onnx.cpp 53 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
#include <migraphx/fallthrough.hpp>
#include <migraphx/program.hpp>
#include <migraphx/operators.hpp>
#include <migraphx/ranges.hpp>
#include <migraphx/instruction.hpp>
#include <migraphx/config.hpp>
18
#include <migraphx/onnx.hpp>
Paul's avatar
Paul committed
19
20

namespace migraphx {
Paul's avatar
Paul committed
21
inline namespace MIGRAPHX_INLINE_NS {
Paul's avatar
Paul committed
22
23
24
25
26

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
27
28
    using op_func =
        std::function<std::vector<instruction_ref>(attribute_map, std::vector<instruction_ref>)>;
Paul's avatar
Paul committed
29
30
    node_map nodes;
    std::unordered_map<std::string, instruction_ref> instructions;
Scott Thornton's avatar
Scott Thornton committed
31
    program prog    = program();
32
    bool is_pytorch = false;
Paul's avatar
Paul committed
33
34

    std::unordered_map<std::string, op_func> ops;
35
    std::unordered_map<std::string, operation> map_actv_funcs;
Paul's avatar
Paul committed
36
37
38

    onnx_parser()
    {
Shucai Xiao's avatar
Shucai Xiao committed
39
        add_generic_op("MatMul", op::dot{});
Khalique's avatar
Khalique committed
40
        add_generic_op("Relu", op::relu{});
Khalique's avatar
Khalique committed
41
42
        add_generic_op("Sigmoid", op::sigmoid{});
        add_generic_op("Abs", op::abs{});
Shucai Xiao's avatar
Shucai Xiao committed
43
44
        add_generic_op("Exp", op::exp{});
        add_generic_op("Log", op::log{});
Khalique's avatar
Khalique committed
45
46
        // disable dropout for inference
        add_generic_op("Dropout", op::identity{});
Khalique's avatar
Khalique committed
47
        add_generic_op("Identity", op::identity{});
Shucai Xiao's avatar
Shucai Xiao committed
48
49
50
        add_generic_op("Sin", op::sin{});
        add_generic_op("Cos", op::cos{});
        add_generic_op("Tan", op::tan{});
51
52
        add_generic_op("Sinh", op::sinh{});
        add_generic_op("Cosh", op::cosh{});
53
        add_generic_op("Tanh", op::tanh{});
54
55
56
        add_generic_op("Asin", op::asin{});
        add_generic_op("Acos", op::acos{});
        add_generic_op("Atan", op::atan{});
Paul's avatar
Paul committed
57

Khalique's avatar
Khalique committed
58
59
60
61
62
        add_binary_op("Add", op::add{});
        add_binary_op("Div", op::div{});
        add_binary_op("Mul", op::mul{});
        add_binary_op("Sub", op::sub{});

Khalique's avatar
Khalique committed
63
64
65
        add_variadic_op("Sum", op::add{});
        add_variadic_op("Max", op::max{});
        add_variadic_op("Min", op::min{});
Paul's avatar
Paul committed
66

Khalique's avatar
Khalique committed
67
        add_mem_op("LRN", &onnx_parser::parse_lrn);
Khalique's avatar
Khalique committed
68
        add_mem_op("ImageScaler", &onnx_parser::parse_imagescaler);
69
        add_mem_op("LeakyRelu", &onnx_parser::parse_leaky_relu);
Khalique's avatar
Khalique committed
70
        add_mem_op("Elu", &onnx_parser::parse_elu);
Paul's avatar
Paul committed
71
72
        add_mem_op("Constant", &onnx_parser::parse_constant);
        add_mem_op("Conv", &onnx_parser::parse_conv);
Paul's avatar
Paul committed
73
74
        add_mem_op("MaxPool", &onnx_parser::parse_pooling);
        add_mem_op("AveragePool", &onnx_parser::parse_pooling);
75
76
        add_mem_op("GlobalMaxPool", &onnx_parser::parse_pooling);
        add_mem_op("GlobalAveragePool", &onnx_parser::parse_pooling);
Paul's avatar
Paul committed
77
        add_mem_op("Reshape", &onnx_parser::parse_reshape);
Paul's avatar
Paul committed
78
79
        add_mem_op("Flatten", &onnx_parser::parse_flatten);
        add_mem_op("Gemm", &onnx_parser::parse_gemm);
80
        add_mem_op("BatchNormalization", &onnx_parser::parse_batchnorm);
Paul's avatar
Paul committed
81
        add_mem_op("Softmax", &onnx_parser::parse_softmax);
82
83
84
        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
85
        add_mem_op("Concat", &onnx_parser::parse_concat);
86
87
88
        add_mem_op("Gather", &onnx_parser::parse_gather);
        add_mem_op("Shape", &onnx_parser::parse_shape);
        add_mem_op("ConstantFill", &onnx_parser::parse_constant_fill);
Khalique's avatar
Khalique committed
89
        add_mem_op("Transpose", &onnx_parser::parse_transpose);
Shucai Xiao's avatar
Shucai Xiao committed
90
        add_mem_op("RNN", &onnx_parser::parse_rnn);
91
        add_mem_op("GRU", &onnx_parser::parse_gru);
Shucai Xiao's avatar
Shucai Xiao committed
92
        add_mem_op("LSTM", &onnx_parser::parse_lstm);
Khalique's avatar
Khalique committed
93
        add_mem_op("Pad", &onnx_parser::parse_pad);
94
95
96
97
98
99
100

        // init the activation function map
        init_actv_func();
    }

    void init_actv_func()
    {
101
102
103
104
105
        map_actv_funcs.insert(std::make_pair("tanh", op::tanh{}));
        map_actv_funcs.insert(std::make_pair("relu", op::relu{}));
        map_actv_funcs.insert(std::make_pair("sigmoid", op::sigmoid{}));
        map_actv_funcs.insert(std::make_pair("leakyrelu", op::leaky_relu{}));
        map_actv_funcs.insert(std::make_pair("elu", op::elu{}));
Paul's avatar
Paul committed
106
107
108
109
    }

    template <class F>
    void add_op(std::string name, F f)
Paul's avatar
Paul committed
110
111
112
113
114
115
116
117
118
    {
        ops.emplace(name, [=](auto&&... xs) {
            return std::vector<instruction_ref>{f(std::forward<decltype(xs)>(xs)...)};
        });
    }

    // Multi output op
    template <class F>
    void add_multi_op(std::string name, F f)
Paul's avatar
Paul committed
119
120
121
122
123
124
125
    {
        ops.emplace(name, f);
    }

    template <class F>
    void add_mem_op(std::string name, F f)
    {
Paul's avatar
Paul committed
126
        add_op(name, [=](auto&&... xs) {
Paul's avatar
Paul committed
127
128
129
            return std::mem_fn(f)(*this, name, std::forward<decltype(xs)>(xs)...);
        });
    }
Khalique's avatar
Khalique committed
130

