tucan9389

/* 
 * original code of mlir.js in netron repo 
 * https://github.com/lutzroeder/netron/blob/main/source/mlir.js
*/

if (process.argv.length >= 3) {
    const fs = require('fs');

    function runParser(textContent) {

tucan9389

import argparse
import os
import subprocess
from tqdm import tqdm
import random

def get_mlir_file_paths(directory):
    mlir_file_paths = []
    for root, dirs, files in os.walk(directory):
        for file in files:

tucan9389

{
  "functions": [
    {
      "name": "@test_onnx_conv_simple_pattern",
      "inputs": [
        "%arg0",
        "%arg1"
      ],
      "inputTypes": [
        "tensor<5x3x32x32xf32>",

tucan9389

module {
func.func @test_onnx_conv_simple_pattern(%arg0: tensor<5x3x32x32xf32>, %arg1: tensor<?x3x2x2xf32>) -> tensor<5x?x31x31xf32> {
%0 = "onnx.NoValue"() {value} : () -> none
%1 = "onnx.Conv"(%arg0, %arg1, %0) {auto_pad = "NOTSET", kernel_shape = [2, 2], pads = [0, 0, 0, 0]} : (tensor<5x3x32x32xf32>, tensor<?x3x2x2xf32>, none) -> tensor<5x?x31x31xf32>
return %1 : tensor<5x?x31x31xf32>
}
}

tucan9389

{
  "functions": [
    {
      "name": "@__inference_predict_3320",
      "inputs": [
        "%arg0",
        "%arg1",
        "%arg2",
        "%arg3",
        "%arg4",

tucan9389

{
  "functions": [
    {
      "name": "@main",
      "inputs": [
        "%image",
        "%weights",
        "%bias"
      ],
      "inputTypes": [

tucan9389

module {
func.func @test_onnx_conv_simple_pattern(%arg0: tensor<5x3x32x32xf32>, %arg1: tensor<?x3x2x2xf32>) -> tensor<5x?x31x31xf32> {
%0 = "onnx.NoValue"() {value} : () -> none
%1 = "onnx.Conv"(%arg0, %arg1, %0) {auto_pad = "NOTSET", kernel_shape = [2, 2], pads = [0, 0, 0, 0]} : (tensor<5x3x32x32xf32>, tensor<?x3x2x2xf32>, none) -> tensor<5x?x31x31xf32>
return %1 : tensor<5x?x31x31xf32>
}
}

tucan9389

module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 440 : i32}, tf_saved_model.semantics} {
  "tf_saved_model.global_tensor"() {is_mutable, sym_name = "__sm_node4__optimizer.iter", tf_saved_model.exported_names = [], type = tensor<i64>, value = dense<0> : tensor<i64>} : () -> ()
  "tf_saved_model.global_tensor"() {sym_name = "__sm_node6__optimizer.learning_rate", tf_saved_model.exported_names = [], type = tensor<f32>, value = dense<0.00999999977> : tensor<f32>} : () -> ()
  "tf_saved_model.global_tensor"() {is_mutable, sym_name = "__sm_node17__model.conv1.kernel", tf_saved_model.exported_names = [], type = tensor<5x5x1x32xf32>, value = dense<""> : tensor<5x5x1x32xf32>} : () -> ()
  "tf_saved_model.global_tensor"() {is_mutable, sym_name = "__sm_node26__model.conv2.kernel", tf_saved_model.exported_names = [], type = tensor<5x5x32x32xf32>, value = dense<""> : tensor<5x5x32x32xf32>} : () -> ()
  "tf_saved_model.global_tensor"() {is_mutable, sym_name = "__sm_node39__model.de

tucan9389

{
  "functions": [
    {
      "name": "@test_onnx_conv_simple_pattern",
      "inputs": [
        "%arg0",
        "%arg1"
      ],
      "inputTypes": [
        "tensor<5x3x32x32xf32>",

tucan9389

module {
    func.func @test_onnx_conv_simple_pattern(%arg0: tensor<5x3x32x32xf32>, %arg1: tensor<?x3x2x2xf32>) -> tensor<5x?x31x31xf32> {
        %0 = "onnx.NoValue"() {value} : () -> none
        %1 = "onnx.Conv"(%arg0, %arg1, %0) {auto_pad = "NOTSET", kernel_shape = [2, 2], pads = [0, 0, 0, 0]} : (tensor<5x3x32x32xf32>, tensor<?x3x2x2xf32>, none) -> tensor<5x?x31x31xf32>
        return %1 : tensor<5x?x31x31xf32>
    }
}