onnx.cpp 26.5 KB
Newer Older
Paul's avatar
Paul committed
1
2
3
4
5
6
7
8
#include <google/protobuf/text_format.h>
#include <google/protobuf/io/zero_copy_stream_impl.h>
#include <onnx.pb.h>
#include <iostream>
#include <fstream>
#include <unordered_map>
#include <functional>
#include <array>
Paul's avatar
Paul committed
9
#include <utility>
10
#include <vector>
Paul's avatar
Paul committed
11

Paul's avatar
Paul committed
12
13
14
15
#include <migraph/fallthrough.hpp>
#include <migraph/program.hpp>
#include <migraph/operators.hpp>
#include <migraph/ranges.hpp>
16
#include <migraph/instruction.hpp>
Paul's avatar
Paul committed
17

Paul's avatar
Paul committed
18
namespace migraph {
Paul's avatar
Paul committed
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41

struct unknown
{
    std::string op;
    std::string name() const { return "unknown:" + op; }
    shape compute_shape(std::vector<shape> input) const
    {
        if(input.empty())
            return {};
        else
            return input.front();
    }
    friend std::ostream& operator<<(std::ostream& os, const unknown& x)
    {
        os << x.name();
        return os;
    }
};

struct onnx_parser
{
    using attribute_map = std::unordered_map<std::string, onnx::AttributeProto>;
    using node_map      = std::unordered_map<std::string, onnx::NodeProto>;
Paul's avatar
Paul committed
42
    using op_func = std::function<instruction_ref(attribute_map, std::vector<instruction_ref>)>;
Paul's avatar
Paul committed
43
44
45
46
47
48
49
50
    node_map nodes;
    std::unordered_map<std::string, instruction_ref> instructions;
    program prog = program();

    std::unordered_map<std::string, op_func> ops;

    onnx_parser()
    {
Shucai Xiao's avatar
Shucai Xiao committed
51
        add_generic_op("MatMul", op::dot{});
Khalique's avatar
Khalique committed
52
        add_generic_op("Relu", op::relu{});
53
54
55
56
57
58

        add_broadcastable_binary_op("Add", op::add{});
        add_broadcastable_binary_op("Div", op::div{});
        add_broadcastable_binary_op("Mul", op::mul{});
        add_broadcastable_binary_op("Sub", op::sub{});
        add_broadcastable_binary_op("Sum", op::add{});
Paul's avatar
Paul committed
59

Khalique's avatar
Khalique committed
60
        add_mem_op("ImageScaler", &onnx_parser::parse_imagescaler);
61
        add_mem_op("LeakyRelu", &onnx_parser::parse_leaky_relu);
Paul's avatar
Paul committed
62
63
        add_mem_op("Constant", &onnx_parser::parse_constant);
        add_mem_op("Conv", &onnx_parser::parse_conv);
Paul's avatar
Paul committed
64
65
        add_mem_op("MaxPool", &onnx_parser::parse_pooling);
        add_mem_op("AveragePool", &onnx_parser::parse_pooling);
66
67
        add_mem_op("GlobalMaxPool", &onnx_parser::parse_pooling);
        add_mem_op("GlobalAveragePool", &onnx_parser::parse_pooling);
Paul's avatar
Paul committed
68
        add_mem_op("Reshape", &onnx_parser::parse_reshape);
Paul's avatar
Paul committed
69
70
        add_mem_op("Flatten", &onnx_parser::parse_flatten);
        add_mem_op("Gemm", &onnx_parser::parse_gemm);
71
        add_mem_op("BatchNormalization", &onnx_parser::parse_batchnorm);
Paul's avatar
Paul committed
72
        add_mem_op("Softmax", &onnx_parser::parse_softmax);
73
74
75
        add_mem_op("Squeeze", &onnx_parser::parse_squeeze);
        add_mem_op("Unsqueeze", &onnx_parser::parse_unsqueeze);
        add_mem_op("Slice", &onnx_parser::parse_slice);
Scott Thornton's avatar
Scott Thornton committed
76
        add_mem_op("Concat", &onnx_parser::parse_concat);
Paul's avatar
Paul committed
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
    }

