1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
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
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
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
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
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
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
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
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
|
# SPDX-FileCopyrightText: Copyright 2020-2023 Arm Limited and/or its affiliates <open-source-office@arm.com>
#
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the License); you may
# not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an AS IS BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# Description:
# Internal representation of a Neural Network Tensor.
import copy
import enum
import uuid
from collections import defaultdict
from enum import auto
from functools import lru_cache
from functools import total_ordering
from typing import Dict
from typing import List
from typing import Optional
from typing import Tuple
from typing import Union
from uuid import UUID
import numpy as np
from . import numeric_util
from .data_type import BaseType
from .data_type import DataType
from .errors import UnsupportedFeatureError
from .errors import VelaError
from .numeric_util import full_shape
from .operation import Op
from .operation import Operation
from .shape4d import Shape4D
Shape = List
class MemType(enum.IntFlag):
Unknown = 0
Permanent_NPU = 1
Permanent_CPU = 2
Scratch = 3
Scratch_fast = 4
Size = Scratch_fast + 1
def display_name(self) -> str:
return ("Unknown", "Permanent_NPU", "Permanent_CPU", "Scratch", "Scratch_fast", "Size")[self.value]
def identifier_name(self) -> str:
return ("unknown", "permanent_npu", "permanent_cpu", "scratch", "scratch_fast", "size")[self.value]
@staticmethod
def all():
return (MemType.Permanent_NPU, MemType.Permanent_CPU, MemType.Scratch, MemType.Scratch_fast)
def __str__(self):
return self.name
class BandwidthDirection(enum.IntEnum):
Read = 0
Write = auto()
Size = auto()
def display_name(self):
return self.name
def identifier_name(self):
return self.name.lower()
@staticmethod
def all():
return (BandwidthDirection.Read, BandwidthDirection.Write)
class MemArea(enum.IntFlag):
Unknown = 0
Sram = 1
Dram = 2
OnChipFlash = 3
OffChipFlash = 4
Shram = 5 # for LUT
Size = Shram + 1
def display_name(self) -> str:
return ("Unknown", "SRAM", "DRAM", "On-chip Flash", "Off-chip Flash", "SHRAM", "Size")[self.value]
def identifier_name(self) -> str:
return ("unknown", "sram", "dram", "on_chip_flash", "off_chip_flash", "shram", "size")[self.value]
@staticmethod
def all():
return (MemArea.Sram, MemArea.Dram, MemArea.OnChipFlash, MemArea.OffChipFlash, MemArea.Shram)
def __str__(self):
return self.name
class TensorPurpose(enum.IntFlag):
Unknown = 0
Weights = 1
FeatureMap = 2
Scratch = 3
ScratchFast = 4
LUT = 5
FSBias = 6
Virtual = 7
Size = 8
def display_name(self) -> str:
return ("Unknown", "Weights", "FeatureMap", "Scratch", "ScratchFast", "LUT", "FastStorageBias", "Size")[
self.value
]
def identifier_name(self) -> str:
return ("unknown", "weights", "feature_map", "scratch", "scratch_fast", "lut", "fast_storage_bias", "size")[
self.value
]
@staticmethod
def all():
return (TensorPurpose.Weights, TensorPurpose.FeatureMap, TensorPurpose.FSBias)
class TensorSubPurpose(enum.Enum):
Standard = 0
DoubleBuffer = 1
RollingBufferX = 2
RollingBufferY = 3
RollingBufferXY = 4
def display_name(self) -> str:
return ("Standard", "Double Buffer", "Rolling Buffer X", "Rolling Buffer Y", "Rolling Buffer XY")[self.value]
def identifier_name(self) -> str:
return ("standard", "double_buffer", "rolling_buffer_x", "rolling_buffer_y", "rolling_buffer_xy")[self.value]
@staticmethod
def all():
return (
TensorSubPurpose.Standard,
TensorSubPurpose.DoubleBuffer,
TensorSubPurpose.RollingBufferX,
TensorSubPurpose.RollingBufferY,
TensorSubPurpose.RollingBufferXY,
)
class TensorFormat(enum.Flag):
Unknown = 0
WeightsCompressed = 1
NHWC = 2
NHCWB16 = 3
def __str__(self):
return self.name
class TensorBlockTraversal(enum.Enum):
Default = 0
DepthWise = 1
DepthFirst = 2
PartKernelFirst = 3
def shape_num_elements(shp: Shape) -> Optional[int]:
elems = 1
if shp is None:
return None
for d in shp:
if d is None:
return None
elems *= d
return elems
def shape_fully_defined(shp: Shape) -> bool:
if shp is None:
return False
for d in shp:
if d is None:
return False
return True
def shape_round_to_quantum(shp: Shape, quantum: Tuple) -> Shape:
new_shp = list(shp)
# Traverse backwards using length of shape since there may be more rounding quantums than shape elements
for i in range(-1, -len(shp) - 1, -1):
if new_shp[i] is not None:
new_shp[i] = numeric_util.round_up(new_shp[i], quantum[i])
return new_shp
@lru_cache(maxsize=None)
def create_equivalence_id(key) -> UUID:
# Generates equivalence_id based on the given key.
return uuid.uuid4()
class QuantizationParameters:
__slots__ = (
"min",
"max",
"num_bits",
"narrow_range",
"next_after",
"scale_f32",
"zero_point",
"quant_min",
"quant_max",
"quant_dim",
)
def __init__(
self,
min: Union[float, np.ndarray, None] = None,
max: Union[float, np.ndarray, None] = None,
num_bits=None,
narrow_range=None,
scale_f32: Union[float, np.ndarray, None] = None,
zero_point: Union[int, np.ndarray, None] = None,
):
self.min = min
self.max = max
self.num_bits = num_bits
self.narrow_range = narrow_range
# Use the 'next after' float value of scale_f32 when converting to scale and shift. It can be combined with
# natural rounding to perform rounding away from zero. This only affects the ofm scale and bias tensor, it has
# no affect on global scaling i.e. the ofm_scale register
self.next_after = False
self.scale_f32: Union[float, np.ndarray, None] = scale_f32
self.zero_point: Union[int, np.ndarray, None] = zero_point
self.quant_min: Optional[float] = None
self.quant_max: Optional[float] = None
self.quant_dim: Optional[int] = None
def __str__(self):
return (
f"<nng.QuantizationParameters min={self.min}, max={self.max}, num_bits={self.num_bits}, "
f"scale={self.scale_f32}, zero_point={self.zero_point}, next={self.next_after}>"
)
__repr__ = __str__
def clone(self) -> "QuantizationParameters":
res = QuantizationParameters()
res.min = self.min
res.max = self.max
res.num_bits = self.num_bits
res.narrow_range = self.narrow_range
res.next_after = self.next_after
res.scale_f32 = self.scale_f32
res.zero_point = self.zero_point
res.quant_min = self.quant_min
res.quant_max = self.quant_max
res.quant_dim = self.quant_dim
return res
def dequantize(self, values) -> np.ndarray:
return np.subtract(values, self.zero_point) * self.scale_f32
def is_scaling_equal(self, other: Optional["QuantizationParameters"]) -> bool:
"""
Returns True if the scale and zero point of self and other are equal. If other is None then the scaling is
not considered equal because the tensor is assumed to not be quantised and False will be returned
"""
if not isinstance(other, QuantizationParameters):
return False
return self.scale_f32 == other.scale_f32 and self.zero_point == other.zero_point
def is_valid(self) -> bool:
"""Return True if the quantisation parameters have a scale and zero point"""
return self.scale_f32 is not None and self.zero_point is not None
def is_per_axis(self) -> bool:
"""Returns True if either the scale, zero point, minimum or maximum values have more than one value"""
for attr in ("scale_f32", "zero_point", "min", "max"):
if np.size(getattr(self, attr)) > 1:
return True
return False
def create_virtual_tensor(
name: str,
):
virtual_tensor = Tensor([], DataType.int8, name)
virtual_tensor.purpose = TensorPurpose.Virtual
return virtual_tensor
def create_const_tensor(
name: str,
shape: Shape,
dtype: DataType, # datatype of the tensor
values: Optional[Union[np.ndarray, list]], # list-like data of some type, or scalar (skip mypy), or None
purpose: TensorPurpose = TensorPurpose.Unknown,
quantization: Optional[QuantizationParameters] = None,
):
assert isinstance(dtype, DataType)
# Tensor
const_tensor = Tensor(shape, dtype, name + "_0")
const_tensor.purpose = purpose
const_tensor.quantization = quantization
# if the tensor datatype does not match that of the values then np.array() will perform a cast operation. this can
# result in undefined behaviour if casting from a numpy float to a numpy unsigned integer. therefore, we need to
# avoid this undefined behaviour by converting the numpy floats to python floats as these give the desired behaviour
# when casting to unsigned integers
if (
values is not None
and shape != [] # values are not a scalar
and isinstance(values[0], np.floating)
and dtype.type == BaseType.Unsigned
