onnx.cpp 27.7 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
16
17
18
19
#include <migraphx/fallthrough.hpp>
#include <migraphx/program.hpp>
#include <migraphx/operators.hpp>
#include <migraphx/ranges.hpp>
#include <migraphx/instruction.hpp>
#include <migraphx/config.hpp>

namespace migraphx {
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
        add_generic_op("Sigmoid", op::sigmoid{});
        add_generic_op("Abs", op::abs{});
Shucai Xiao's avatar
Shucai Xiao committed
56
57
        add_generic_op("Exp", op::exp{});
        add_generic_op("Log", op::log{});
Khalique's avatar
Khalique committed
58
59
        // disable dropout for inference
        add_generic_op("Dropout", op::identity{});
Khalique's avatar
Khalique committed
60
        add_generic_op("Identity", op::identity{});
Shucai Xiao's avatar
Shucai Xiao committed
61
62
63
        add_generic_op("Sin", op::sin{});
        add_generic_op("Cos", op::cos{});
        add_generic_op("Tan", op::tan{});
64
65
        add_generic_op("Sinh", op::sinh{});
        add_generic_op("Cosh", op::cosh{});
66
        add_generic_op("Tanh", op::tanh{});
67
68
69
        add_generic_op("Asin", op::asin{});
        add_generic_op("Acos", op::acos{});
        add_generic_op("Atan", op::atan{});
Paul's avatar
Paul committed
70

71
72
73
74
75
        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
76

Khalique's avatar
Khalique committed
77
        add_mem_op("ImageScaler", &onnx_parser::parse_imagescaler);
78
        add_mem_op("LeakyRelu", &onnx_parser::parse_leaky_relu);
Khalique's avatar
Khalique committed
79
        add_mem_op("Elu", &onnx_parser::parse_elu);
Paul's avatar
Paul committed
80
81
        add_mem_op("Constant", &onnx_parser::parse_constant);
        add_mem_op("Conv", &onnx_parser::parse_conv);
Paul's avatar
Paul committed
82
83
        add_mem_op("MaxPool", &onnx_parser::parse_pooling);
        add_mem_op("AveragePool", &onnx_parser::parse_pooling);
84
85
        add_mem_op("GlobalMaxPool", &onnx_parser::parse_pooling);
        add_mem_op("GlobalAveragePool", &onnx_parser::parse_pooling);
Paul's avatar
Paul committed
86
        add_mem_op("Reshape", &onnx_parser::parse_reshape);
Paul's avatar
Paul committed
87
88
        add_mem_op("Flatten", &onnx_parser::parse_flatten);
        add_mem_op("Gemm", &onnx_parser::parse_gemm);
89
        add_mem_op("BatchNormalization", &onnx_parser::parse_batchnorm);
Paul's avatar
Paul committed
90
        add_mem_op("Softmax", &onnx_parser::parse_softmax);
91
92
93
        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
94
        add_mem_op("Concat", &onnx_parser::parse_concat);
Khalique's avatar
Khalique committed
95
        add_mem_op("Transpose", &onnx_parser::parse_transpose);
Paul's avatar
Paul committed
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
    }

    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)...);
        });
    }
111
112
113
114
    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
115
            if(args.size() != 2)
Paul's avatar
Paul committed
116
                MIGRAPH_THROW("binary operators should have 2 operands");
117
118
119
120
121
122
123
124
125
126
127
128
            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);
                }
129
                return prog.add_instruction(x, args);
130
            }
131
            else if(args[0]->get_shape() != args[1]->get_shape())
132
133
134
135
            {
                // Example:
                // s0 = (3,2,4,5) and s1 = (2,1,1)
                //
Scott Thornton's avatar
Scott Thornton committed
136
137
                // In this case we need to broadcast (:,1,1) portion of
                // s1 plus broadcast the 1st dimension of s1
138
139
140
141
142
143
144
145
146
                // 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
147
148
149
150
151
152
                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);
153
154

                // Copy the larger vector to output_lens
Paul's avatar
Paul committed
155
156
                std::vector<std::size_t> output_lens(s1->size());
                auto offset = s1->size() - s0->size();
Paul's avatar
Paul committed
157
158
159
160
161
162
                std::transform(s0->begin(),
                               s0->end(),
                               s1->begin() + offset,
                               output_lens.begin() + offset,
                               [](auto a, auto b) { return std::max(a, b); });

163
164
165
                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
166
167
            }
            else
168
169
            {
                return prog.add_instruction(x, args);
170
171
172
173
            }
        });
    }

