onnx.cpp 26.7 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>
17
#include <migraph/config.hpp>
Paul's avatar
Paul committed
18

Paul's avatar
Paul committed
19
namespace migraph {
20
inline namespace MIGRAPH_INLINE_NS {
Paul's avatar
Paul committed
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
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
43
    using op_func = std::function<instruction_ref(attribute_map, std::vector<instruction_ref>)>;
Paul's avatar
Paul committed
44
45
46
47
48
49
50
51
    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
52
        add_generic_op("MatMul", op::dot{});
Khalique's avatar
Khalique committed
53
        add_generic_op("Relu", op::relu{});
Khalique's avatar
Khalique committed
54
55
        // disable dropout for inference
        add_generic_op("Dropout", op::identity{});
Paul's avatar
Paul committed
56

57
58
59
60
61
        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
62

Khalique's avatar
Khalique committed
63
        add_mem_op("ImageScaler", &onnx_parser::parse_imagescaler);
64
        add_mem_op("LeakyRelu", &onnx_parser::parse_leaky_relu);
Paul's avatar
Paul committed
65
66
        add_mem_op("Constant", &onnx_parser::parse_constant);
        add_mem_op("Conv", &onnx_parser::parse_conv);
Paul's avatar
Paul committed
67
68
        add_mem_op("MaxPool", &onnx_parser::parse_pooling);
        add_mem_op("AveragePool", &onnx_parser::parse_pooling);
69
70
        add_mem_op("GlobalMaxPool", &onnx_parser::parse_pooling);
        add_mem_op("GlobalAveragePool", &onnx_parser::parse_pooling);
Paul's avatar
Paul committed
71
        add_mem_op("Reshape", &onnx_parser::parse_reshape);
Paul's avatar
Paul committed
72
73
        add_mem_op("Flatten", &onnx_parser::parse_flatten);
        add_mem_op("Gemm", &onnx_parser::parse_gemm);
74
        add_mem_op("BatchNormalization", &onnx_parser::parse_batchnorm);
Paul's avatar
Paul committed
75
        add_mem_op("Softmax", &onnx_parser::parse_softmax);
76
77
78
        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
79
        add_mem_op("Concat", &onnx_parser::parse_concat);
Khalique's avatar
Khalique committed
80
        add_mem_op("Transpose", &onnx_parser::parse_transpose);
Paul's avatar
Paul committed
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
    }

    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)...);
        });
    }
96
97
98
99
    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
100
            if(args.size() != 2)
Paul's avatar
Paul committed
101
                MIGRAPH_THROW("binary operators should have 2 operands");
102
103
104
105
106
107
108
109
110
111
112
113
            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);
                }
114
                return prog.add_instruction(x, args);
115
            }
116
            else if(args[0]->get_shape() != args[1]->get_shape())
117
118
119
120
            {
                // Example:
                // s0 = (3,2,4,5) and s1 = (2,1,1)
                //
Scott Thornton's avatar
Scott Thornton committed
121
122
                // In this case we need to broadcast (:,1,1) portion of
                // s1 plus broadcast the 1st dimension of s1
123
124
125
126
127
128
129
130
131
                // 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
Paul's avatar
Paul committed
132
133
134
135
136
137
                const std::vector<std::size_t>* s0 = &args[0]->get_shape().lens();
                const std::vector<std::size_t>* s1 = &args[1]->get_shape().lens();

                // Make sure s0 is the smaller size
                if(s0->size() > s1->size())
                    std::swap(s0, s1);
138
139

                // Copy the larger vector to output_lens
Paul's avatar
Paul committed
140
141
                std::vector<std::size_t> output_lens(s1->size());
                auto offset = s1->size() - s0->size();
Paul's avatar
Paul committed
142
143
144
145
146
147
                std::transform(s0->begin(),
                               s0->end(),
                               s1->begin() + offset,
                               output_lens.begin() + offset,
                               [](auto a, auto b) { return std::max(a, b); });

148
149
150
                auto l0 = prog.add_instruction(op::multibroadcast{output_lens}, args[0]);
                auto l1 = prog.add_instruction(op::multibroadcast{output_lens}, args[1]);
                return prog.add_instruction(x, l0, l1);
Paul's avatar
Paul committed
151
152
            }
            else
153
154
            {
                return prog.add_instruction(x, args);
155
156
157
158
            }
        });
    }

Paul's avatar
Paul committed
159
    template <class T>
Paul's avatar
Paul committed
160
161
    void add_generic_op(std::string name, T x)
    {
Paul's avatar
Paul committed
162
        ops.emplace(name, [this, x](attribute_map, std::vector<instruction_ref> args) {
Paul's avatar
Paul committed
163
164
165
166
            return prog.add_instruction(x, args);
        });
    }

