx11`, or `gl`).
* `WGPU_POWER_PREFERENCE` with the power preference to choose when a specific adapter name isn't specified (`high` or `low`)

## Contribution: On implementing a new Operator

Contribution are very much welcomed even without large experience in DL, WGSL, or Rust. I hope that, this project can be a sandbox for all of us to learn more about those technologies beyond this project initial scope.

To implement an operator all you have to do is:
1. Add a new matching pattern in `compiler.rs`
2. Retrieve its attributes values using the `get_attribute` function:
```Rust
    let alpha = get_attribute("alpha", Some(1.0), node);
    // or without default value
    let alpha = get_attribute::<f32>("alpha", None, node);
```
3. Add any variable you want to use in the WGSL shader using `context`.
4. Write a new WGSL template in the `templates` folder.
> Available types are in `structs.wgsl` but you can also generate new ones within your templates.
5. Respect the binding layout that each entry is incremented by 1 starting from 0, with input first and output last. If the number of binding is above 4. Increment the binding group. You can change the input within `sequencer.rs`
6. Write the logic.

There is default variables in the context: 
- `{{ i_lens[0] }}`: the length of the input 0. This also work for output: `{{ o_lens[0] }}` and other input `{{ i_lens[1] }}`
- `{{ i_shape[0] }}`: the array of dimensions of input 0. To get the first dimension of the array, just use: `{{ i_shape[0][0] }}` 
- `{{ i_chunks[0] }}`: the size of the chunks of each dimensions of input 0. By default, each variable is represented as a long array of values where to get to specific values you have to move by chunks. Those chunks are represented within this variable. To get the size of the chunks of the first dimensions use: `{{ i_chunks[0][0] }}`.
- `{{ op_type }}` the op type as some op_type like activation are using the same template.

7. Test it using the utils function and place it in the tests folder. The test can look as follows:
```Rust
#[test]
fn test_matmul_square_matrix() {
    // USER INPUT

    let n = 16;
    let mut input_data = HashMap::new();

    let data_a = ndarray::Array2::eye(n);
    let mut data_b = ndarray::Array2::<f32>::zeros((n, n));
    data_b[[0, 0]] = 0.2;
    data_b[[0, 1]] = 0.5;

    let sum = data_a.dot(&data_b);

    input_data.insert("A".to_string(), data_a.as_slice().unwrap());
    input_data.insert("B".to_string(), data_b.as_slice().unwrap());

    let n = n as i64;
    let model = model(graph(
        vec![tensor("A", &[n, n]), tensor("B", &[n, n])],
        vec![tensor("C", &[n, n])],
        vec![],
        vec![],
        vec![node(vec!["A", "B"], vec!["C"], "MatMul", "MatMul", vec![])],
    ));

    let session =
        pollster::block_on(wonnx::Session::from_model(model)).expect("Session did not create");

    let result = pollster::block_on(session.run(input_data)).unwrap();

    // Note: it is better to use a method that compares floats with a tolerance to account for differences
    // between implementations; see `wonnx/tests/common/mod.rs` for an example.
    assert_eq!((&result["C"]).try_into().unwrap(),sum.as_slice().unwrap());
}
```
> Check out tera documentation for other templating operation: https://tera.netlify.app/docs/