Paul's avatar
Paul committed
174
    template <class T>
Paul's avatar
Paul committed
175
176
    void add_generic_op(std::string name, T x)
    {
Paul's avatar
Paul committed
177
        ops.emplace(name, [this, x](attribute_map, std::vector<instruction_ref> args) {
Paul's avatar
Paul committed
178
179
180
181
            return prog.add_instruction(x, args);
        });
    }

Paul's avatar
Paul committed
182
    instruction_ref
Paul's avatar
Paul committed
183
    parse_softmax(const std::string&, const attribute_map&, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
184
185
    {
        auto dims = args.front()->get_shape().lens();
Scott Thornton's avatar
Scott Thornton committed
186
187
        auto r =
            prog.add_instruction(op::reshape{{long(dims[0]), long(dims[1]), 1, 1}}, args.front());
188
189
        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
190
191
    }

Paul's avatar
Paul committed
192
    instruction_ref
Paul's avatar
Paul committed
193
    parse_conv(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
194
    {
195
        op::convolution op;
Paul's avatar
Paul committed
196
197
198
        if(contains(attributes, "pads"))
        {
            copy(attributes["pads"].ints(), op.padding.begin());
Paul's avatar
Paul committed
199
        }
Paul's avatar
Paul committed
200
201
202
203
204
205
206
207
208
209
210
211
        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
212
            auto l2       = prog.add_instruction(op::broadcast{axis, l1->get_shape()}, args[2]);
213
            return prog.add_instruction(op::add{}, l1, l2);
Paul's avatar
Paul committed
214
        }
Paul's avatar
Paul committed
215
216
        return prog.add_instruction(op, args);
    }
Paul's avatar
Paul committed
217

Paul's avatar
Paul committed
218
219
220
    instruction_ref parse_pooling(const std::string& name,
                                  attribute_map attributes,
                                  std::vector<instruction_ref> args)
Paul's avatar
Paul committed
221
    {
Khalique's avatar
Khalique committed
222
223
        op::pooling op{ends_with(name, "MaxPool") ? "max" : "average"};
        if(starts_with(name, "Global"))
224
        {
Khalique's avatar
Khalique committed
225
226
            auto lens  = args.front()->get_shape().lens();
            op.lengths = {lens[2], lens[3]};
227
        }
Paul's avatar
Paul committed
228
229
230
231
232
233
234
235
236
237
238
239
        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
240
        return prog.add_instruction(op, std::move(args));
Paul's avatar
Paul committed
241
242
    }

Paul's avatar
Paul committed
243
    instruction_ref
Paul's avatar
Paul committed
244
    parse_reshape(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
245
    {
246
        op::reshape op;
Paul's avatar
Paul committed
247
248
249
250
251
252
253
        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
254
            literal s = args[1]->get_literal();
Paul's avatar
Paul committed
255
            s.visit([&](auto v) { copy(v, std::back_inserter(op.dims)); });
Paul's avatar
Paul committed
256
        }
Paul's avatar
Paul committed
257
258
259
        return prog.add_instruction(op, args[0]);
    }

Paul's avatar
Paul committed
260
    instruction_ref
Paul's avatar
Paul committed
261
    parse_flatten(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
262
263
    {
        uint64_t axis = 0;
Paul's avatar
Paul committed
264
265
266
267
        if(contains(attributes, "axis"))
        {
            axis = parse_value(attributes.at("axis")).at<int>();
        }
268
        return prog.add_instruction(op::flatten{axis}, args[0]);
Paul's avatar
Paul committed
269
270
    }

271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
    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
289
290
291
292
293
294
295
    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));
    }
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316

    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
317
318
319
    instruction_ref parse_constant(const std::string&,
                                   attribute_map attributes,
                                   const std::vector<instruction_ref>&)
Paul's avatar
Paul committed
320
321
322
323
    {
        literal v = parse_value(attributes.at("value"));
        return prog.add_literal(v);
    }
Paul's avatar
Paul committed
324

Paul's avatar
Paul committed
325
    instruction_ref
Paul's avatar
Paul committed
326
    parse_gemm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
    {
        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};
349
350
        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
351
352
353
        if(args.size() == 3)
        {
            uint64_t axis = 1;
Shucai Xiao's avatar
Shucai Xiao committed
354
            auto l3       = prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Scott Thornton's avatar
Scott Thornton committed
355
            auto l4       = prog.add_instruction(op::broadcast{axis, l3->get_shape()}, args[2]);
356
            return prog.add_instruction(op::add{}, l3, l4);
Paul's avatar
Paul committed
357
        }
Shucai Xiao's avatar
Shucai Xiao committed
358
        return prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Paul's avatar
Paul committed
359
360
    }