Paul's avatar
Paul committed
167
    instruction_ref
Paul's avatar
Paul committed
168
    parse_softmax(const std::string&, const attribute_map&, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
169
170
    {
        auto dims = args.front()->get_shape().lens();
Scott Thornton's avatar
Scott Thornton committed
171
172
        auto r =
            prog.add_instruction(op::reshape{{long(dims[0]), long(dims[1]), 1, 1}}, args.front());
173
174
        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
175
176
    }

Paul's avatar
Paul committed
177
    instruction_ref
Paul's avatar
Paul committed
178
    parse_conv(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
179
    {
180
        op::convolution op;
Paul's avatar
Paul committed
181
182
183
        if(contains(attributes, "pads"))
        {
            copy(attributes["pads"].ints(), op.padding.begin());
Paul's avatar
Paul committed
184
        }
Paul's avatar
Paul committed
185
186
187
188
189
190
191
192
193
194
195
196
        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
197
            auto l2       = prog.add_instruction(op::broadcast{axis, l1->get_shape()}, args[2]);
198
            return prog.add_instruction(op::add{}, l1, l2);
Paul's avatar
Paul committed
199
        }
Paul's avatar
Paul committed
200
201
        return prog.add_instruction(op, args);
    }
Paul's avatar
Paul committed
202

Paul's avatar
Paul committed
203
204
205
    instruction_ref parse_pooling(const std::string& name,
                                  attribute_map attributes,
                                  std::vector<instruction_ref> args)
Paul's avatar
Paul committed
206
    {
Khalique's avatar
Khalique committed
207
208
        op::pooling op{ends_with(name, "MaxPool") ? "max" : "average"};
        if(starts_with(name, "Global"))
209
        {
Khalique's avatar
Khalique committed
210
211
            auto lens  = args.front()->get_shape().lens();
            op.lengths = {lens[2], lens[3]};
212
        }
Paul's avatar
Paul committed
213
214
215
216
217
218
219
220
221
222
223
224
        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
225
        return prog.add_instruction(op, std::move(args));
Paul's avatar
Paul committed
226
227
    }

Paul's avatar
Paul committed
228
    instruction_ref
Paul's avatar
Paul committed
229
    parse_reshape(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
230
    {
231
        op::reshape op;
Paul's avatar
Paul committed
232
233
234
235
236
237
238
        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
239
            literal s = args[1]->get_literal();
Paul's avatar
Paul committed
240
            s.visit([&](auto v) { copy(v, std::back_inserter(op.dims)); });
Paul's avatar
Paul committed
241
        }
Paul's avatar
Paul committed
242
243
244
        return prog.add_instruction(op, args[0]);
    }

Paul's avatar
Paul committed
245
    instruction_ref
Paul's avatar
Paul committed
246
    parse_flatten(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
247
248
    {
        uint64_t axis = 0;
Paul's avatar
Paul committed
249
250
251
252
        if(contains(attributes, "axis"))
        {
            axis = parse_value(attributes.at("axis")).at<int>();
        }
253
        return prog.add_instruction(op::flatten{axis}, args[0]);
Paul's avatar
Paul committed
254
255
    }

256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
    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
274
275
276
277
278
279
280
    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));
    }
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301

    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
302
303
304
    instruction_ref parse_constant(const std::string&,
                                   attribute_map attributes,
                                   const std::vector<instruction_ref>&)
Paul's avatar
Paul committed
305
306
307
308
    {
        literal v = parse_value(attributes.at("value"));
        return prog.add_literal(v);
    }
Paul's avatar
Paul committed
309

Paul's avatar
Paul committed
310
    instruction_ref
Paul's avatar
Paul committed
311
    parse_gemm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
    {
        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};
334
335
        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
336
337
338
        if(args.size() == 3)
        {
            uint64_t axis = 1;
Shucai Xiao's avatar
Shucai Xiao committed
339
            auto l3       = prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Scott Thornton's avatar
Scott Thornton committed
340
            auto l4       = prog.add_instruction(op::broadcast{axis, l3->get_shape()}, args[2]);
341
            return prog.add_instruction(op::add{}, l3, l4);
Paul's avatar
Paul committed
342
        }
Shucai Xiao's avatar
Shucai Xiao committed
343
        return prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Paul's avatar
Paul committed
344
345
    }