8. If at any point you want to do optimisation of several node you can do it within `sequencer.rs`.

## Supported Operators (ref [ONNX IR](https://github.com/onnx/onnx/blob/master/docs/Operators.md?plain=1)) 

|**Operator**|**Since version**|**Implemented**|**Shape inference supported**|
|-|-|-|-|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Abs">Abs</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Abs-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Abs-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Abs-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Acos">Acos</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Acos-7">7</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Acosh">Acosh</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Acosh-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Add">Add</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Add-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Add-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Add-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Add-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Add-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#And">And</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#And-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#And-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ArgMax">ArgMax</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMax-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMax-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMax-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMax-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ArgMin">ArgMin</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMin-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMin-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMin-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ArgMin-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Asin">Asin</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Asin-7">7</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Asinh">Asinh</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Asinh-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Atan">Atan</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Atan-7">7</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Atanh">Atanh</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Atanh-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#AveragePool">AveragePool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#AveragePool-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#AveragePool-10">10</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#AveragePool-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#AveragePool-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#BatchNormalization">BatchNormalization</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-15">15</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#BitShift">BitShift</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BitShift-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Cast">Cast</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cast-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cast-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cast-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cast-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Ceil">Ceil</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Ceil-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Ceil-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Ceil-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Clip">Clip</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Clip-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Clip-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Clip-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Clip-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Clip-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Compress">Compress</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Compress-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Compress-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Concat">Concat</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Concat-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Concat-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Concat-4">4</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Concat-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ConcatFromSequence">ConcatFromSequence</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ConcatFromSequence-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Constant">Constant</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Constant-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Constant-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Constant-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Constant-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Constant-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ConstantOfShape">ConstantOfShape</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ConstantOfShape-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Conv">Conv</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Conv-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Conv-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ConvInteger">ConvInteger</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ConvInteger-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ConvTranspose">ConvTranspose</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ConvTranspose-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ConvTranspose-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Cos">Cos</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cos-7">7</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Cosh">Cosh</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cosh-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#CumSum">CumSum</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#CumSum-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#CumSum-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#DepthToSpace">DepthToSpace</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#DepthToSpace-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#DepthToSpace-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#DepthToSpace-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#DequantizeLinear">DequantizeLinear</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#DequantizeLinear-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#DequantizeLinear-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Det">Det</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Det-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Div">Div</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Div-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Div-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Div-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Div-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Div-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Dropout">Dropout</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Dropout-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Dropout-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Dropout-10">10</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Dropout-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Dropout-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Dropout-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Einsum">Einsum</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Einsum-12">12</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Elu">Elu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Elu-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Elu-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Equal">Equal</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Equal-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Equal-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Equal-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Equal-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Erf">Erf</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Erf-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Erf-9">9</a>||✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Exp">Exp</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Exp-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Exp-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Exp-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Expand">Expand</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Expand-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Expand-8">8</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#EyeLike">EyeLike</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#EyeLike-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Flatten">Flatten</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Flatten-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Flatten-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Flatten-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Flatten-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Floor">Floor</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Floor-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Floor-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Floor-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GRU">GRU</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GRU-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GRU-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GRU-3">3</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GRU-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Gather">Gather</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gather-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gather-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gather-1">1</a>|✅ (axis=0)|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GatherElements">GatherElements</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GatherElements-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GatherElements-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GatherND">GatherND</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GatherND-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GatherND-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GatherND-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Gemm">Gemm</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-1">1</a>|✅*|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GlobalAveragePool">GlobalAveragePool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GlobalAveragePool-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GlobalLpPool">GlobalLpPool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GlobalLpPool-2">2</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GlobalLpPool-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GlobalMaxPool">GlobalMaxPool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GlobalMaxPool-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Greater">Greater</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Greater-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Greater-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Greater-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Greater-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GridSample">GridSample</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GridSample-16">16</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#HardSigmoid">HardSigmoid</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#HardSigmoid-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#HardSigmoid-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Hardmax">Hardmax</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Hardmax-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Hardmax-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Hardmax-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Identity">Identity</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Identity-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Identity-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Identity-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Identity-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#If">If</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#If-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#If-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#If-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#If-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#InstanceNormalization">InstanceNormalization</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#InstanceNormalization-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#InstanceNormalization-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#IsInf">IsInf</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#IsInf-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#IsNaN">IsNaN</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#IsNaN-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#IsNaN-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LRN">LRN</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LRN-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LRN-1">1</a>||