131
    template <class T>
Khalique's avatar
Khalique committed
132
    void add_binary_op(std::string name, T x)
133
    {
Paul's avatar
Paul committed
134
        add_op(name, [this, x](attribute_map attributes, std::vector<instruction_ref> args) {
Scott Thornton's avatar
Scott Thornton committed
135
            if(args.size() != 2)
Paul's avatar
Paul committed
136
                MIGRAPHX_THROW("binary operators should have 2 operands");
137
            if(contains(attributes, "broadcast") and contains(attributes, "axis"))
138
139
140
141
            {
                uint64_t broadcasted = parse_value(attributes.at("broadcast")).at<uint64_t>();
                if(broadcasted != 0)
                {
142
                    uint64_t axis = parse_value(attributes.at("axis")).at<uint64_t>();
143
144
145
146
                    auto l =
                        prog.add_instruction(op::broadcast{axis, args[0]->get_shape()}, args[1]);
                    return prog.add_instruction(x, args[0], l);
                }
147
                return prog.add_instruction(x, args);
148
            }
Paul's avatar
Paul committed
149
            else
150
            {
Khalique's avatar
Khalique committed
151
                return add_broadcastable_binary_op(args[0], args[1], x);
152
153
154
155
            }
        });
    }

Khalique's avatar
Khalique committed
156
157
158
159
160
    template <class T>
    instruction_ref add_broadcastable_binary_op(instruction_ref arg0, instruction_ref arg1, T x)
    {
        if(arg0->get_shape() != arg1->get_shape())
        {
Khalique's avatar
Khalique committed
161
162
163
164
165
166
167
168
169
170
171
172
173
            // Example:
            // s0 = (3,2,4,5) and s1 = (2,1,1)
            //
            // In this case we need to broadcast (:,1,1) portion of
            // s1 plus broadcast the 1st dimension of s1
            // 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)
            //
Khalique's avatar
Khalique committed
174
175
176
177
178
179
180
181
            // Get lengths for both arguments
            const std::vector<std::size_t>* s0 = &arg0->get_shape().lens();
            const std::vector<std::size_t>* s1 = &arg1->get_shape().lens();

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

Khalique's avatar
Khalique committed
182
            std::vector<std::size_t> output_lens(*s1);
Khalique's avatar
Khalique committed
183
184
            auto offset = s1->size() - s0->size();
            std::transform(s0->begin(),
Khalique's avatar
Khalique committed
185
186
187
188
                           s0->end(),
                           s1->begin() + offset,
                           output_lens.begin() + offset,
                           [](auto a, auto b) { return std::max(a, b); });
Khalique's avatar
Khalique committed
189
190
191
192
193
194
195
196
197

            auto l0 = prog.add_instruction(op::multibroadcast{output_lens}, arg0);
            auto l1 = prog.add_instruction(op::multibroadcast{output_lens}, arg1);
            return prog.add_instruction(x, l0, l1);
        }
        else
        {
            return prog.add_instruction(x, {arg0, arg1});
        }
198
199
    }

Paul's avatar
Paul committed
200
    template <class T>
Paul's avatar
Paul committed
201
202
    void add_generic_op(std::string name, T x)
    {
Paul's avatar
Paul committed
203
        add_op(name, [this, x](attribute_map, std::vector<instruction_ref> args) {
Paul's avatar
Paul committed
204
205
206
207
            return prog.add_instruction(x, args);
        });
    }

Khalique's avatar
Khalique committed
208
    template <class T>
Khalique's avatar
Khalique committed
209
    void add_variadic_op(std::string name, T x)
Khalique's avatar
Khalique committed
210
    {
Paul's avatar
Paul committed
211
        add_op(name, [this, x](attribute_map, std::vector<instruction_ref> args) {
Khalique's avatar
Khalique committed
212
            return std::accumulate(std::next(args.begin()),
Khalique's avatar
Khalique committed
213
214
215
216
217
                                   args.end(),
                                   args.front(),
                                   [this, x](instruction_ref a, instruction_ref b) {
                                       return add_broadcastable_binary_op(a, b, x);
                                   });
Khalique's avatar
Khalique committed
218
        });
Khalique's avatar
Khalique committed
219
220
    }

Paul's avatar
Paul committed
221
    instruction_ref
Paul's avatar
Paul committed
222
    parse_softmax(const std::string&, const attribute_map&, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
223
224
    {
        auto dims = args.front()->get_shape().lens();
Scott Thornton's avatar
Scott Thornton committed
225
226
        auto r =
            prog.add_instruction(op::reshape{{long(dims[0]), long(dims[1]), 1, 1}}, args.front());
227
228
        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
229
230
    }

Paul's avatar
Paul committed
231
    instruction_ref
Paul's avatar
Paul committed
232
    parse_conv(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
233
    {
234
        op::convolution op;
235
        auto l0 = args[0];
Paul's avatar
Paul committed
236
237
        if(contains(attributes, "pads"))
        {
Scott Thornton's avatar
Scott Thornton committed
238
            if(contains(attributes, "auto_pad"))
239
            {
Paul's avatar
Paul committed
240
                MIGRAPHX_THROW("auto_pad and padding cannot be specified simultaneously");
241
            }
242
243
            std::vector<std::int64_t> padding;
            copy(attributes["pads"].ints(), std::back_inserter(padding));
Scott Thornton's avatar
Scott Thornton committed
244
            if(padding.size() != 4)
245
            {
Paul's avatar
Paul committed
246
                MIGRAPHX_THROW("padding should have 4 values");
247
            }
Scott Thornton's avatar
Scott Thornton committed
248
            if(padding[0] != padding[2] || padding[1] != padding[3])
249
            {
250
251
                // insert zeros for pad op (args[0] has 4 dims)
                padding = {0, 0, padding[0], padding[1], 0, 0, padding[2], padding[3]};
Khalique's avatar
Khalique committed
252
                l0      = prog.add_instruction(op::pad{padding}, l0);
253
            }
254
255
256
257
            else
            {
                op.padding[0] = padding[0];
                op.padding[1] = padding[1];
258
            }
Paul's avatar
Paul committed
259
        }
Paul's avatar
Paul committed
260
261
262
263
264
265
266
267
        if(contains(attributes, "strides"))
        {
            copy(attributes["strides"].ints(), op.stride.begin());
        }
        if(contains(attributes, "dilations"))
        {
            copy(attributes["dilations"].ints(), op.dilation.begin());
        }
Scott Thornton's avatar
Scott Thornton committed
268
        if(contains(attributes, "auto_pad"))
269
270
        {
            auto s = attributes["auto_pad"].s();
Scott Thornton's avatar
Scott Thornton committed
271
            if(contains(attributes, "pads") and to_upper(s) != "NOTSET")
272
            {
Paul's avatar
Paul committed
273
                MIGRAPHX_THROW("auto_pad and padding cannot be specified simultaneously");
274
275
            }

wsttiger's avatar
fixes  
wsttiger committed
276
            if(s.find("SAME") != std::string::npos)
277
            {
278
                op.padding_mode = op::padding_mode_t::same;
279
280
            }
        }
Khalique's avatar
Khalique committed
281
282
283
284
        if(contains(attributes, "group"))
        {
            op.group = parse_value(attributes.at("group")).at<int>();
        }
Paul's avatar
Paul committed
285
286
287
288
        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
289
            auto l2       = prog.add_instruction(op::broadcast{axis, l1->get_shape()}, args[2]);
290
            return prog.add_instruction(op::add{}, l1, l2);
Paul's avatar
Paul committed
291
        }
292
        return prog.add_instruction(op, l0, args[1]);
Paul's avatar
Paul committed
293
    }
Paul's avatar
Paul committed
294

