Command Line: `python -m yobx print`#

This example builds a small ONNX model, saves it to a temporary file, and then demonstrates the different output formats produced by the python -m yobx print command.

The same result can be achieved from the terminal with:

python -m yobx print pretty  model.onnx
python -m yobx print printer model.onnx
python -m yobx print dot     model.onnx

import os
import tempfile
import onnx
import onnx.helper as oh
from yobx._command_lines_parser import _cmd_print

TFLOAT = onnx.TensorProto.FLOAT

Build a small ONNX model#

The graph computes Z = Relu(X + Y) with static shapes (2, 3).

model = oh.make_model(
    oh.make_graph(
        [oh.make_node("Add", ["X", "Y"], ["added"]), oh.make_node("Relu", ["added"], ["Z"])],
        "add_relu",
        [
            oh.make_tensor_value_info("X", TFLOAT, [2, 3]),
            oh.make_tensor_value_info("Y", TFLOAT, [2, 3]),
        ],
        [oh.make_tensor_value_info("Z", TFLOAT, [2, 3])],
    ),
    opset_imports=[oh.make_opsetid("", 18)],
    ir_version=10,
)

# Save to a temporary file so the CLI helpers can load it.
fd, tmp = tempfile.mkstemp(suffix=".onnx")
os.close(fd)
onnx.save(model, tmp)

`pretty` format#

The default format produced by yobx.helpers.onnx_helper.pretty_onnx(). It shows opset, inputs/outputs, and every node in a compact, human-readable layout.

print("python -m yobx print pretty model.onnx")
print("-" * 40)
_cmd_print(["print", "pretty", tmp])

python -m yobx print pretty model.onnx
----------------------------------------
opset: domain='' version=18
input: name='X' type=dtype('float32') shape=[2, 3]
input: name='Y' type=dtype('float32') shape=[2, 3]
Add(X, Y) -> added
  Relu(added) -> Z
output: name='Z' type=dtype('float32') shape=[2, 3]

`printer` format#

Uses the built-in onnx.printer.to_text renderer which produces the official ONNX text representation.

print("python -m yobx print printer model.onnx")
print("-" * 40)
_cmd_print(["print", "printer", tmp])

python -m yobx print printer model.onnx
----------------------------------------
<
   ir_version: 10,
   opset_import: ["" : 18]
>
add_relu (float[2,3] X, float[2,3] Y) => (float[2,3] Z) {
   added = Add (X, Y)
   Z = Relu (added)
}

`dot` format#

Dumps the DOT graph source. Pipe the output to dot -Tsvg to get a visual representation of the graph (see also yobx.helpers.dot_helper.to_dot()).

print("python -m yobx print dot model.onnx")
print("-" * 40)
_cmd_print(["print", "dot", tmp])

python -m yobx print dot model.onnx
----------------------------------------
digraph {
  graph [rankdir=TB, splines=true, overlap=false, nodesep=0.2, ranksep=0.2, fontsize=8];
  node [style="rounded,filled", color="#888888", fontcolor="#222222", shape=box];
  edge [arrowhead=vee, fontsize=7, labeldistance=-5, labelangle=0];
  I_0 [label="X\nFLOAT(2,3)", fillcolor="#aaeeaa"];
  I_1 [label="Y\nFLOAT(2,3)", fillcolor="#aaeeaa"];
  Add_2 [label="Add(., .)", fillcolor="#cccccc"];
  Relu_3 [label="Relu(.)", fillcolor="#cccccc"];
  I_0 -> Add_2 [label="FLOAT(2,3)"];
  I_1 -> Add_2 [label="FLOAT(2,3)"];
  Add_2 -> Relu_3 [label="FLOAT(2,3)"];
  O_4 [label="Z\nFLOAT(2,3)", fillcolor="#aaaaee"];
  Relu_3 -> O_4;
}

Total running time of the script: (0 minutes 0.170 seconds)