    template <class F>
    void add_op(std::string name, F f)
    {
        ops.emplace(name, f);
    }

    template <class F>
    void add_mem_op(std::string name, F f)
    {
        ops.emplace(name, [=](auto&&... xs) {
            return std::mem_fn(f)(*this, name, std::forward<decltype(xs)>(xs)...);
        });
    }
92
93
94
95
    template <class T>
    void add_broadcastable_binary_op(std::string name, T x)
    {
        ops.emplace(name, [this, x](attribute_map attributes, std::vector<instruction_ref> args) {
Scott Thornton's avatar
Scott Thornton committed
96
97
            if(args.size() != 2)
                MIGRAPH_THROW("binaGry operators should have 2 operands");
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
            if(contains(attributes, "broadcast"))
            {
                uint64_t broadcasted = parse_value(attributes.at("broadcast")).at<uint64_t>();
                if(broadcasted != 0)
                {
                    uint64_t axis = (contains(attributes, "axis"))
                                        ? parse_value(attributes.at("axis")).at<uint64_t>()
                                        : 0;
                    auto l =
                        prog.add_instruction(op::broadcast{axis, args[0]->get_shape()}, args[1]);
                    return prog.add_instruction(x, args[0], l);
                }
            }
            else
            {
                // Example:
                // s0 = (3,2,4,5) and s1 = (2,1,1)
                //
Scott Thornton's avatar
Scott Thornton committed
116
117
                // In this case we need to broadcast (:,1,1) portion of
                // s1 plus broadcast the 1st dimension of s1
118
119
120
121
122
123
124
125
126
127
128
129
130
                // giving output_lens = (3,2,4,5)
                //
                // Another example:
                // s0 = (3,2,1,5) and s1 = (2,7,5)
                // In this case we need to broadcast the (:,:,1:,:) axis
                // of s0 plus the 1st dimension of s1 giving
                // output_lens = (3,2,7,5)
                //
                // Get lengths for both arguments
                const std::vector<std::size_t>& s0 = args[0]->get_shape().lens();
                const std::vector<std::size_t>& s1 = args[1]->get_shape().lens();

                // Copy the larger vector to output_lens
Scott Thornton's avatar
Scott Thornton committed
131
132
                std::vector<std::size_t> output_lens = (s0.size() >= s1.size()) ? s0 : s1;
                if(s0.size() >= s1.size())
133
134
135
                {
                    // s0 is bigger, so iterate over the range of s1
                    auto offset = s0.size() - s1.size();
Scott Thornton's avatar
Scott Thornton committed
136
                    for(std::size_t i = 0; i < s1.size(); i++)
137
                    {
Scott Thornton's avatar
Scott Thornton committed
138
                        output_lens[i + offset] = std::max(s0[i + offset], s1[i]);
139
140
141
142
143
144
                    }
                }
                else
                {
                    // s1 is bigger, so iterate over the range of s0
                    auto offset = s1.size() - s0.size();
Scott Thornton's avatar
Scott Thornton committed
145
                    for(std::size_t i = 0; i < s0.size(); i++)
146
                    {
Scott Thornton's avatar
Scott Thornton committed
147
                        output_lens[i + offset] = std::max(s0[i], s1[i + offset]);
148
149
150
151
152
153
154
                    }
                }
            }
            return prog.add_instruction(x, args);
        });
    }