Paul's avatar
Paul committed
295
296
297
    instruction_ref parse_pooling(const std::string& name,
                                  attribute_map attributes,
                                  std::vector<instruction_ref> args)
Paul's avatar
Paul committed
298
    {
Khalique's avatar
Khalique committed
299
        op::pooling op{ends_with(name, "MaxPool") ? "max" : "average"};
300
        auto l0 = args[0];
Khalique's avatar
Khalique committed
301
        if(starts_with(name, "Global"))
302
        {
Khalique's avatar
Khalique committed
303
304
            auto lens  = args.front()->get_shape().lens();
            op.lengths = {lens[2], lens[3]};
305
        }
Paul's avatar
Paul committed
306
307
        if(contains(attributes, "pads"))
        {
308
309
            std::vector<std::int64_t> padding;
            copy(attributes["pads"].ints(), std::back_inserter(padding));
Scott Thornton's avatar
Scott Thornton committed
310
            if(padding.size() != 4)
311
            {
Paul's avatar
Paul committed
312
                MIGRAPHX_THROW("padding should have 4 values");
313
            }
Scott Thornton's avatar
Scott Thornton committed
314
            if(padding[0] != padding[2] || padding[1] != padding[3])
315
            {
316
317
                // insert zeros for pad op (args[0] has 4 dims)
                padding = {0, 0, padding[0], padding[1], 0, 0, padding[2], padding[3]};
Khalique's avatar
Khalique committed
318
                l0      = prog.add_instruction(op::pad{padding}, l0);
319
320
321
322
323
            }
            else
            {
                op.padding[0] = padding[0];
                op.padding[1] = padding[1];
324
            }
Paul's avatar
Paul committed
325
326
327
328
329
330
331
332
333
        }
        if(contains(attributes, "strides"))
        {
            copy(attributes["strides"].ints(), op.stride.begin());
        }
        if(contains(attributes, "kernel_shape"))
        {
            copy(attributes["kernel_shape"].ints(), op.lengths.begin());
        }
Scott Thornton's avatar
Scott Thornton committed
334
        if(contains(attributes, "auto_pad"))
335
336
        {
            auto s = attributes["auto_pad"].s();
337
            if(s.find("SAME_UPPER") == std::string::npos)
338
            {
339
                MIGRAPHX_THROW("auto_pad only supports SAME_UPPER for pooling");
340
            }
341
            op.padding_mode = op::padding_mode_t::same;
342
343
        }

344
        return prog.add_instruction(op, l0);
Paul's avatar
Paul committed
345
346
    }

Paul's avatar
Paul committed
347
    instruction_ref
Paul's avatar
Paul committed
348
    parse_reshape(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
349
    {
350
        op::reshape op;
Paul's avatar
Paul committed
351
352
353
354
355
356
357
        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
358
            auto s = args[1]->eval();
Paul's avatar
Paul committed
359
            if(s.empty())
Paul's avatar
Paul committed
360
                MIGRAPHX_THROW("Dynamic shape is not supported.");
Paul's avatar
Paul committed
361
            s.visit([&](auto v) { copy(v, std::back_inserter(op.dims)); });
Paul's avatar
Paul committed
362
        }
Paul's avatar
Paul committed
363
364
365
        return prog.add_instruction(op, args[0]);
    }

Paul's avatar
Paul committed
366
    instruction_ref
Paul's avatar
Paul committed
367
    parse_flatten(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
368
    {
369
        uint64_t axis = 1;
Paul's avatar
Paul committed
370
371
372
373
        if(contains(attributes, "axis"))
        {
            axis = parse_value(attributes.at("axis")).at<int>();
        }
374
        return prog.add_instruction(op::flatten{axis}, args[0]);
Paul's avatar
Paul committed
375
376
    }

377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
    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
395
396
397
398
399
400
401
    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));
    }
402

403
404
405
    instruction_ref
    parse_gather(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
406
        int axis = 0;
407
408
409
410
        if(contains(attributes, "axis"))
        {
            axis = parse_value(attributes.at("axis")).at<int>();
        }
411
        op::gather op{axis};
412
413
414
        return prog.add_instruction(op, std::move(args));
    }

415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
    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
435
436
437
    instruction_ref parse_constant(const std::string&,
                                   attribute_map attributes,
                                   const std::vector<instruction_ref>&)
Paul's avatar
Paul committed
438
    {
Shucai Xiao's avatar
Shucai Xiao committed
439
        literal v     = parse_value(attributes.at("value"));
440
441
442
        auto dim_size = attributes.at("value").t().dims_size();
        // if dim_size is 0, it is a scalar
        if(dim_size == 0)
443
        {
444
            migraphx::shape scalar_shape{v.get_shape().type()};
445
446
447
            return prog.add_literal(migraphx::literal{scalar_shape, v.data()});
        }

Paul's avatar
Paul committed
448
449
        return prog.add_literal(v);
    }
Paul's avatar
Paul committed
450

Paul's avatar
Paul committed
451
    instruction_ref
Paul's avatar
Paul committed
452
    parse_gemm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Paul's avatar
Paul committed
453
454
    {
        float alpha = 1.0f;
Khalique's avatar
Khalique committed
455
        float beta  = 1.0f;
Paul's avatar
Paul committed
456
457
458
459
460
461
462
463
        bool transa = false;
        bool transb = false;
        if(contains(attributes, "alpha"))
        {
            alpha = parse_value(attributes.at("alpha")).at<float>();
        }
        if(contains(attributes, "beta"))
        {
464
            beta = parse_value(attributes.at("beta")).at<float>();
Paul's avatar
Paul committed
465
466
467
468
469
470
471
472
473
        }
        if(contains(attributes, "transA"))
        {
            transa = parse_value(attributes.at("transA")).at<bool>();
        }
        if(contains(attributes, "transB"))
        {
            transb = parse_value(attributes.at("transB")).at<bool>();
        }
474
475
476
477
478
479

        std::vector<int64_t> perm(args[0]->get_shape().lens().size());
        std::iota(perm.begin(), perm.end(), int64_t{0});
        // swap the last two elements
        std::swap(*perm.rbegin(), *(perm.rbegin() + 1));

480
481
        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
482
483
        if(args.size() == 3)
        {
Khalique's avatar
Khalique committed
484
            if(beta != 0.f)
485
            {
Khalique's avatar
Khalique committed
486
                auto l3 = prog.add_instruction(op::dot{alpha}, l1, l2);
Khalique's avatar
Khalique committed
487
                auto l4 = args[2];
Khalique's avatar
Khalique committed
488
                if(l4->get_shape().scalar()) // ignore args[2] (no C value added to alpha*A*B)
Khalique's avatar
Khalique committed
489
                    return l3;
Khalique's avatar
Khalique committed
490
                if(beta != 1.f)
Khalique's avatar
Khalique committed
491
492
                {
                    auto beta_val = prog.add_literal(beta);
Khalique's avatar
Khalique committed
493
494
                    auto l5 = prog.add_instruction(op::scalar{args[2]->get_shape()}, beta_val);
                    l4      = prog.add_instruction(op::mul{}, args[2], l5);
Khalique's avatar
Khalique committed
495
496
                }
                return add_broadcastable_binary_op(l3, l4, op::add{});
497
            }
Paul's avatar
Paul committed
498
        }
499
        return prog.add_instruction(op::dot{alpha, beta}, l1, l2);
Paul's avatar
Paul committed
500
501
    }