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LSTM">LSTM</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LSTM-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LSTM-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LSTM-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LeakyRelu">LeakyRelu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LeakyRelu-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LeakyRelu-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Less">Less</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Less-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Less-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Less-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Less-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Log">Log</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Log-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Log-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Log-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Loop">Loop</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Loop-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Loop-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Loop-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Loop-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LpNormalization">LpNormalization</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LpNormalization-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LpPool">LpPool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LpPool-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LpPool-2">2</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LpPool-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#MatMul">MatMul</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MatMul-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MatMul-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MatMul-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#MatMulInteger">MatMulInteger</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MatMulInteger-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Max">Max</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-8">8</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#MaxPool">MaxPool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxPool-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxPool-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxPool-10">10</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxPool-8">8</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxPool-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#MaxRoiPool">MaxRoiPool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxRoiPool-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#MaxUnpool">MaxUnpool</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxUnpool-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MaxUnpool-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Mean">Mean</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mean-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mean-8">8</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mean-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mean-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Min">Min</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Min-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Min-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Min-8">8</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Min-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Min-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Mod">Mod</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mod-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mod-10">10</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Mul">Mul</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mul-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mul-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mul-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mul-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mul-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Multinomial">Multinomial</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Multinomial-7">7</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Neg">Neg</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Neg-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Neg-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Neg-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#NonMaxSuppression">NonMaxSuppression</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#NonMaxSuppression-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#NonMaxSuppression-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#NonZero">NonZero</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#NonZero-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#NonZero-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Not">Not</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Not-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#OneHot">OneHot</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#OneHot-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#OneHot-9">9</a>|✅ (axis=-1)|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Optional">Optional</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Optional-15">15</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#OptionalGetElement">OptionalGetElement</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#OptionalGetElement-15">15</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#OptionalHasElement">OptionalHasElement</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#OptionalHasElement-15">15</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Or">Or</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Or-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Or-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#PRelu">PRelu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#PRelu-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#PRelu-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#PRelu-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#PRelu-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Pad">Pad</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pad-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pad-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pad-2">2</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pad-1">1</a>|✅ (mode=constant, pads>=0)|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Pow">Pow</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pow-15">15</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pow-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pow-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pow-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pow-1">1</a>|✅ (broadcast=0 and data type is f32)|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#QLinearConv">QLinearConv</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#QLinearConv-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#QLinearMatMul">QLinearMatMul</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#QLinearMatMul-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#QuantizeLinear">QuantizeLinear</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#QuantizeLinear-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#QuantizeLinear-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#RNN">RNN</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RNN-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RNN-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RNN-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#RandomNormal">RandomNormal</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RandomNormal-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#RandomNormalLike">RandomNormalLike</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RandomNormalLike-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#RandomUniform">RandomUniform</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RandomUniform-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#RandomUniformLike">RandomUniformLike</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RandomUniformLike-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Reciprocal">Reciprocal</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reciprocal-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reciprocal-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reciprocal-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceL1">ReduceL1</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceL1-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceL1-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceL1-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceL2">ReduceL2</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceL2-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceL2-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceL2-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceLogSum">ReduceLogSum</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceLogSum-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceLogSum-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceLogSum-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceLogSumExp">ReduceLogSumExp</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceLogSumExp-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceLogSumExp-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceLogSumExp-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceMax">ReduceMax</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMax-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMax-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMax-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMax-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceMean">ReduceMean</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMean-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMean-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMean-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceMin">ReduceMin</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMin-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMin-12">12</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMin-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceMin-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceProd">ReduceProd</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceProd-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceProd-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceProd-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceSum">ReduceSum</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceSum-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceSum-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceSum-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReduceSumSquare">ReduceSumSquare</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceSumSquare-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceSumSquare-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReduceSumSquare-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Relu">Relu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