Paul's avatar
Paul committed
155
    template <class T>
Paul's avatar
Paul committed
156
157
158
159
160
161
162
163
164
165
166
    void add_generic_op(std::string name, T x)
    {
        ops.emplace(name, [this, x](attribute_map attributes, std::vector<instruction_ref> args) {
            if(args.size() == 2 and contains(attributes, "broadcast"))
            {
                uint64_t broadcasted = parse_value(attributes.at("broadcast")).at<uint64_t>();
                if(broadcasted != 0)
                {
                    uint64_t axis = (contains(attributes, "axis"))
                                        ? parse_value(attributes.at("axis")).at<uint64_t>()
                                        : 0;
Scott Thornton's avatar
Scott Thornton committed
167
168
                    auto l =
                        prog.add_instruction(op::broadcast{axis, args[0]->get_shape()}, args[1]);
Paul's avatar
Paul committed
169
170
171
172
173
174
175
                    return prog.add_instruction(x, args[0], l);
                }
            }
            return prog.add_instruction(x, args);
        });
    }

Paul's avatar
Paul committed
176
    instruction_ref
Paul's avatar
Paul committed
177
    parse_softmax(const std::string&, const attribute_map&, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
178
179
    {
        auto dims = args.front()->get_shape().lens();
Scott Thornton's avatar
Scott Thornton committed
180
181
        auto r =
            prog.add_instruction(op::reshape{{long(dims[0]), long(dims[1]), 1, 1}}, args.front());
182
183
        auto s = prog.add_instruction(op::softmax{}, r);
        return prog.add_instruction(op::reshape{{long(dims[0]), long(dims[1])}}, s);
Paul's avatar
Paul committed
184
185
    }

Paul's avatar
Paul committed
186
    instruction_ref
Paul's avatar
Paul committed
187
    parse_conv(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
188
    {
189
        op::convolution op;
Paul's avatar
Paul committed
190
191
192
        if(contains(attributes, "pads"))
        {
            copy(attributes["pads"].ints(), op.padding.begin());
Paul's avatar
Paul committed
193
        }
Paul's avatar
Paul committed
194
195
196
197
198
199
200
201
202
203
204
205
        if(contains(attributes, "strides"))
        {
            copy(attributes["strides"].ints(), op.stride.begin());
        }
        if(contains(attributes, "dilations"))
        {
            copy(attributes["dilations"].ints(), op.dilation.begin());
        }
        if(args.size() == 3)
        {
            uint64_t axis = 1;
            auto l1       = prog.add_instruction(op, args[0], args[1]);
Scott Thornton's avatar
Scott Thornton committed
206
            auto l2       = prog.add_instruction(op::broadcast{axis, l1->get_shape()}, args[2]);
207
            return prog.add_instruction(op::add{}, l1, l2);
Paul's avatar
Paul committed
208
        }
Paul's avatar
Paul committed
209
210
        return prog.add_instruction(op, args);
    }
Paul's avatar
Paul committed
211

Paul's avatar
Paul committed
212
213
214
    instruction_ref parse_pooling(const std::string& name,
                                  attribute_map attributes,
                                  std::vector<instruction_ref> args)
Paul's avatar
Paul committed
215
    {
Khalique's avatar
Khalique committed
216
217
        op::pooling op{ends_with(name, "MaxPool") ? "max" : "average"};
        if(starts_with(name, "Global"))
218
        {
Khalique's avatar
Khalique committed
219
220
            auto lens  = args.front()->get_shape().lens();
            op.lengths = {lens[2], lens[3]};
221
        }
Paul's avatar
Paul committed
222
223
224
225
226
227
228
229
230
231
232
233
        if(contains(attributes, "pads"))
        {
            copy(attributes["pads"].ints(), op.padding.begin());
        }
        if(contains(attributes, "strides"))
        {
            copy(attributes["strides"].ints(), op.stride.begin());
        }
        if(contains(attributes, "kernel_shape"))
        {
            copy(attributes["kernel_shape"].ints(), op.lengths.begin());
        }
Paul's avatar
Paul committed
234
        return prog.add_instruction(op, std::move(args));
Paul's avatar
Paul committed
235
236
    }