502
    instruction_ref
Paul's avatar
Paul committed
503
    parse_batchnorm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
504
    {
Scott Thornton's avatar
Scott Thornton committed
505
506
        float epsilon                                     = 1e-5f;
        float momentum                                    = 0.9f;
507
        op::batch_norm_inference::bn_infer_mode_t bn_mode = op::batch_norm_inference::spatial;
Scott Thornton's avatar
Scott Thornton committed
508
        bool is_test                                      = false;
509
510
511
512
513
514
        if(contains(attributes, "epsilon"))
        {
            epsilon = parse_value(attributes.at("epsilon")).at<float>();
        }
        if(contains(attributes, "momentum"))
        {
515
            momentum = parse_value(attributes.at("momentum")).at<float>();
516
517
518
        }
        if(contains(attributes, "is_test"))
        {
wsttiger's avatar
wsttiger committed
519
            is_test = parse_value(attributes.at("is_test")).at<uint64_t>() > 0;
520
521
522
        }
        if(contains(attributes, "spatial"))
        {
523
            bn_mode = (parse_value(attributes.at("spatial")).at<uint64_t>() > 0)
524
525
                          ? op::batch_norm_inference::spatial
                          : op::batch_norm_inference::per_activation;
526
        }
Paul's avatar
Paul committed
527
        (void)is_test;
Paul's avatar
Paul committed
528
        op::batch_norm_inference op{epsilon, momentum, bn_mode};
Paul's avatar
Paul committed
529
        return prog.add_instruction(op, std::move(args));
530
531
    }

532
533
534
535
    instruction_ref parse_leaky_relu(const std::string&,
                                     attribute_map attributes,
                                     std::vector<instruction_ref> args)
    {
Khalique's avatar
Khalique committed
536
        float alpha = 0.01; // default alpha val for leaky relu
537
538
539
540
541
542
543
544
        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
545
546
    instruction_ref
    parse_elu(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Khalique's avatar
Khalique committed
547
548
549
550
551
552
553
554
555
556
    {
        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
557
558
    instruction_ref
    parse_lrn(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Khalique's avatar
Khalique committed
559
560
    {
        float alpha = 0.0001;
Khalique's avatar
Khalique committed
561
562
563
        float beta  = 0.75;
        float bias  = 1.0;
        int size    = 1;
Khalique's avatar
Khalique committed
564
565
566
567
568
569
570
571
572
573
574
575
        if(contains(attributes, "alpha"))
            alpha = parse_value(attributes.at("alpha")).at<float>();
        if(contains(attributes, "beta"))
            beta = parse_value(attributes.at("beta")).at<float>();
        if(contains(attributes, "bias"))
            bias = parse_value(attributes.at("bias")).at<float>();
        if(contains(attributes, "size"))
            size = parse_value(attributes.at("size")).at<int>();
        op::lrn op{alpha, beta, bias, size};
        return prog.add_instruction(op, args.front());
    }

Khalique's avatar
Khalique committed
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
    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
593

Khalique's avatar
Khalique committed
594
595
        auto scale_val = prog.add_literal(scale);
        auto bias_vals = prog.add_literal(
Paul's avatar
Paul committed
596
            migraphx::literal{migraphx::shape{migraphx::shape::float_type, {bias.size()}}, bias});
Khalique's avatar
Khalique committed
597

Paul's avatar
Paul committed
598
599
        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
600
        auto bias_bcast = prog.add_instruction(migraphx::op::broadcast{1, input_shape}, bias_vals);
Paul's avatar
Paul committed
601
        return prog.add_instruction(migraphx::op::add{}, img_scaled, bias_bcast);
Khalique's avatar
Khalique committed
602
    }
Khalique's avatar
Khalique committed
603

Khalique's avatar
Khalique committed
604
605
    instruction_ref
    parse_transpose(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
Khalique's avatar
Khalique committed
606
607
608
609
610
611
612
    {
        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
613
        return prog.add_instruction(migraphx::op::transpose{perm}, args.front());
Khalique's avatar
Khalique committed
614
615
    }

Khalique's avatar
Khalique committed
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
    instruction_ref
    parse_pad(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        std::vector<int64_t> pads{};
        float value = 0.0f;
        if(contains(attributes, "pads"))
        {
            auto&& pad_vals = attributes["pads"].ints();
            pads            = std::vector<int64_t>(pad_vals.begin(), pad_vals.end());
        }
        if(contains(attributes, "value"))
        {
            value = parse_value(attributes.at("value")).at<float>();
        }
        if(contains(attributes, "mode"))
        {
            auto mode = attributes.at("mode").s();
            if(mode != "constant")
                MIGRAPHX_THROW("migraphx currently only supports constant padding");
        }
        return prog.add_instruction(migraphx::op::pad{pads, value}, args.front());
    }
638
639
640
    // Use a literal instruction to replace the shape since, output of
    // shape operator are literals in migraphx
    instruction_ref
Shucai Xiao's avatar
Shucai Xiao committed
641
    parse_shape(const std::string&, const attribute_map&, std::vector<instruction_ref> args)
642
643
    {
        if(args.size() != 1)
644
            MIGRAPHX_THROW("Shape: operator should have 1 operand");
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
        std::vector<std::size_t> arg_shape = args[0]->get_shape().lens();
        std::vector<int64_t> vec_shape(arg_shape.size());
        migraphx::shape s(migraphx::shape::int64_type, {arg_shape.size()});
        std::transform(arg_shape.begin(), arg_shape.end(), vec_shape.begin(), [](auto i) {
            return int64_t(i);
        });
        return prog.add_literal(migraphx::literal{s, vec_shape});
    }

    // Use a literal instruction to replace the constantFill operator. In RNN, input shape
    // and value are fixed, so no need to do the actual computation for the constantFill
    // operator
    instruction_ref parse_constant_fill(const std::string&,
                                        attribute_map attributes,
                                        std::vector<instruction_ref> args)
    {
        int input_as_shape = 0;
        int dtype          = 1;
        float value        = 0.0f;

        if(contains(attributes, "dtype"))
        {
            dtype = parse_value(attributes.at("dtype")).at<int>();
        }
        migraphx::shape::type_t type = get_type(dtype);

        if(contains(attributes, "input_as_shape"))
        {
            input_as_shape = parse_value(attributes.at("input_as_shape")).at<int>();
        }

        if(contains(attributes, "value"))
        {
            value = parse_value(attributes.at("value")).at<float>();
        }

Shucai Xiao's avatar
Shucai Xiao committed
681
682
        if(contains(attributes, "extra_shape"))
        {
683
            MIGRAPHX_THROW("ConstantFill: cannot handle extra shape attribute");
684
685
        }

686
687
        if(input_as_shape == 1)
        {
Shucai Xiao's avatar
Shucai Xiao committed
688
            if(args.size() != 1)
689
            {
690
                MIGRAPHX_THROW("ConstantFill: need an input argument as output shape");
691
692
            }

Shucai Xiao's avatar
Shucai Xiao committed
693
694
            if(contains(attributes, "shape"))
            {
695
                MIGRAPHX_THROW("ConstantFill: cannot set the shape argument and pass in an input "
Shucai Xiao's avatar
Shucai Xiao committed
696
                               "at the same time");
697
698
            }