):
values = [float(v) for v in values]
const_tensor.values = np.array(values, dtype=dtype.as_numpy_type())
# Operator
const_op = Operation(Op.Const, name)
const_op.set_output_tensor(const_tensor)
const_op.set_ifm_ofm_shapes()
return const_tensor
# class that keeps track of all tensor addresses in the different memory types
class TensorAddressMap:
address_map: Dict = defaultdict(dict) # dict (tens.equivalence_id -> dict (mem_type -> address))
@classmethod
def get_address_for_tens(cls, tens_id: UUID, mem_type: MemType) -> int:
return cls.address_map[tens_id].get(mem_type)
@classmethod
def set_address_for_tens(cls, tens_id: UUID, mem_type: MemType, address: int):
# Check previous address if there is one
previous_address = cls.address_map[tens_id].get(mem_type)
if address is not None and previous_address is not None:
assert previous_address == address, "Two different addresses cannot be assigned to the same tensor."
# Set tensor's address for memory type
cls.address_map[tens_id][mem_type] = address
@total_ordering
class Tensor:
__slots__ = (
"shape",
"_original_shape",
"storage_shape",
"bandwidth_shape",
"dtype",
"name",
"is_variable",
"pre_buffer",
"ops",
"consumer_list",
"values",
"compressed_values",
"compressed_values_substream_offsets",
"mem_area",
"mem_type",
"format",
"purpose",
"sub_purpose",
"alignment",
"weight_transpose_depthwise",
"storage_compression_scale",
"bandwidth_compression_scale",
"compression_scale_for_worst_weight_stream",
"weight_compression_scales",
"weight_compression_config",
"value_id",
"storage_rounding_quantum",
"brick_size",
"quantization",
"weight_compressed_offsets",
"element_size_bytes",
"block_traversal",
"equivalence_id",
"src_tensor",
"force_linear_format",
"ifm_write_protected",
)
AllocationQuantum = 16
def __init__(self, shape: Shape, dtype: DataType, name: str):
self.shape = shape
self._original_shape = shape
self.storage_shape = shape
self.bandwidth_shape = shape
self.dtype = dtype
self.name = name
self.is_variable = False
self.pre_buffer = False
self.equivalence_id: UUID = uuid.uuid4()
self.ops: List[Operation] = []
self.consumer_list: List[Operation] = []
self.values: Optional[np.ndarray] = None # elements are of type self.dtype
self.compressed_values: Optional[np.ndarray] = None
self.compressed_values_substream_offsets: Optional[List] = None
self.mem_area: MemArea = MemArea.Unknown
self.mem_type: MemType = MemType.Unknown
self.format: TensorFormat = TensorFormat.Unknown
self.purpose: TensorPurpose = TensorPurpose.Unknown
self.sub_purpose: TensorSubPurpose = TensorSubPurpose.Standard
self.alignment: int = Tensor.AllocationQuantum
self.weight_transpose_depthwise: bool = False
self.storage_compression_scale: float = 1.0
self.bandwidth_compression_scale: float = 1.0
self.compression_scale_for_worst_weight_stream: float = 1.0
self.weight_compression_scales: Optional[np.ndarray] = None
# if two tensors have the same weight_compression_config, then they have the same compressed values
self.weight_compression_config = None
# if two tensors have the same value_id, then they have the same values
self.value_id: UUID = uuid.uuid4()
self.weight_compressed_offsets: List = []
self.storage_rounding_quantum: Tuple = (1, 1, 1, 1)
self.brick_size: Tuple = (1, 1, 1, 1)
self.element_size_bytes: int = 0
# quantization parameters
self.quantization: Optional[QuantizationParameters] = None
self.block_traversal: TensorBlockTraversal = TensorBlockTraversal.Default
# Keep track of whether the linear format should be enforced
self.force_linear_format: Optional[bool] = None
self.ifm_write_protected = False
# Reference to parent-tensor if this tensor is a clone
self.src_tensor: Optional[Tensor] = None
@property
def use_linear_format(self) -> bool:
"""Return whether the tensor should use linear format or not."""