361
    instruction_ref
Paul's avatar
Paul committed
362
    parse_batchnorm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
363
    {
Scott Thornton's avatar
Scott Thornton committed
364
365
        float epsilon                                     = 1e-5f;
        float momentum                                    = 0.9f;
366
        op::batch_norm_inference::bn_infer_mode_t bn_mode = op::batch_norm_inference::spatial;
Scott Thornton's avatar
Scott Thornton committed
367
        bool is_test                                      = false;
368
369
370
371
372
373
        if(contains(attributes, "epsilon"))
        {
            epsilon = parse_value(attributes.at("epsilon")).at<float>();
        }
        if(contains(attributes, "momentum"))
        {
374
            momentum = parse_value(attributes.at("momentum")).at<float>();
375
376
377
        }
        if(contains(attributes, "is_test"))
        {
wsttiger's avatar
wsttiger committed
378
            is_test = parse_value(attributes.at("is_test")).at<uint64_t>() > 0;
379
380
381
        }
        if(contains(attributes, "spatial"))
        {
382
            bn_mode = (parse_value(attributes.at("spatial")).at<uint64_t>() > 0)
383
384
                          ? op::batch_norm_inference::spatial
                          : op::batch_norm_inference::per_activation;
385
        }
Paul's avatar
Paul committed
386
        (void)is_test;
Paul's avatar
Paul committed
387
        op::batch_norm_inference op{epsilon, momentum, bn_mode};
Paul's avatar
Paul committed
388
        return prog.add_instruction(op, std::move(args));
389
390
    }

391
392
393
394
    instruction_ref parse_leaky_relu(const std::string&,
                                     attribute_map attributes,
                                     std::vector<instruction_ref> args)
    {
Khalique's avatar
Khalique committed
395
        float alpha = 0.01; // default alpha val for leaky relu
396
397
398
399
400
401
402
403
        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
404
405
    instruction_ref
    parse_elu(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Khalique's avatar
Khalique committed
406
407
408
409
410
411
412
413
414
415
    {
        float alpha = 1.0; // default alpha val for elu
        if(contains(attributes, "alpha"))
        {
            alpha = parse_value(attributes.at("alpha")).at<float>();
        }
        op::elu op{alpha};
        return prog.add_instruction(op, args.front());
    }

Khalique's avatar
Khalique committed
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
    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
433

Khalique's avatar
Khalique committed
434
435
        auto scale_val = prog.add_literal(scale);
        auto bias_vals = prog.add_literal(
Paul's avatar
Paul committed
436
            migraphx::literal{migraphx::shape{migraphx::shape::float_type, {bias.size()}}, bias});
Khalique's avatar
Khalique committed
437

Paul's avatar
Paul committed
438
439
        auto scale_tensor = prog.add_instruction(migraphx::op::scalar{input_shape}, scale_val);
        auto img_scaled   = prog.add_instruction(migraphx::op::mul{}, args.front(), scale_tensor);
Paul's avatar
Paul committed
440
        auto bias_bcast = prog.add_instruction(migraphx::op::broadcast{1, input_shape}, bias_vals);
Paul's avatar
Paul committed
441
        return prog.add_instruction(migraphx::op::add{}, img_scaled, bias_bcast);
Khalique's avatar
Khalique committed
442
    }
Khalique's avatar
Khalique committed
443

Khalique's avatar
Khalique committed
444
445
    instruction_ref
    parse_transpose(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Khalique's avatar
Khalique committed
446
447
448
449
450
451
452
    {
        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());
        }
Paul's avatar
Paul committed
453
        return prog.add_instruction(migraphx::op::transpose{perm}, args.front());
Khalique's avatar
Khalique committed
454
455
    }

Paul's avatar
Paul committed
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
    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);
475
476
477
478
479
        std::unordered_map<std::string, onnx::TensorProto> initializer_data;
        for(auto&& f : graph.initializer())
        {
            initializer_data[f.name()] = f;
        }
Paul's avatar
Paul committed
480
481
482
        for(auto&& input : graph.input())
        {
            const std::string& name = input.name();
483
484
485
486
487
488
489
490
491
492
493
494
            // 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
495
496
497
        }
        for(auto&& p : nodes)
        {
498
            this->parse_node(get_name(p.second));
Paul's avatar
Paul committed
499
500
501
        }
    }