699
700
701
            migraphx::argument in = args[0]->eval();
            if(in.empty())
            {
702
                MIGRAPHX_THROW("ConstantFill: cannot handle dynamic shape as input");
703
            }
704

705
706
707
            std::vector<std::size_t> dims;
            in.visit([&](auto input) { dims.assign(input.begin(), input.end()); });
            migraphx::shape s(type, dims);
708
709
            std::vector<float> values(s.elements(), value);
            return prog.add_literal(migraphx::literal(s, values));
710
711
712
        }
        else if(input_as_shape == 0)
        {
Shucai Xiao's avatar
Shucai Xiao committed
713
714
            if(!contains(attributes, "shape"))
            {
715
                MIGRAPHX_THROW("ConstantFill: attribute output shape is needed");
716
717
718
            }

            literal ls = parse_value(attributes.at("shape"));
719
            std::vector<std::size_t> dims;
Shucai Xiao's avatar
Shucai Xiao committed
720
            ls.visit([&](auto s) { dims.assign(s.begin(), s.end()); });
721
            migraphx::shape s{type, dims};
722
723
            std::vector<float> values(s.elements(), value);
            return prog.add_literal(migraphx::literal(s, values));
724
725
726
        }
        else
        {
727
            MIGRAPHX_THROW("ConstantFill: wrong value of attribute input_as_shape");
728
729
730
        }
    }

Shucai Xiao's avatar
Shucai Xiao committed
731
    std::vector<instruction_ref>
Shucai Xiao's avatar
Shucai Xiao committed
732
733
734
    parse_rnn(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        migraphx::shape input_shape = args[0]->get_shape();
735
        std::size_t hidden_size     = args[1]->get_shape().lens()[1];
Shucai Xiao's avatar
Shucai Xiao committed
736
737
738

        if(contains(attributes, "hidden_size"))
        {
Shucai Xiao's avatar
Shucai Xiao committed
739
            std::size_t hidden_size_att = parse_value(attributes.at("hidden_size")).at<int>();
Shucai Xiao's avatar
Shucai Xiao committed
740
            if(hidden_size != hidden_size_att)
Shucai Xiao's avatar
Shucai Xiao committed
741
742
743
            {
                MIGRAPHX_THROW("RNN: hidden size mismatch in input and attribute");
            }
Shucai Xiao's avatar
Shucai Xiao committed
744
745
746
747
748
749
750
751
752
        }

        // Handling of direction to be added later
        std::string direction{"forward"};
        if(contains(attributes, "direction"))
        {
            direction = attributes.at("direction").s();
        }

753
        op::rnn_direction dirct = op::rnn_direction::forward;
Shucai Xiao's avatar
Shucai Xiao committed
754
755
        if(direction == "bidirectional")
        {
756
            dirct = op::rnn_direction::bidirectional;
Shucai Xiao's avatar
Shucai Xiao committed
757
758
759
        }
        else if(direction == "reverse")
        {
760
            dirct = op::rnn_direction::reverse;
Shucai Xiao's avatar
Shucai Xiao committed
761
762
        }

763
764
765
766
767
        std::vector<std::string> vec_names{"tanh"};
        if(contains(attributes, "activations"))
        {
            auto names = attributes.at("activations").strings();
            vec_names.clear();
768
            vec_names.resize(names.size());
769
            std::copy(names.begin(), names.end(), vec_names.begin());
770
771
        }

772
773
774
        auto name_it = std::find_if(vec_names.begin(), vec_names.end(), [&](auto& name) {
            return (map_actv_funcs.count(name) == 0);
        });
Shucai Xiao's avatar
Shucai Xiao committed
775
        if(name_it != vec_names.end())
776
777
778
        {
            MIGRAPHX_THROW("RNN: activation function " + std::string(*name_it) + " not supported");
        }
779

Shucai Xiao's avatar
Shucai Xiao committed
780
        // bidirectional case should have two activation functions.
Shucai Xiao's avatar
Shucai Xiao committed
781
        // one is for forward, and the other is for reverse.
Shucai Xiao's avatar
Shucai Xiao committed
782
        // if only one actv function is provided, we use it in both
783
        // forward and reverse direction
784
        if(dirct == op::rnn_direction::bidirectional)
785
        {
Shucai Xiao's avatar
Shucai Xiao committed
786
            if(vec_names.size() == 1)
787
788
789
790
791
            {
                vec_names.push_back(vec_names.at(0));
            }
        }

Shucai Xiao's avatar
Shucai Xiao committed
792
793
794
        std::vector<operation> vec_actv_funcs(vec_names.size());
        std::transform(vec_names.begin(), vec_names.end(), vec_actv_funcs.begin(), [&](auto& fn) {
            return map_actv_funcs[fn];
795
        });
Shucai Xiao's avatar
Shucai Xiao committed
796

Shucai Xiao's avatar
Shucai Xiao committed
797
798
799
800
801
802
803
        // To be added later
        float clip = 0.0;
        if(contains(attributes, "clip"))
        {
            clip = parse_value(attributes.at("clip")).at<float>();
        }

804
805
        // if the number of arguments is less than 6, append
        // undefined operator to have 6 arguments
Shucai Xiao's avatar
Shucai Xiao committed
806
        if(args.size() < 6)
807
808
809
810
811
        {
            auto ins = prog.add_instruction(op::undefined{});
            args.insert(args.end(), (6 - args.size()), ins);
        }

Shucai Xiao's avatar
Shucai Xiao committed
812
813
        // first output for the concatenation of hidden states
        auto hidden_states = prog.add_instruction(op::rnn{hidden_size, vec_actv_funcs, dirct, clip},
Shucai Xiao's avatar
Shucai Xiao committed
814
                                                  std::move(args));
Shucai Xiao's avatar
Shucai Xiao committed
815

816
        // second output for the last hidden state
Shucai Xiao's avatar
Shucai Xiao committed
817
        auto last_output = prog.add_instruction(op::rnn_last_output{}, hidden_states);
Shucai Xiao's avatar
Shucai Xiao committed
818

Shucai Xiao's avatar
Shucai Xiao committed
819
        return {hidden_states, last_output};
Shucai Xiao's avatar
Shucai Xiao committed
820
821
    }

822
    std::vector<instruction_ref>
823
824
825
826
827
828
829
    parse_gru(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        migraphx::shape input_shape = args[0]->get_shape();
        std::size_t hidden_size     = args[2]->get_shape().lens()[2];

        if(contains(attributes, "hidden_size"))
        {
Shucai Xiao's avatar
Shucai Xiao committed
830
            std::size_t hidden_size_att = parse_value(attributes.at("hidden_size")).at<int>();
Shucai Xiao's avatar
Shucai Xiao committed
831
            if(hidden_size != hidden_size_att)
Shucai Xiao's avatar
Shucai Xiao committed
832
833
834
            {
                MIGRAPHX_THROW("GRU: hidden size mismatch in input and attribute");
            }
835
836
837
838
839
840
841
842
843
        }

        // Handling of direction to be added later
        std::string direction{"forward"};
        if(contains(attributes, "direction"))
        {
            direction = attributes.at("direction").s();
        }

844
        op::rnn_direction dirct = op::rnn_direction::forward;
845
846
        if(direction == "bidirectional")
        {
847
            dirct = op::rnn_direction::bidirectional;
848
849
850
        }
        else if(direction == "reverse")
        {
851
            dirct = op::rnn_direction::reverse;
852
853
        }