346
    instruction_ref
Paul's avatar
Paul committed
347
    parse_batchnorm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
348
    {
Scott Thornton's avatar
Scott Thornton committed
349
350
        float epsilon                                     = 1e-5f;
        float momentum                                    = 0.9f;
351
        op::batch_norm_inference::bn_infer_mode_t bn_mode = op::batch_norm_inference::spatial;
Scott Thornton's avatar
Scott Thornton committed
352
        bool is_test                                      = false;
353
354
355
356
357
358
        if(contains(attributes, "epsilon"))
        {
            epsilon = parse_value(attributes.at("epsilon")).at<float>();
        }
        if(contains(attributes, "momentum"))
        {
359
            momentum = parse_value(attributes.at("momentum")).at<float>();
360
361
362
        }
        if(contains(attributes, "is_test"))
        {
wsttiger's avatar
wsttiger committed
363
            is_test = parse_value(attributes.at("is_test")).at<uint64_t>() > 0;
364
365
366
        }
        if(contains(attributes, "spatial"))
        {
367
            bn_mode = (parse_value(attributes.at("spatial")).at<uint64_t>() > 0)
368
369
                          ? op::batch_norm_inference::spatial
                          : op::batch_norm_inference::per_activation;
370
        }
Paul's avatar
Paul committed
371
        (void)is_test;
Paul's avatar
Paul committed
372
        op::batch_norm_inference op{epsilon, momentum, bn_mode};
Paul's avatar
Paul committed
373
        return prog.add_instruction(op, std::move(args));
374
375
    }

376
377
378
379
    instruction_ref parse_leaky_relu(const std::string&,
                                     attribute_map attributes,
                                     std::vector<instruction_ref> args)
    {
Khalique's avatar
Khalique committed
380
        float alpha = 0.01; // default alpha val for leaky relu
381
382
383
384
385
386
387
388
        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
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
    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
406

Khalique's avatar
Khalique committed
407
408
409
410
411
412
413
414
415
        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
416

Khalique's avatar
Khalique committed
417
418
    instruction_ref
    parse_transpose(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Khalique's avatar
Khalique committed
419
420
421
422
423
424
425
426
427
428
    {
        std::vector<int64_t> perm{};
        if(contains(attributes, "perm"))
        {
            auto&& perm_vals = attributes["perm"].ints();
            perm             = std::vector<int64_t>(perm_vals.begin(), perm_vals.end());
        }
        return prog.add_instruction(migraph::op::transpose{perm}, args.front());
    }

Paul's avatar
Paul committed
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
    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);
448
449
450
451
452
        std::unordered_map<std::string, onnx::TensorProto> initializer_data;
        for(auto&& f : graph.initializer())
        {
            initializer_data[f.name()] = f;
        }
Paul's avatar
Paul committed
453
454
455
        for(auto&& input : graph.input())
        {
            const std::string& name = input.name();
456
457
458
459
460
461
462
463
464
465
466
467
            // 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
468
469
470
        }
        for(auto&& p : nodes)
        {
471
            this->parse_node(get_name(p.second));
Paul's avatar
Paul committed
472
473
474
        }
    }

Paul's avatar
Paul committed
475
    void parse_node(const std::string& name)
Paul's avatar
Paul committed
476
    {
Paul's avatar
Paul committed
477
        if(name.empty())
Paul's avatar
Paul committed
478
            MIGRAPH_THROW("Onnx node must have a name");
Paul's avatar
Paul committed
479
480
481
482
483
484
485
486
        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)
                {
487
                    auto&& iname = get_name(nodes.at(input));
Paul's avatar
Paul committed
488
                    assert(name != iname);
Paul's avatar
Paul committed
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
515
516
517
                    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;
    }

518
519
520
521
522
    static std::string get_name(const onnx::NodeProto& node)
    {
        if(node.name().empty())
        {
            std::string generated = "migraph_unnamed_node";
Paul's avatar
Paul committed
523
524
525
526
            return std::accumulate(node.output().begin(),
                                   node.output().end(),
                                   generated,
                                   [](auto x, auto y) { return x + "_" + y; });
527
528
529
530
        }
        return node.name();
    }

Paul's avatar
Paul committed
531
532
533
534
535
    static node_map get_nodes(const onnx::GraphProto& graph)
    {
        std::unordered_map<std::string, onnx::NodeProto> result;
        for(auto&& node : graph.node())
        {
536
            result[get_name(node)] = node;
Paul's avatar
Paul committed
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
            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
562
        case onnx::AttributeProto::FLOATS: return from_repeated(shape::float_type, attr.floats());
Paul's avatar
Paul committed
563
564
565
566
567
        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
568
        MIGRAPH_THROW("Invalid attribute type");
Paul's avatar
Paul committed
569
570
571
572
573
    }

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

    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");
Paul's avatar
Paul committed
649
        case onnx::TensorProto::FLOAT16: shape_type = shape::half_type; break;
Paul's avatar
Paul committed
650
651
652
653
654
655
656
657
658
        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
659
        auto&& tensor_dims = t.tensor_type().shape().dim();
660
661
662
663
664
665
666
667
668
669
670
        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
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
        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);
}

696
} // namespace MIGRAPH_INLINE_NS
Paul's avatar
Paul committed
697
} // namespace migraph