return self.force_linear_format in (True, None)
@property
def original_shape(self):
return self._original_shape
@property
def address(self) -> int:
return TensorAddressMap.get_address_for_tens(self.equivalence_id, self.mem_type)
@address.setter
def address(self, address: int):
TensorAddressMap.set_address_for_tens(self.equivalence_id, self.mem_type, address)
@property
def is_standard_fm(self) -> bool:
return self.sub_purpose == TensorSubPurpose.Standard and self.purpose == TensorPurpose.FeatureMap
@property
def is_const(self) -> bool:
return self.ops != [] and self.ops[0].type == Op.Const
@property
def is_scalar(self) -> bool:
return self.shape == [] and self.elements() == 1
def is_broadcast(self, ofm) -> bool:
return self.shape != ofm.shape
def element_size(self) -> int:
if self.element_size_bytes == 0:
return self.dtype.size_in_bits() // 8
return self.element_size_bytes
# Returns a copy, renamed to self.name + suffix
# The references to Operators will be empty when returned
# Depending on set_unique, the copy is shallow, or deep
# For set_unique==True, a new equivalence_id will be set
def clone(self, suffix="_clone", set_unique: bool = False) -> "Tensor":
res = copy.copy(self)
if set_unique:
res.equivalence_id = uuid.uuid4()
res.storage_shape = list(self.storage_shape)
res.bandwidth_shape = list(self.bandwidth_shape)
if self.quantization is not None:
res.quantization = self.quantization.clone()
res.name = res.name + suffix
res.ops = []
res.consumer_list = []
return res
def clone_into_shram(self, arch) -> "Tensor":
res = self.clone(suffix="_shram")
res.mem_area = MemArea.Shram
res.src_tensor = self
return res
def as_1D(self):
self.shape = [np.prod(self.shape)]
if self.values is not None:
self.values = self.values.reshape(self.shape)
def transpose(self, reorder):
self.shape = [self.shape[idx] for idx in reorder]
self._original_shape = [self._original_shape[idx] for idx in reorder]
if self.values is not None:
self.values = self.values.transpose(reorder)
def copy_compressed_weight_info(self, src_tens: "Tensor"):
# Copies compressed values + all related weight compression info from the given tensor
self.equivalence_id = src_tens.equivalence_id
self.compressed_values = src_tens.compressed_values
self.compressed_values_substream_offsets = src_tens.compressed_values_substream_offsets
self.storage_shape = src_tens.storage_shape
self.brick_size = src_tens.brick_size
self.weight_compression_scales = src_tens.weight_compression_scales
self.weight_compressed_offsets = src_tens.weight_compressed_offsets
self.weight_transpose_depthwise = src_tens.weight_transpose_depthwise
self.compression_scale_for_worst_weight_stream = src_tens.compression_scale_for_worst_weight_stream
self.storage_compression_scale = src_tens.storage_compression_scale
self.bandwidth_compression_scale = src_tens.bandwidth_compression_scale
self.block_traversal = src_tens.block_traversal
self.weight_compression_config = src_tens.weight_compression_config
self.value_id = src_tens.value_id
def set_format(self, fmt: TensorFormat, arch):
self.format = fmt
shape_len = 0
try:
shape_len = len(self.shape)
except TypeError:
pass
if shape_len > 4:
return
assert not (self.use_linear_format and fmt == TensorFormat.NHCWB16)
self.storage_rounding_quantum = arch.storage_rounding_quantums[self.format]
self.storage_rounding_quantum = tuple(self.storage_rounding_quantum[-shape_len:])
self.brick_size = arch.brick_sizes[self.format]
self.brick_size = tuple(self.brick_size[-shape_len:])
if self.shape is None:
return
self.bandwidth_shape = shape_round_to_quantum(self.shape, self.brick_size)
self.storage_shape = shape_round_to_quantum(self.shape, self.storage_rounding_quantum)
if fmt == TensorFormat.WeightsCompressed:
compression_ratio = 5 / 8
self.storage_compression_scale = compression_ratio
self.bandwidth_compression_scale = compression_ratio
self.compression_scale_for_worst_weight_stream = compression_ratio