Paul's avatar
Paul committed
237
    instruction_ref
Paul's avatar
Paul committed
238
    parse_reshape(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
239
    {
240
        op::reshape op;
Paul's avatar
Paul committed
241
242
243
244
245
246
247
        if(args.size() == 1)
        {
            literal s = parse_value(attributes.at("shape"));
            s.visit([&](auto v) { copy(v, std::back_inserter(op.dims)); });
        }
        if(args.size() == 2)
        {
Paul's avatar
Paul committed
248
            literal s = args[1]->get_literal();
Paul's avatar
Paul committed
249
            s.visit([&](auto v) { copy(v, std::back_inserter(op.dims)); });
Paul's avatar
Paul committed
250
        }
Paul's avatar
Paul committed
251
252
253
        return prog.add_instruction(op, args[0]);
    }

Paul's avatar
Paul committed
254
    instruction_ref
Paul's avatar
Paul committed
255
    parse_flatten(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
256
257
    {
        uint64_t axis = 0;
Paul's avatar
Paul committed
258
259
260
261
        if(contains(attributes, "axis"))
        {
            axis = parse_value(attributes.at("axis")).at<int>();
        }
262
        return prog.add_instruction(op::flatten{axis}, args[0]);
Paul's avatar
Paul committed
263
264
    }

265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
    instruction_ref
    parse_squeeze(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        op::squeeze op;
        literal s = parse_value(attributes.at("axes"));
        s.visit([&](auto v) { copy(v, std::back_inserter(op.axes)); });
        return prog.add_instruction(op, args[0]);
    }

    instruction_ref
    parse_unsqueeze(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        op::unsqueeze op;
        literal s = parse_value(attributes.at("axes"));
        s.visit([&](auto v) { copy(v, std::back_inserter(op.axes)); });
        return prog.add_instruction(op, args[0]);
    }

Scott Thornton's avatar
Scott Thornton committed
283
284
285
286
287
288
289
    instruction_ref
    parse_concat(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        std::size_t axis = parse_value(attributes.at("axis")).at<int>();
        op::concat op{axis};
        return prog.add_instruction(op, std::move(args));
    }
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310

    instruction_ref
    parse_slice(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        op::slice op;
        if(contains(attributes, "axes"))
        {
            literal s = parse_value(attributes.at("axes"));
            s.visit([&](auto v) { copy(v, std::back_inserter(op.axes)); });
        }
        {
            literal s = parse_value(attributes.at("ends"));
            s.visit([&](auto v) { copy(v, std::back_inserter(op.ends)); });
        }
        {
            literal s = parse_value(attributes.at("starts"));
            s.visit([&](auto v) { copy(v, std::back_inserter(op.starts)); });
        }
        return prog.add_instruction(op, args[0]);
    }

Paul's avatar
Paul committed
311
312
313
    instruction_ref parse_constant(const std::string&,
                                   attribute_map attributes,
                                   const std::vector<instruction_ref>&)
Paul's avatar
Paul committed
314
315
316
317
    {
        literal v = parse_value(attributes.at("value"));
        return prog.add_literal(v);
    }
Paul's avatar
Paul committed
318

Paul's avatar
Paul committed
319
    instruction_ref
Paul's avatar
Paul committed
320
    parse_gemm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
    {
        float alpha = 1.0f;
        float beta  = 0.0f;
        bool transa = false;
        bool transb = false;
        if(contains(attributes, "alpha"))
        {
            alpha = parse_value(attributes.at("alpha")).at<float>();
        }
        if(contains(attributes, "beta"))
        {
            alpha = parse_value(attributes.at("beta")).at<float>();
        }
        if(contains(attributes, "transA"))
        {
            transa = parse_value(attributes.at("transA")).at<bool>();
        }
        if(contains(attributes, "transB"))
        {
            transb = parse_value(attributes.at("transB")).at<bool>();
        }
        std::vector<int64_t> perm = {1, 0};
343
344
        auto l1 = (transa) ? prog.add_instruction(op::transpose{perm}, args[0]) : args[0];
        auto l2 = (transb) ? prog.add_instruction(op::transpose{perm}, args[1]) : args[1];
Paul's avatar
Paul committed
345
346
347
        if(args.size() == 3)
        {
            uint64_t axis = 1;
Shucai Xiao's avatar
Shucai Xiao committed
348
            auto l3       = prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Scott Thornton's avatar
Scott Thornton committed
349
            auto l4       = prog.add_instruction(op::broadcast{axis, l3->get_shape()}, args[2]);
350
            return prog.add_instruction(op::add{}, l3, l4);
Paul's avatar
Paul committed
351
        }
Shucai Xiao's avatar
Shucai Xiao committed
352
        return prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Paul's avatar
Paul committed
353
354
    }