Relu-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Relu-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Relu-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Relu-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Reshape">Reshape</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reshape-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reshape-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reshape-5">5</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Reshape-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Resize">Resize</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Resize-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Resize-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Resize-10">10</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ReverseSequence">ReverseSequence</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ReverseSequence-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#RoiAlign">RoiAlign</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RoiAlign-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#RoiAlign-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Round">Round</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Round-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Scan">Scan</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Scan-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Scan-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Scan-8">8</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Scatter">Scatter</a> (deprecated)|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Scatter-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Scatter-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ScatterElements">ScatterElements</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ScatterElements-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ScatterElements-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ScatterElements-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ScatterND">ScatterND</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ScatterND-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ScatterND-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ScatterND-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Selu">Selu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Selu-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Selu-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SequenceAt">SequenceAt</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SequenceAt-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SequenceConstruct">SequenceConstruct</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SequenceConstruct-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SequenceEmpty">SequenceEmpty</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SequenceEmpty-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SequenceErase">SequenceErase</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SequenceErase-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SequenceInsert">SequenceInsert</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SequenceInsert-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SequenceLength">SequenceLength</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SequenceLength-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Shape">Shape</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Shape-15">15</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Shape-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Shape-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Shrink">Shrink</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Shrink-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sigmoid">Sigmoid</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sigmoid-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sigmoid-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sigmoid-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sign">Sign</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sign-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sign-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sin">Sin</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sin-7">7</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sinh">Sinh</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sinh-9">9</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Size">Size</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Size-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Size-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Slice">Slice</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Slice-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Slice-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Slice-10">10</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Slice-1">1</a>||✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Softplus">Softplus</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Softplus-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Softsign">Softsign</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Softsign-1">1</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SpaceToDepth">SpaceToDepth</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SpaceToDepth-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SpaceToDepth-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Split">Split</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Split-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Split-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Split-2">2</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Split-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SplitToSequence">SplitToSequence</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SplitToSequence-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sqrt">Sqrt</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sqrt-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sqrt-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sqrt-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Squeeze">Squeeze</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Squeeze-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Squeeze-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Squeeze-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#StringNormalizer">StringNormalizer</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#StringNormalizer-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sub">Sub</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sub-14">14</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sub-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sub-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sub-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sub-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Sum">Sum</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sum-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sum-8">8</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sum-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sum-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Tan">Tan</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tan-7">7</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Tanh">Tanh</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tanh-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tanh-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tanh-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#TfIdfVectorizer">TfIdfVectorizer</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#TfIdfVectorizer-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#ThresholdedRelu">ThresholdedRelu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ThresholdedRelu-10">10</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Tile">Tile</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tile-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tile-6">6</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tile-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#TopK">TopK</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#TopK-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#TopK-10">10</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#TopK-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Transpose">Transpose</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Transpose-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Transpose-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Trilu">Trilu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Trilu-14">14</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Unique">Unique</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Unique-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Unsqueeze">Unsqueeze</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Unsqueeze-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Unsqueeze-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Unsqueeze-1">1</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Upsample">Upsample</a> (deprecated)|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Upsample-10">10</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Upsample-9">9</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Upsample-7">7</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Where">Where</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Where-16">16</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Where-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Xor">Xor</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Xor-7">7</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Xor-1">1</a>|
|**Function**|**Since version**|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Bernoulli">Bernoulli</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Bernoulli-15">15</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#CastLike">CastLike</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#CastLike-15">15</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Celu">Celu</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Celu-12">12</a>|✅|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#DynamicQuantizeLinear">DynamicQuantizeLinear</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#DynamicQuantizeLinear-11">11</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#GreaterOrEqual">GreaterOrEqual</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GreaterOrEqual-12">12</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#HardSwish">HardSwish</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#HardSwish-14">14</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LessOrEqual">LessOrEqual</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LessOrEqual-12">12</a>|✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#LogSoftmax">LogSoftmax</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LogSoftmax-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LogSoftmax-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LogSoftmax-1">1</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#MeanVarianceNormalization">MeanVarianceNormalization</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MeanVarianceNormalization-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MeanVarianceNormalization-9">9</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#NegativeLogLikelihoodLoss">NegativeLogLikelihoodLoss</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#NegativeLogLikelihoodLoss-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#NegativeLogLikelihoodLoss-12">12</a>|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Range">Range</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Range-11">11</a>||✅|
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#Softmax">Softmax</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Softmax-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Softmax-11">11</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Softmax-1">1</a>|✅ |
|<a href="https://github.com/onnx/onnx/blob/main/docs/Operators.md#SoftmaxCrossEntropyLoss">SoftmaxCrossEntropyLoss</a>|<a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SoftmaxCrossEntropyLoss-13">13</a>, <a href="https://github.com/onnx/onnx/blob/main/docs/Changelog.md#SoftmaxCrossEntropyLoss-12">12</a>|