854
        std::vector<std::string> vec_names = {"sigmoid", "tanh"};
855
856
        if(contains(attributes, "activations"))
        {
857
            auto names = attributes.at("activations").strings();
858
            vec_names.clear();
Shucai Xiao's avatar
Shucai Xiao committed
859
            vec_names.resize(names.size());
860
            std::copy(names.begin(), names.end(), vec_names.begin());
861
862
        }

863
        // need 4 activation functions
864
        if(dirct == op::rnn_direction::bidirectional)
865
        {
Shucai Xiao's avatar
Shucai Xiao committed
866
            // 4 activation functions are used in the bidirectional
867
            // scenario. No spec is provided in onnx::operator. we
Shucai Xiao's avatar
Shucai Xiao committed
868
869
            // use the algorithm that: if 1 actv function is provided,
            // repeat 1 four times. If 2 actv functins are provided,
870
871
            // assume forward and reverse use the same pair of actv
            // functions. For the case of 3 actv functions provided,
Shucai Xiao's avatar
Shucai Xiao committed
872
873
874
            // assume the 3rd one is repeated once and used by the
            // reverse direction.
            // This may need change later
875
            if(vec_names.size() == 1)
876
            {
877
                vec_names.insert(vec_names.end(), 3, vec_names.at(0));
878
            }
879
            else if(vec_names.size() == 2)
880
            {
881
882
883
                // repeat the activation functions
                vec_names.push_back(vec_names.at(0));
                vec_names.push_back(vec_names.at(1));
884
            }
885
            else if(vec_names.size() == 3)
886
            {
887
                vec_names.push_back(vec_names.at(2));
888
889
            }
        }
Shucai Xiao's avatar
Shucai Xiao committed
890
        else
891
        {
892
            if(vec_names.size() == 1)
893
            {
894
                vec_names.push_back(vec_names.at(0));
895
896
897
            }
        }

898
899
900
        auto name_it = std::find_if(vec_names.begin(), vec_names.end(), [&](auto& name) {
            return (map_actv_funcs.count(name) == 0);
        });
Shucai Xiao's avatar
Shucai Xiao committed
901
        if(name_it != vec_names.end())
902
903
904
        {
            MIGRAPHX_THROW("GRU: activation function " + std::string(*name_it) + " not supported");
        }
905

Shucai Xiao's avatar
Shucai Xiao committed
906
907
908
        std::vector<operation> vec_actv_funcs(vec_names.size());
        std::transform(vec_names.begin(), vec_names.end(), vec_actv_funcs.begin(), [&](auto& name) {
            return map_actv_funcs[name];
Shucai Xiao's avatar
Shucai Xiao committed
909
        });
910
911
912
913
914
915
916
917

        float clip = 0.0;
        if(contains(attributes, "clip"))
        {
            clip = parse_value(attributes.at("clip")).at<float>();
        }

        int linear_before_reset = 0;
Shucai Xiao's avatar
Shucai Xiao committed
918
        if(contains(attributes, "linear_before_reset"))
919
920
921
922
        {
            linear_before_reset = parse_value(attributes.at("linear_before_reset")).at<int>();
        }

Shucai Xiao's avatar
Shucai Xiao committed
923
        // append undefined opeator to make 6 arguments
Shucai Xiao's avatar
Shucai Xiao committed
924
        if(args.size() < 6)
Shucai Xiao's avatar
Shucai Xiao committed
925
926
927
928
929
        {
            auto ins = prog.add_instruction(op::undefined{});
            args.insert(args.end(), 6 - args.size(), ins);
        }

930
931
        // first output for concatenation of hidden states
        auto hidden_states = prog.add_instruction(
Shucai Xiao's avatar
Shucai Xiao committed
932
            op::gru{hidden_size, vec_actv_funcs, dirct, clip, linear_before_reset},
Shucai Xiao's avatar
Shucai Xiao committed
933
            std::move(args));
934
935

        // second output for last gru output
936
        auto last_output = prog.add_instruction(op::rnn_last_output{}, hidden_states);
937

Shucai Xiao's avatar
Shucai Xiao committed
938
        return {hidden_states, last_output};
939
940
    }

Shucai Xiao's avatar
Shucai Xiao committed
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
    std::vector<instruction_ref>
    parse_lstm(const std::string&, attribute_map attributes, std::vector<instruction_ref> args)
    {
        migraphx::shape input_shape = args[0]->get_shape();
        std::size_t hidden_size     = args[2]->get_shape().lens()[2];

        if(contains(attributes, "hidden_size"))
        {
            std::size_t hidden_size_att = parse_value(attributes.at("hidden_size")).at<int>();
            if(hidden_size != hidden_size_att)
            {
                MIGRAPHX_THROW("LSTM: hidden size mismatch in input and attribute");
            }
        }

        // Handling of direction to be added later
        std::string direction{"forward"};
        if(contains(attributes, "direction"))
        {
            direction = attributes.at("direction").s();
        }

Shucai Xiao's avatar
Shucai Xiao committed
963
        op::rnn_direction dirct = op::rnn_direction::forward;
Shucai Xiao's avatar
Shucai Xiao committed
964
965
        if(direction == "bidirectional")
        {
Shucai Xiao's avatar
Shucai Xiao committed
966
            dirct = op::rnn_direction::bidirectional;
Shucai Xiao's avatar
Shucai Xiao committed
967
968
969
        }
        else if(direction == "reverse")
        {
Shucai Xiao's avatar
Shucai Xiao committed
970
            dirct = op::rnn_direction::reverse;
Shucai Xiao's avatar
Shucai Xiao committed
971
        }
Shucai Xiao's avatar
Shucai Xiao committed
972
        else if(direction == "forward")
Shucai Xiao's avatar
Shucai Xiao committed
973
        {
Shucai Xiao's avatar
Shucai Xiao committed
974
            dirct = op::rnn_direction::forward;
Shucai Xiao's avatar
Shucai Xiao committed
975
976
977
978
979
980
981
982
983
984
985
986
        }
        else
        {
            MIGRAPHX_THROW("LSTM: incorrect direction attribute");
        }

        std::vector<std::string> vec_names = {"sigmoid", "tanh", "tanh"};
        if(contains(attributes, "activations"))
        {
            auto names = attributes.at("activations").strings();
            vec_names.clear();
            vec_names.resize(names.size());
987
            std::copy(names.begin(), names.end(), vec_names.begin());
Shucai Xiao's avatar
Shucai Xiao committed
988
989
990
        }