355
    instruction_ref
Paul's avatar
Paul committed
356
    parse_batchnorm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
357
    {
Scott Thornton's avatar
Scott Thornton committed
358
359
        float epsilon                                     = 1e-5f;
        float momentum                                    = 0.9f;
360
        op::batch_norm_inference::bn_infer_mode_t bn_mode = op::batch_norm_inference::spatial;
Scott Thornton's avatar
Scott Thornton committed
361
        bool is_test                                      = false;
362
363
364
365
366
367
        if(contains(attributes, "epsilon"))
        {
            epsilon = parse_value(attributes.at("epsilon")).at<float>();
        }
        if(contains(attributes, "momentum"))
        {
368
            momentum = parse_value(attributes.at("momentum")).at<float>();
369
370
371
        }
        if(contains(attributes, "is_test"))
        {
wsttiger's avatar
wsttiger committed
372
            is_test = parse_value(attributes.at("is_test")).at<uint64_t>() > 0;
373
374
375
        }
        if(contains(attributes, "spatial"))
        {
376
            bn_mode = (parse_value(attributes.at("spatial")).at<uint64_t>() > 0)
377
378
                          ? op::batch_norm_inference::spatial
                          : op::batch_norm_inference::per_activation;
379
        }
Paul's avatar
Paul committed
380
        (void)is_test;
Paul's avatar
Paul committed
381
        op::batch_norm_inference op{epsilon, momentum, bn_mode};
Paul's avatar
Paul committed
382
        return prog.add_instruction(op, std::move(args));
383
384
    }

385
386
387
388
    instruction_ref parse_leaky_relu(const std::string&,
                                     attribute_map attributes,
                                     std::vector<instruction_ref> args)
    {
Khalique's avatar
Khalique committed
389
        float alpha = 0.01; // default alpha val for leaky relu
390
391
392
393
394
395
396
397
        if(contains(attributes, "alpha"))
        {
            alpha = parse_value(attributes.at("alpha")).at<float>();
        }
        op::leaky_relu op{alpha};
        return prog.add_instruction(op, args.front());
    }

Khalique's avatar
Khalique committed
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
    instruction_ref parse_imagescaler(const std::string&,
                                      attribute_map attributes,
                                      std::vector<instruction_ref> args)
    {
        float scale = 1.0;
        std::vector<float> bias{};
        if(contains(attributes, "scale"))
        {
            scale = parse_value(attributes.at("scale")).at<float>();
        }

        if(contains(attributes, "bias"))
        {
            auto&& bias_floats = attributes["bias"].floats();
            bias               = std::vector<float>(bias_floats.begin(), bias_floats.end());
        }
        auto input_shape = args.front()->get_shape();
Khalique's avatar
Khalique committed
415

Khalique's avatar
Khalique committed
416
417
418
419
420
421
422
423
424
        auto scale_val = prog.add_literal(scale);
        auto bias_vals = prog.add_literal(
            migraph::literal{migraph::shape{migraph::shape::float_type, {bias.size()}}, bias});

        auto scale_tensor = prog.add_instruction(migraph::op::scalar{input_shape}, scale_val);
        auto img_scaled   = prog.add_instruction(migraph::op::mul{}, args.front(), scale_tensor);
        auto bias_bcast   = prog.add_instruction(migraph::op::broadcast{1, input_shape}, bias_vals);
        return prog.add_instruction(migraph::op::add{}, img_scaled, bias_bcast);
    }
Khalique's avatar
Khalique committed
425