Paul's avatar
Paul committed
502
    void parse_node(const std::string& name)
Paul's avatar
Paul committed
503
    {
Paul's avatar
Paul committed
504
        if(name.empty())
Paul's avatar
Paul committed
505
            MIGRAPH_THROW("Onnx node must have a name");
Paul's avatar
Paul committed
506
507
508
509
510
511
512
513
        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)
                {
514
                    auto&& iname = get_name(nodes.at(input));
Paul's avatar
Paul committed
515
                    assert(name != iname);
Paul's avatar
Paul committed
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
                    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;
    }

545
546
547
548
    static std::string get_name(const onnx::NodeProto& node)
    {
        if(node.name().empty())
        {
Paul's avatar
Paul committed
549
            std::string generated = "migraphx_unnamed_node";
Paul's avatar
Paul committed
550
551
552
553
            return std::accumulate(node.output().begin(),
                                   node.output().end(),
                                   generated,
                                   [](auto x, auto y) { return x + "_" + y; });
554
555
556
557
        }
        return node.name();
    }

Paul's avatar
Paul committed
558
559
560
561
562
    static node_map get_nodes(const onnx::GraphProto& graph)
    {
        std::unordered_map<std::string, onnx::NodeProto> result;
        for(auto&& node : graph.node())
        {
563
            result[get_name(node)] = node;
Paul's avatar
Paul committed
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
            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
589
        case onnx::AttributeProto::FLOATS: return from_repeated(shape::float_type, attr.floats());
Paul's avatar
Paul committed
590
591
592
593
594
        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
595
        MIGRAPH_THROW("Invalid attribute type");
Paul's avatar
Paul committed
596
597
598
599
600
    }

    static literal parse_tensor(const onnx::TensorProto& t)
    {
        std::vector<std::size_t> dims(t.dims().begin(), t.dims().end());
601
602
        if(t.has_raw_data())
        {
wsttiger's avatar
wsttiger committed
603
            const std::string& s = t.raw_data();
Scott Thornton's avatar
Scott Thornton committed
604
605
606
607
608
609
610
611
612
613
614
615
            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
616
            case onnx::TensorProto::FLOAT16: return literal{{shape::half_type, dims}, s.data()};
Scott Thornton's avatar
Scott Thornton committed
617
618
619
620
621
622
623
            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");
624
        }
Paul's avatar
Paul committed
625
626
627
628
        switch(t.data_type())
        {
        case onnx::TensorProto::UNDEFINED: throw std::runtime_error("");
        case onnx::TensorProto::FLOAT:
Paul's avatar
Paul committed
629
            return literal{{shape::float_type, dims}, t.float_data().begin(), t.float_data().end()};
Paul's avatar
Paul committed
630
631
        case onnx::TensorProto::UINT8: throw std::runtime_error("");
        case onnx::TensorProto::INT8:
Paul's avatar
Paul committed
632
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
633
        case onnx::TensorProto::UINT16:
Paul's avatar
Paul committed
634
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
635
        case onnx::TensorProto::INT16:
Paul's avatar
Paul committed
636
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
637
        case onnx::TensorProto::INT32:
Paul's avatar
Paul committed
638
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
639
        case onnx::TensorProto::INT64:
Paul's avatar
Paul committed
640
            return literal{{shape::int64_type, dims}, t.int64_data().begin(), t.int64_data().end()};
Paul's avatar
Paul committed
641
642
        case onnx::TensorProto::STRING: throw std::runtime_error("");
        case onnx::TensorProto::BOOL:
Paul's avatar
Paul committed
643
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
644
645
        case onnx::TensorProto::FLOAT16:
            return literal{{shape::half_type, dims}, t.float_data().begin(), t.float_data().end()};
Paul's avatar
Paul committed
646
647
648
649
650
651
652
653
        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
654
        MIGRAPH_THROW("Invalid tensor type");
Paul's avatar
Paul committed
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
    }

    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
676
        case onnx::TensorProto::FLOAT16: shape_type = shape::half_type; break;
Paul's avatar
Paul committed
677
678
679
680
681
682
683
684
685
        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
686
        auto&& tensor_dims = t.tensor_type().shape().dim();
687
688
689
690
691
692
693
694
695
696
697
        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
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
        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);
}

723
} // namespace MIGRAPH_INLINE_NS
Paul's avatar
Paul committed
724
} // namespace migraphx