        // need 6 activation functions for bidirectional directions
Shucai Xiao's avatar
Shucai Xiao committed
991
        if(dirct == op::rnn_direction::bidirectional)
Shucai Xiao's avatar
Shucai Xiao committed
992
993
994
995
996
997
        {
            // 6 activation functions are used in the bidirectional
            // scenario. No spec is provided in onnx::operator. we
            // use the algorithm that: if 1 actv function is provided,
            // repeat 1st six times. If 2 actv functins are provided,
            // repeat 2nd once, then repeat all three once
Shucai Xiao's avatar
Shucai Xiao committed
998
            // if 3 actv funcs are provide, repeat all three once.
Shucai Xiao's avatar
Shucai Xiao committed
999
1000
1001
1002
            // the same algorithm is used for 4, 5, and 6 actv funcions
            // provided. This may need change later
            switch(vec_names.size())
            {
1003
            case 1:
Shucai Xiao's avatar
Shucai Xiao committed
1004
1005
1006
1007
1008
1009
                vec_names = {vec_names.at(0),
                             vec_names.at(0),
                             vec_names.at(0),
                             vec_names.at(0),
                             vec_names.at(0),
                             vec_names.at(0)};
1010
                break;
Shucai Xiao's avatar
Shucai Xiao committed
1011
1012
1013

            case 2:
                // repeat the 2nd actv func once, then repeat all three another time
Shucai Xiao's avatar
Shucai Xiao committed
1014
1015
1016
1017
1018
1019
                vec_names = {vec_names.at(0),
                             vec_names.at(1),
                             vec_names.at(1),
                             vec_names.at(0),
                             vec_names.at(1),
                             vec_names.at(1)};
Shucai Xiao's avatar
Shucai Xiao committed
1020
1021
1022
1023
                break;

            case 3:
                // repeat all three actv funcs once
Shucai Xiao's avatar
Shucai Xiao committed
1024
1025
1026
1027
1028
1029
                vec_names = {vec_names.at(0),
                             vec_names.at(1),
                             vec_names.at(2),
                             vec_names.at(0),
                             vec_names.at(1),
                             vec_names.at(2)};
Shucai Xiao's avatar
Shucai Xiao committed
1030
1031
                break;

Shucai Xiao's avatar
Shucai Xiao committed
1032
1033
1034
1035
1036
1037
1038
            case 4:
                vec_names = {vec_names.at(0),
                             vec_names.at(1),
                             vec_names.at(2),
                             vec_names.at(3),
                             vec_names.at(3),
                             vec_names.at(3)};
1039
                break;
Shucai Xiao's avatar
Shucai Xiao committed
1040

Shucai Xiao's avatar
Shucai Xiao committed
1041
1042
1043
1044
1045
1046
1047
            case 5:
                vec_names = {vec_names.at(0),
                             vec_names.at(1),
                             vec_names.at(2),
                             vec_names.at(3),
                             vec_names.at(4),
                             vec_names.at(4)};
1048
                break;
Shucai Xiao's avatar
Shucai Xiao committed
1049

Shucai Xiao's avatar
Shucai Xiao committed
1050
            default: break;
Shucai Xiao's avatar
Shucai Xiao committed
1051
1052
1053
1054
1055
1056
            }
        }
        else
        {
            switch(vec_names.size())
            {
Shucai Xiao's avatar
Shucai Xiao committed
1057
            case 1: vec_names = {vec_names.at(0), vec_names.at(0), vec_names.at(0)}; break;
Shucai Xiao's avatar
Shucai Xiao committed
1058
1059
1060

            case 2:
                // repeat the 2nd actv func once, so we have 3 actv funcs
Shucai Xiao's avatar
Shucai Xiao committed
1061
                vec_names = {vec_names.at(0), vec_names.at(1), vec_names.at(1)};
Shucai Xiao's avatar
Shucai Xiao committed
1062
1063
                break;

Shucai Xiao's avatar
Shucai Xiao committed
1064
            default: break;
Shucai Xiao's avatar
Shucai Xiao committed
1065
1066
1067
            }
        }

1068
1069
1070
        auto name_it = std::find_if(vec_names.begin(), vec_names.end(), [&](auto& name) {
            return (map_actv_funcs.count(name) == 0);
        });
Shucai Xiao's avatar
Shucai Xiao committed
1071
        if(name_it != vec_names.end())
1072
1073
1074
        {
            MIGRAPHX_THROW("LSTM: activation function " + std::string(*name_it) + " not supported");
        }
Shucai Xiao's avatar
Shucai Xiao committed
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096

        std::vector<operation> vec_actv_funcs(vec_names.size());
        std::transform(vec_names.begin(), vec_names.end(), vec_actv_funcs.begin(), [&](auto& name) {
            return map_actv_funcs[name];
        });

        float clip = 0.0;
        if(contains(attributes, "clip"))
        {
            clip = parse_value(attributes.at("clip")).at<float>();
        }

        int input_forget = 0;
        if(contains(attributes, "input_forget"))
        {
            input_forget = parse_value(attributes.at("input_forget")).at<int>();
        }

        // append undefined opeator to make 6 arguments
        if(args.size() < 8)
        {
            auto ins = prog.add_instruction(op::undefined{});
Shucai Xiao's avatar
Shucai Xiao committed
1097
            args.insert(args.end(), 8 - args.size(), ins);
Shucai Xiao's avatar
Shucai Xiao committed
1098
1099
1100
1101
        }

        // first output for concatenation of hidden states
        auto hidden_states = prog.add_instruction(
Shucai Xiao's avatar
Shucai Xiao committed
1102
            op::lstm{hidden_size, vec_actv_funcs, dirct, clip, input_forget}, std::move(args));
Shucai Xiao's avatar
Shucai Xiao committed
1103
1104

        // second output for last lstm output
Shucai Xiao's avatar
Shucai Xiao committed
1105
        auto last_output = prog.add_instruction(op::rnn_last_output{}, hidden_states);
Shucai Xiao's avatar
Shucai Xiao committed
1106
1107
1108
1109
1110
1111
1112

        // third output for last cell output
        auto last_cell_output = prog.add_instruction(op::lstm_last_cell_output{}, hidden_states);

        return {hidden_states, last_output, last_cell_output};
    }

Paul's avatar
Paul committed
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
    void parse_from(std::istream& is)
    {
        onnx::ModelProto model;
        if(model.ParseFromIstream(&is))
        {
            if(model.has_graph())
            {
                this->parse_graph(model.graph());
            }
        }
        else
        {
Paul's avatar
Paul committed
1125
            MIGRAPHX_THROW("Failed reading onnx file.");
Paul's avatar
Paul committed
1126
1127
1128
1129
1130
1131
        }
    }

    void parse_graph(const onnx::GraphProto& graph)
    {
        nodes = get_nodes(graph);
1132
1133
1134
1135
1136
        std::unordered_map<std::string, onnx::TensorProto> initializer_data;
        for(auto&& f : graph.initializer())
        {
            initializer_data[f.name()] = f;
        }
Paul's avatar
Paul committed
1137
1138
1139
        for(auto&& input : graph.input())
        {
            const std::string& name = input.name();
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
            // 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
1152
1153
1154
        }
        for(auto&& p : nodes)
        {
Paul's avatar
Paul committed
1155
            this->parse_node(p.first);
Paul's avatar
Paul committed
1156
1157
1158
        }
    }

Shucai Xiao's avatar
Shucai Xiao committed
1159
    void parse_undefined(const std::string& name)
1160
    {
Shucai Xiao's avatar
Shucai Xiao committed
1161
        auto ins           = prog.add_instruction(op::undefined{});
1162
1163
1164
        instructions[name] = ins;
    }