Paul's avatar
Paul committed
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
    void parse_from(std::istream& is)
    {
        onnx::ModelProto model;
        if(model.ParseFromIstream(&is))
        {
            if(model.has_graph())
            {
                this->parse_graph(model.graph());
            }
        }
        else
        {
            throw std::runtime_error("Failed reading");
        }
    }

    void parse_graph(const onnx::GraphProto& graph)
    {
        nodes = get_nodes(graph);
445
446
447
448
449
        std::unordered_map<std::string, onnx::TensorProto> initializer_data;
        for(auto&& f : graph.initializer())
        {
            initializer_data[f.name()] = f;
        }
Paul's avatar
Paul committed
450
451
452
        for(auto&& input : graph.input())
        {
            const std::string& name = input.name();
453
454
455
456
457
458
459
460
461
462
463
464
            // Does the input have an initializer?
            if(contains(initializer_data, name))
            {
                auto t             = initializer_data[name];
                instructions[name] = prog.add_literal(parse_tensor(t));
            }
            else
            {
                // TODO: Get shape of input parameter
                shape s            = parse_type(input.type());
                instructions[name] = prog.add_parameter(name, s);
            }
Paul's avatar
Paul committed
465
466
467
        }
        for(auto&& p : nodes)
        {
468
            this->parse_node(get_name(p.second));
Paul's avatar
Paul committed
469
470
471
        }
    }

Paul's avatar
Paul committed
472
    void parse_node(const std::string& name)
Paul's avatar
Paul committed
473
    {
Paul's avatar
Paul committed
474
        if(name.empty())
Paul's avatar
Paul committed
475
            MIGRAPH_THROW("Onnx node must have a name");
Paul's avatar
Paul committed
476
477
478
479
480
481
482
483
        if(instructions.count(name) == 0)
        {
            auto&& node = nodes.at(name);
            std::vector<instruction_ref> args;
            for(auto&& input : node.input())
            {
                if(nodes.count(input) > 0)
                {
484
                    auto&& iname = get_name(nodes.at(input));
Paul's avatar
Paul committed
485
                    assert(name != iname);
Paul's avatar
Paul committed
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
                    this->parse_node(iname);
                    args.push_back(instructions.at(iname));
                }
                else
                {
                    args.push_back(instructions.at(input));
                }
            }
            if(ops.count(node.op_type()) == 0)
            {
                instructions[name] = prog.add_instruction(unknown{node.op_type()}, args);
            }
            else
            {
                instructions[name] = ops[node.op_type()](get_attributes(node), args);
            }
        }
    }

    static attribute_map get_attributes(const onnx::NodeProto& node)
    {
        std::unordered_map<std::string, onnx::AttributeProto> result;
        for(auto&& attr : node.attribute())
        {
            result[attr.name()] = attr;
        }
        return result;
    }

515
516
517
518
519
    static std::string get_name(const onnx::NodeProto& node)
    {
        if(node.name().empty())
        {
            std::string generated = "migraph_unnamed_node";
Paul's avatar
Paul committed
520
521
522
523
            return std::accumulate(node.output().begin(),
                                   node.output().end(),
                                   generated,
                                   [](auto x, auto y) { return x + "_" + y; });
524
525
526
527
        }
        return node.name();
    }