### Known limitations

* The `Clip`, `Resize`, `Reshape`, `Split`, `Pad` and `ReduceSum` ops accept (typically optional) secondary inputs to set various
  parameters (i.e. axis). These inputs are only supported if they are supplied as initializer tensors (i.e. do not depend 
  on inputs and are not outputs of other ops), because wonnx pre-compiles all operations to shaders in advance (and must know
  these parameters up front).

* Internally 64-bit integers are not supported (the reason is they are not supported in the current version of WGSL); 
  inputs and initializers with 64-bit scalars are converted to 32-bit values (possibly overflowing).

* For `MatMul` and `Gemm`, the matrix dimensions must be divisible by 2, or the output matrix must be of size (1, N). Matrix 
  multiplication only supports floats, not integers (this is a WebGPU/WGSL limitation).

### Shape inference

WONNX needs to know the shape of input and output tensors for each operation in order to generate shader code for executing
it. ONNX models however do not always contain this information for intermediate values. Shape inference is the process of
deducing the shape of intermediate values from the shape of inputs and outputs and the characteristics of each operation.

WONNX supports a limited form of shape inference (the process of determining what the shapes are of the various nodes in 
a model's graph). Shape inference is available programmatically as well as through the CLI. Before shape inference can be
performed, all dynamic dimension parameters need to be replaced with static values. Shape inference only infers output shapes
from input shapes for specific supported ops (see the table above). Inference cannot succeed if the shape for any input of
a node is not known. Nodes that already have fully defined shapes for their outputs are left unchanged (and the outputs are
used for shape inference on nodes that use these outputs as inputs).

To perform shape inference using the CLI, run a command similar to this (here `batch_size` and `sequence_length` are dynamic
dimension parameters; the `-i` flag enables shape inference):

````bash
nnx prepare model.onnx model-prepared.onnx --set batch_size=1 --set sequence_length=255 -i
````

To perform shape inference programmatically, use `apply_dynamic_dimensions` and `infer_shapes` from the 
`wonnx_preprocessing::shape_inference` module.

### Constant folding

Some models contain subgraphs whose output can be determined statically, as they do not depend on the specific inputs provided
during inference. WONNX can replace such constant intermediate values with static values ('constant folding'). This is
supported in the following cases:

* Output of nodes of the `Constant` op type (these are replaced with initializers)
* Output of nodes of the `Shape` op type where the shape of the input is known (up front or during inference)
* Output of nodes of which all inputs are constant (possibly after folding), *and* for which the operator is supported by WONNX.

Constant folding is performed as part of shape inference, unless disabled (from the CLI pass `--no-fold-constants` to disable). This
is done in order to support models that dynamically calculate shapes using operators such as `Shape`/`Squeeze`/`Unsqueeze` depending
on dynamically set dimension parameters (e.g. batch size).

## License

Licensed under either of
 * Apache License, Version 2.0 ([LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0)
 * MIT license ([LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT)
at your option.

Except for the following files:

* `data/models`: 
  * `mobilenetv2-7.onnx`: [source](https://github.com/onnx/models/blob/main/vision/classification/mobilenet/model/mobilenetv2-7.onnx), Apache-2.0 license only.
  * `squeezenet-labels.txt`: [source](https://github.com/onnx/models/blob/main/vision/classification/synset.txt), Apache-2.0 license only.

* `data/images`:
  * `pelican.jpeg`: [source](https://en.wikipedia.org/wiki/Pelican#/media/File:Pelikan_Walvis_Bay.jpg), (C) Rui Ornelas, [CC-BY 2.0](https://creativecommons.org/licenses/by/2.0/).
  * `bald_eagle.jpeg`: [source](https://en.wikipedia.org/wiki/Bald_eagle#/media/File:Bald-Eagle-9114-cropped.jpg), (C) David R. Tribble, [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)


### Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted
for inclusion in the work by you shall be dual licensed as above, without any
additional terms or conditions.