Paul's avatar
Paul committed
1165
    void parse_node(const std::string& name)
Paul's avatar
Paul committed
1166
    {
Paul's avatar
Paul committed
1167
        if(name.empty())
Paul's avatar
Paul committed
1168
            MIGRAPHX_THROW("Onnx node must have a name");
Paul's avatar
Paul committed
1169
1170
1171
1172
1173
1174
1175
1176
        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)
                {
Paul's avatar
Paul committed
1177
1178
                    assert(name != input);
                    this->parse_node(input);
Paul's avatar
Paul committed
1179
                }
Shucai Xiao's avatar
Shucai Xiao committed
1180
                else if(input.empty())
Paul's avatar
Paul committed
1181
                {
1182
                    this->parse_undefined(input);
Paul's avatar
Paul committed
1183
                }
1184
                args.push_back(instructions.at(input));
Paul's avatar
Paul committed
1185
            }
Paul's avatar
Paul committed
1186
            std::vector<instruction_ref> result;
Paul's avatar
Paul committed
1187
1188
            if(ops.count(node.op_type()) == 0)
            {
Paul's avatar
Paul committed
1189
                result.push_back(prog.add_instruction(unknown{node.op_type()}, args));
Paul's avatar
Paul committed
1190
1191
1192
            }
            else
            {
Paul's avatar
Paul committed
1193
                result = ops[node.op_type()](get_attributes(node), args);
Paul's avatar
Paul committed
1194
            }
Paul's avatar
Paul committed
1195
            // Even no output nodes produce output in migraphx
Paul's avatar
Paul committed
1196
            if(node.output().empty() and result.size() == 1)
Paul's avatar
Paul committed
1197
1198
            {
                instructions[name] = result.front();
Paul's avatar
Paul committed
1199
1200
1201
            }
            else
            {
Paul's avatar
Paul committed
1202
1203
1204
1205
1206
1207
                assert(node.output().size() >= result.size());
                std::transform(result.begin(),
                               result.end(),
                               node.output().begin(),
                               std::inserter(instructions, instructions.end()),
                               [](auto&& x, auto&& y) { return std::make_pair(y, x); });
Paul's avatar
Paul committed
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
            }
        }
    }

    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;
    }

    static node_map get_nodes(const onnx::GraphProto& graph)
    {
        std::unordered_map<std::string, onnx::NodeProto> result;
Paul's avatar
Paul committed
1225
        std::size_t n = 0;
Paul's avatar
Paul committed
1226
1227
        for(auto&& node : graph.node())
        {
Paul's avatar
Paul committed
1228
            if(node.output().empty())
Paul's avatar
Paul committed
1229
            {
Paul's avatar
Paul committed
1230
                if(node.name().empty())
Paul's avatar
Paul committed
1231
1232
1233
1234
1235
1236
1237
1238
1239
                {
                    result["migraphx_unamed_node_" + std::to_string(n)] = node;
                    n++;
                }
                else
                {
                    result[node.name()] = node;
                }
            }
Paul's avatar
Paul committed
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
            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
1265
        case onnx::AttributeProto::FLOATS: return from_repeated(shape::float_type, attr.floats());
Paul's avatar
Paul committed
1266
1267
1268
1269
1270
        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
1271
        MIGRAPHX_THROW("Invalid attribute type");
Paul's avatar
Paul committed
1272
1273
1274
1275
1276
    }

    static literal parse_tensor(const onnx::TensorProto& t)
    {
        std::vector<std::size_t> dims(t.dims().begin(), t.dims().end());
Khalique's avatar
Khalique committed
1277
        // in case of scalar constants in onnx file, use dims=1 to fill initializer data
1278
        if(dims.empty())
Khalique's avatar
Khalique committed
1279
1280
1281
        {
            dims = {1};
        }
1282
1283
        if(t.has_raw_data())
        {
wsttiger's avatar
wsttiger committed
1284
            const std::string& s = t.raw_data();
Scott Thornton's avatar
Scott Thornton committed
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
            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
1297
            case onnx::TensorProto::FLOAT16: return literal{{shape::half_type, dims}, s.data()};
Scott Thornton's avatar
Scott Thornton committed
1298
1299
1300
1301
1302
1303
            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("");
            }
Paul's avatar
Paul committed
1304
            MIGRAPHX_THROW("Invalid tensor type");
1305
        }
Paul's avatar
Paul committed
1306
1307
1308
1309
        switch(t.data_type())
        {
        case onnx::TensorProto::UNDEFINED: throw std::runtime_error("");
        case onnx::TensorProto::FLOAT:
Paul's avatar
Paul committed
1310
            return literal{{shape::float_type, dims}, t.float_data().begin(), t.float_data().end()};
Paul's avatar
Paul committed
1311
1312
        case onnx::TensorProto::UINT8: throw std::runtime_error("");
        case onnx::TensorProto::INT8:
Paul's avatar
Paul committed
1313
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
1314
        case onnx::TensorProto::UINT16:
Paul's avatar
Paul committed
1315
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
1316
        case onnx::TensorProto::INT16:
Paul's avatar
Paul committed
1317
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
1318
        case onnx::TensorProto::INT32:
Paul's avatar
Paul committed
1319
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
1320
        case onnx::TensorProto::INT64:
Paul's avatar
Paul committed
1321
            return literal{{shape::int64_type, dims}, t.int64_data().begin(), t.int64_data().end()};
Paul's avatar
Paul committed
1322
1323
        case onnx::TensorProto::STRING: throw std::runtime_error("");
        case onnx::TensorProto::BOOL:
Paul's avatar
Paul committed
1324
            return literal{{shape::int32_type, dims}, t.int32_data().begin(), t.int32_data().end()};
Paul's avatar
Paul committed
1325
        case onnx::TensorProto::FLOAT16:
Khalique's avatar
Khalique committed
1326
        {
Khalique's avatar
Khalique committed
1327
            std::vector<uint16_t> data_uint16(t.int32_data().begin(), t.int32_data().end());
1328
            std::vector<half> data_half;
Khalique's avatar
Khalique committed
1329
1330
1331
            std::transform(data_uint16.begin(),
                           data_uint16.end(),
                           std::back_inserter(data_half),
1332
                           [](uint16_t raw_val) { return *reinterpret_cast<half*>(&raw_val); });
1333
            return literal{{shape::half_type, dims}, data_half.begin(), data_half.end()};
Khalique's avatar
Khalique committed
1334
        }
Paul's avatar
Paul committed
1335
1336
1337
1338
1339
1340
1341
1342
        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
1343
        MIGRAPHX_THROW("Invalid tensor type");
Paul's avatar
Paul committed
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
    }

    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
1365
        case onnx::TensorProto::FLOAT16: shape_type = shape::half_type; break;
Paul's avatar
Paul committed
1366
1367
1368
1369
1370
1371
1372
1373
1374
        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
1375
        auto&& tensor_dims = t.tensor_type().shape().dim();
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
        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
1387
1388
        return {shape_type, dims};
    }
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410

    shape::type_t get_type(int dtype)
    {
        switch(dtype)
        {
        case 1: return shape::float_type;
        case 2: return shape::uint8_type;
        case 3: return shape::int8_type;
        case 4: return shape::uint16_type;
        case 5: return shape::int16_type;
        case 6: return shape::int32_type;
        case 7: return shape::int64_type;
        case 10: return shape::half_type;
        case 11: return shape::double_type;
        case 12: return shape::uint32_type;
        case 13: return shape::uint64_type;
        default:
        {
            MIGRAPHX_THROW("Prototensor data type " + std::to_string(dtype) + " not supported");
        }
        }
    }
Paul's avatar
Paul committed
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
};

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
1434
} // namespace MIGRAPHX_INLINE_NS
Paul's avatar
Paul committed
1435
} // namespace migraphx