Paul's avatar
Paul committed
528
529
530
531
532
    static node_map get_nodes(const onnx::GraphProto& graph)
    {
        std::unordered_map<std::string, onnx::NodeProto> result;
        for(auto&& node : graph.node())
        {
533
            result[get_name(node)] = node;
Paul's avatar
Paul committed
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
            for(auto&& output : node.output())
            {
                result[output] = node;
            }
        }
        return result;
    }

    template <class T>
    static literal from_repeated(shape::type_t t, const T& r)
    {
        std::size_t size = r.size();
        return literal{{t, {size}}, r.begin(), r.end()};
    }

    static literal parse_value(const onnx::AttributeProto& attr)
    {
        switch(attr.type())
        {
        case onnx::AttributeProto::UNDEFINED: return {};
        case onnx::AttributeProto::FLOAT: return literal{attr.f()};
        case onnx::AttributeProto::INT: return literal{attr.i()};
        case onnx::AttributeProto::STRING: return {};
        case onnx::AttributeProto::TENSOR: return parse_tensor(attr.t());
        case onnx::AttributeProto::GRAPH: return {};
Paul's avatar
Paul committed
559
        case onnx::AttributeProto::FLOATS: return from_repeated(shape::float_type, attr.floats());
Paul's avatar
Paul committed
560
561
562
563
564
        case onnx::AttributeProto::INTS: return from_repeated(shape::int64_type, attr.ints());
        case onnx::AttributeProto::STRINGS: return {};
        case onnx::AttributeProto::TENSORS: return {};
        case onnx::AttributeProto::GRAPHS: return {};
        }
Paul's avatar
Paul committed
565
        MIGRAPH_THROW("Invalid attribute type");
Paul's avatar
Paul committed
566
567
568
569
570
    }

    static literal parse_tensor(const onnx::TensorProto& t)
    {
        std::vector<std::size_t> dims(t.dims().begin(), t.dims().end());
571
572
        if(t.has_raw_data())
        {
wsttiger's avatar
wsttiger committed
573
            const std::string& s = t.raw_data();
Scott Thornton's avatar
Scott Thornton committed
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
            switch(t.data_type())
            {
            case onnx::TensorProto::UNDEFINED: throw std::runtime_error("");
            case onnx::TensorProto::FLOAT: return literal{{shape::float_type, dims}, s.data()};
            case onnx::TensorProto::UINT8: throw std::runtime_error("");
            case onnx::TensorProto::INT8: return literal{{shape::int32_type, dims}, s.data()};
            case onnx::TensorProto::UINT16: return literal{{shape::int32_type, dims}, s.data()};
            case onnx::TensorProto::INT16: return literal{{shape::int32_type, dims}, s.data()};
            case onnx::TensorProto::INT32: return literal{{shape::int32_type, dims}, s.data()};
            case onnx::TensorProto::INT64: return literal{{shape::int64_type, dims}, s.data()};
            case onnx::TensorProto::STRING: throw std::runtime_error("");
            case onnx::TensorProto::BOOL: return literal{{shape::int32_type, dims}, s.data()};
            case onnx::TensorProto::FLOAT16: throw std::runtime_error("");
            case onnx::TensorProto::DOUBLE: return literal{{shape::double_type, dims}, s.data()};
            case onnx::TensorProto::UINT32: throw std::runtime_error("");
            case onnx::TensorProto::UINT64: throw std::runtime_error("");
            case onnx::TensorProto::COMPLEX64: throw std::runtime_error("");
            case onnx::TensorProto::COMPLEX128: throw std::runtime_error("");
            }
            MIGRAPH_THROW("Invalid tensor type");
594
        }
Paul's avatar
Paul committed
595
596
597
598
        switch(t.data_type())
        {
        case onnx::TensorProto::UNDEFINED: throw std::runtime_error("");
        case onnx::TensorProto::FLOAT:
Paul's avatar
Paul committed
599
            return literal{{shape::float_type, dims}, t.float_data().begin(), t.float_data().end()};
Paul's avatar
Paul committed
600
601
        case onnx::TensorProto::UINT8: throw std::runtime_error("");
        case onnx::TensorProto::INT8:
Paul's avatar
Paul committed
602
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
603
        case onnx::TensorProto::UINT16:
Paul's avatar
Paul committed
604
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
605
        case onnx::TensorProto::INT16:
Paul's avatar
Paul committed
606
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
607
        case onnx::TensorProto::INT32:
Paul's avatar
Paul committed
608
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
609
        case onnx::TensorProto::INT64:
Paul's avatar
Paul committed
610
            return literal{{shape::int64_type, dims}, t.int64_data().begin(), t.int64_data().end()};
Paul's avatar
Paul committed
611
612
        case onnx::TensorProto::STRING: throw std::runtime_error("");
        case onnx::TensorProto::BOOL:
Paul's avatar
Paul committed
613
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
614
615
616
617
618
619
620
621
622
        case onnx::TensorProto::FLOAT16: throw std::runtime_error("");
        case onnx::TensorProto::DOUBLE:
            return literal{
                {shape::double_type, dims}, t.double_data().begin(), t.double_data().end()};
        case onnx::TensorProto::UINT32: throw std::runtime_error("");
        case onnx::TensorProto::UINT64: throw std::runtime_error("");
        case onnx::TensorProto::COMPLEX64: throw std::runtime_error("");
        case onnx::TensorProto::COMPLEX128: throw std::runtime_error("");
        }
Paul's avatar
Paul committed
623
        MIGRAPH_THROW("Invalid tensor type");
Paul's avatar
Paul committed
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
    }

    static shape parse_type(const onnx::TypeProto& t)
    {
        shape::type_t shape_type{};
        switch(t.tensor_type().elem_type())
        {
        case onnx::TensorProto::UNDEFINED:
            break; // throw std::runtime_error("Unsupported type UNDEFINED");
        case onnx::TensorProto::FLOAT: shape_type = shape::float_type; break;
        case onnx::TensorProto::UINT8:
            break; // throw std::runtime_error("Unsupported type UINT8");
        case onnx::TensorProto::INT8: shape_type = shape::int8_type; break;
        case onnx::TensorProto::UINT16: shape_type = shape::uint16_type; break;
        case onnx::TensorProto::INT16: shape_type = shape::int16_type; break;
        case onnx::TensorProto::INT32: shape_type = shape::int32_type; break;
        case onnx::TensorProto::INT64: shape_type = shape::int64_type; break;
        case onnx::TensorProto::STRING:
            break; // throw std::runtime_error("Unsupported type STRING");
        case onnx::TensorProto::BOOL:
            break; // throw std::runtime_error("Unsupported type BOOL");
        case onnx::TensorProto::FLOAT16:
            break; // throw std::runtime_error("Unsupported type FLOAT16");
        case onnx::TensorProto::DOUBLE: shape_type = shape::double_type; break;
        case onnx::TensorProto::UINT32: shape_type = shape::uint32_type; break;
        case onnx::TensorProto::UINT64: shape_type = shape::uint64_type; break;
        case onnx::TensorProto::COMPLEX64:
            break; // throw std::runtime_error("Unsupported type COMPLEX64");
        case onnx::TensorProto::COMPLEX128:
            break; // throw std::runtime_error("Unsupported type COMPLEX128");
        }
        std::vector<std::size_t> dims;
Paul's avatar
Paul committed
656
        auto&& tensor_dims = t.tensor_type().shape().dim();
657
658
659
660
661
662
663
664
665
666
667
        std::transform(tensor_dims.begin(),
                       tensor_dims.end(),
                       std::back_inserter(dims),
                       [](auto&& d) -> std::size_t {
                           if(not d.has_dim_value())
                           {
                               long default_batch_size = 1; // FIXME
                               return default_batch_size;
                           }
                           return d.dim_value();
                       });
Paul's avatar
Paul committed
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
        return {shape_type, dims};
    }
};

program parse_onnx(const std::string& name)
{
    std::fstream input(name.c_str(), std::ios::in | std::ios::binary);
    onnx_parser parser;
#ifndef NDEBUG
    // Log the program when it can't be parsed
    try
    {
        parser.parse_from(input);
    }
    catch(...)
    {
        std::cerr << parser.prog << std::endl;
        throw;
    }
#else
    parser.parse_from(input);
#endif
    return std::move(parser.prog);
}

Paul's avatar
Paul committed
693
} // namespace migraph