def storage_elements(self) -> int:
elems = shape_num_elements(self.storage_shape)
if elems is None:
return 0
return elems
def elements(self) -> int:
elems = shape_num_elements(self.shape)
if elems is None:
return 0
return elems
def has_fully_defined_shape(self) -> bool:
return shape_fully_defined(self.shape)
def storage_size(self, scale: float = 1.0) -> int:
raw_size = self.storage_elements() * self.element_size() * scale
if raw_size == 0:
raw_size = 1 # force it to take up space
rounded_size = numeric_util.round_up(numeric_util.round_up_to_int(raw_size), self.alignment)
return rounded_size
def storage_size_for_shape(self, op_storage_shape: Shape) -> int:
elems = shape_num_elements(op_storage_shape)
elems = elems if elems else 0
raw_size = elems * self.element_size()
if raw_size == 0:
raw_size = 1 # force it to take up space
rounded_size = numeric_util.round_up(numeric_util.round_up_to_int(raw_size), self.alignment)
return rounded_size
def storage_shape_for_sub_purpose(
self, sub_purpose: TensorSubPurpose, param_a: Optional[int], param_b: Optional[int]
) -> Shape:
if sub_purpose == TensorSubPurpose.DoubleBuffer:
shp = list(self.shape)
assert len(shp) >= 2
assert param_a is not None
shp[-1] = min(shp[-1], param_a * 2)
else:
shp = full_shape(4, self.storage_shape, 1)
if sub_purpose == TensorSubPurpose.RollingBufferX:
assert len(shp) == 4
assert param_a is not None
shp[0] = 1
shp[2] = min(shp[2], param_a)
elif sub_purpose == TensorSubPurpose.RollingBufferY:
assert len(shp) == 4
assert param_a is not None
shp[0] = 1
shp[1] = min(shp[1], param_a)
elif sub_purpose == TensorSubPurpose.RollingBufferXY:
assert len(shp) == 4
assert param_a is not None
assert param_b is not None
shp[0] = 1
shp[2] = min(shp[2], param_a)
shp[1] = min(shp[1], param_b)
elif sub_purpose == TensorSubPurpose.Standard:
pass
else:
assert 0, "did not expect new sub purpose %s" % (sub_purpose,)
return shp
def set_new_sub_purpose(self, sub_purpose: TensorSubPurpose, param_a=None, param_b=None):
self.storage_shape = self.storage_shape_for_sub_purpose(sub_purpose, param_a, param_b)
self.sub_purpose = sub_purpose
if sub_purpose == TensorSubPurpose.DoubleBuffer:
self.storage_compression_scale = self.compression_scale_for_worst_weight_stream
def bandwidth(self) -> float:
elems = shape_num_elements(self.bandwidth_shape)
if elems is None:
return 0
return elems * self.element_size() * self.bandwidth_compression_scale
def consumers(self) -> List[Operation]:
return self.consumer_list
def get_4D_storage_shape_for_shape(self, op_shape4D: Shape4D) -> Shape4D:
rounding_quantum = full_shape(4, list(self.storage_rounding_quantum), 1)
return Shape4D(shape_round_to_quantum(op_shape4D.as_list(), rounding_quantum))
def addresses_for_rolling_buffer(
self, start_coord: Shape, end_coord: Shape, strides: List[int], op_shape4D: Shape4D
) -> Tuple:
# returns ( box_height0, box_height1, box_width, [address_tl, address_tr, address_bl, address_br] )
if self.storage_shape == []:
return (
1,
1,
1,
[self.address_for_coordinate(start_coord, strides, op_shape4D), 0, 0, 0],
)
if self.is_standard_fm:
storage_shape_4D = self.get_4D_storage_shape_for_shape(op_shape4D)
else:
storage_shape_4D = Shape4D(self.storage_shape)
crossing_y = numeric_util.round_up(start_coord[1] + 1, storage_shape_4D.height)
crossing_x = numeric_util.round_up(start_coord[2] + 1, storage_shape_4D.width)
crossing_y = min(crossing_y, end_coord[1])
crossing_x = min(crossing_x, end_coord[2])
box_height0 = crossing_y - start_coord[1]
box_width = crossing_x - start_coord[2]
addresses: List = [0] * 4
addresses[0] = self.address_for_coordinate(start_coord, strides, op_shape4D)
if end_coord[2] > crossing_x:
addresses[1] = self.address_for_coordinate(
[start_coord[0], start_coord[1], crossing_x, start_coord[3]], strides, op_shape4D
)
raise UnsupportedFeatureError("Striping in vertical direction is not supported")
if end_coord[1] > crossing_y:
addresses[2] = self.address_for_coordinate(
[start_coord[0], crossing_y, start_coord[2], start_coord[3]], strides, op_shape4D
)
if end_coord[1] > crossing_y and end_coord[2] > crossing_x:
addresses[3] = self.address_for_coordinate(
[start_coord[0], crossing_y, crossing_x, start_coord[3]], strides, op_shape4D
)
return box_height0, box_height0, box_width, addresses
def get_strides(self, shape4D: Optional[Shape4D]) -> List[int]:
augmented_shape = self.get_augmented_shape(shape4D)
assert len(augmented_shape) == 5
strides: List = [0] * len(augmented_shape)
stride = self.element_size() * self.storage_compression_scale
if self.format != TensorFormat.NHCWB16:
stride_order = [4, 1, 3, 2, 0]
for i in stride_order:
strides[i] = stride
stride *= augmented_shape[i]
else:
strides[4] = stride
strides[3] = 16 * stride # STRIDE_X
strides[1] = strides[3] * augmented_shape[2] # STRIDE_C
strides[2] = augmented_shape[2] * augmented_shape[3] * stride # STRIDE_Y
strides[0] = strides[2] * augmented_shape[1] # STRIDE_N
return strides
def get_augmented_shape(self, shape4D: Optional[Shape4D] = None) -> Optional[Shape]:
if shape4D and self.is_standard_fm:
augmented_shape = self.get_4D_storage_shape_for_shape(shape4D).as_list()
else:
augmented_shape = full_shape(4, self.storage_shape, 1)
if self.format == TensorFormat.NHWC:
augmented_shape = [augmented_shape[0], augmented_shape[3]] + augmented_shape[1:3] + [1]
elif self.format == TensorFormat.NHCWB16:
augmented_shape = augmented_shape[0:4] + [1]
if augmented_shape[1] == 0:
augmented_shape[1] = 1
else:
assert self.format in (TensorFormat.Unknown, TensorFormat.WeightsCompressed)
return None
return augmented_shape
def get_augmented_coord(self, coord: Optional[Shape] = None) -> Optional[Shape]:
if coord is None:
coord = [0] * min(len(self.storage_shape), 4)
missing_len = 4 - len(coord)
augmented_coord = ([0] * missing_len) + coord
if self.format == TensorFormat.NHWC:
augmented_coord = [augmented_coord[0], augmented_coord[3]] + augmented_coord[1:3] + [0]
elif self.format == TensorFormat.NHCWB16:
channel_divisor = 16
augmented_coord = (
[augmented_coord[0], augmented_coord[3] // channel_divisor]
+ augmented_coord[1:3]
+ [augmented_coord[3] % channel_divisor]
)
else:
assert self.format in (TensorFormat.Unknown, TensorFormat.WeightsCompressed)
return None
return augmented_coord
def find_npu_op(self) -> Optional[Operation]:
# Returns the NPU operator that uses this tensor
for op in self.consumers():
if op.run_on_npu:
return op
return None
def compressed_stream_index_from_coord(self, coord: Shape) -> int:
assert self.format == TensorFormat.WeightsCompressed
assert self.compressed_values is not None
assert len(self.compressed_values) > 0
assert len(self.compressed_values) + 1 == len(self.weight_compressed_offsets)
depth = coord[-1]
brick_depth = self.brick_size[-1]
# Clamp position at final element index
if depth > self.shape[-1]:
depth = self.shape[-1]
# Always round up to next boundary
index = numeric_util.round_up_divide(depth, brick_depth)
# Check boundaries on all but last weight set (which may be shorter
# than the brick we divided it up into)
if index < len(self.weight_compressed_offsets) - 1:
# There are no half-way points in the weights
if (depth % brick_depth) != 0:
raise UnsupportedFeatureError("Offset into weights must be aligned to a brick")
return index
def size_of_compressed_stream(self, index: int) -> int:
assert self.compressed_values is not None
assert 0 <= index < len(self.compressed_values)
return len(self.compressed_values[index])
def is_last_index_in_compressed_stream(self, index: int) -> bool:
assert self.compressed_values is not None
assert 0 <= index < len(self.compressed_values)
return index == len(self.compressed_values) - 1
def address_for_coordinate(
self,
orig_coord: Shape,
strides: Optional[List[int]] = None,
op_shape4D: Optional[Shape4D] = None,
is_top_box: bool = False,
) -> Optional[int]:
address_offset = 0
assert self.purpose != TensorPurpose.Weights
# Strides may be passed as an argument, for example when creating feature maps as the strides may be modified
# by the "ofm_stride_multiplier" operation attribute. If not, they are calculated here.
if not strides:
strides = self.get_strides(op_shape4D)
coord = orig_coord
if is_top_box:
coord = [c - 1 for c in orig_coord]
address_offset += 1 * strides[-1] # one element
if self.sub_purpose == TensorSubPurpose.Standard:
shape = op_shape4D.as_list() if op_shape4D else self.shape
for _coord, _shape in zip(coord, shape):
assert _coord >= 0 and _coord < _shape
if op_shape4D and self.is_standard_fm:
storage_shape = self.get_4D_storage_shape_for_shape(op_shape4D).as_list()
storage_size = self.storage_size_for_shape(storage_shape)
else:
storage_shape = self.storage_shape
coord = coord[-len(storage_shape) :]
storage_size = self.storage_size()
# Handle wraparound for partial buffers. Make sure to do this after subtracting top box
coord = [_coord % _shape for _coord, _shape in zip(coord, storage_shape)]
augmented_coord = self.get_augmented_coord(coord)
assert augmented_coord is not None
address_offset += np.dot(augmented_coord, strides)
assert address_offset >= 0 and address_offset <= storage_size
return self.address + address_offset
def is_allocated_in_tensor_arena(self, scratch_tensor_mem_area: MemArea) -> bool:
return (self.mem_area == scratch_tensor_mem_area) and (self.mem_type in (MemType.Scratch, MemType.Scratch_fast))
def equivalent(self, tens: "Tensor") -> bool:
return self.equivalence_id == tens.equivalence_id
def set_all_shapes(self, shape: Shape):
self.shape = shape
self.storage_shape = shape
self.bandwidth_shape = shape
def get_full_shape(self) -> Shape:
d = len(self.shape)
if d in (1, 3):
return full_shape(4, self.shape, 1)
elif d == 2:
return [self.shape[0], 1, 1, self.shape[1]]
else:
return self.shape.copy()
def is_quantized(self) -> bool:
# a tensor is quantized if it has an integral type and it contains valid quantization params
if not isinstance(self.quantization, QuantizationParameters):
return False
return (self.dtype.type & BaseType.Int) != 0 and self.quantization.is_valid()
def get_scalar(self):
"""
return: Unquantized or dequantized scalar value
rtype: self.dtype (if unquantized) or float (if dequantized)
"""
assert self.values.size == 1, "get_scalar called on non-scalar tensor"
if self.is_quantized():
return self.quantization.dequantize(self.values).item(0)
else:
return self.values.item(0)
def get_shape_as_2d(self, dimension_2_size: int) -> Optional[Shape4D]:
elms = self.elements()
dimension_1_size = elms // dimension_2_size
# Checks if the reduction works and shape is not 1D
is_reducible = dimension_1_size * dimension_2_size == elms and not (len(self.shape) == 1)
new_shape = None
if is_reducible:
new_shape = Shape4D([dimension_1_size, 1, 1, dimension_2_size])
return new_shape
def __lt__(self, other: "Tensor") -> bool:
return self.equivalence_id < other.equivalence_id
def __str__(self):
return "<nng.Tensor '%s' shape=%s dtype=%s>" % (self.name, self.shape, self.dtype)
__repr__ = __str__
def error(self, msg):
"""
Raises a VelaError exception for errors encountered when parsing a Tensor
:param self: Tensor object that resulted in the error
:param msg: str object that contains a description of the specific error encountered
"""
def _print_operators(ops):
lines = []
for idx, op in enumerate(ops):
op_type = getattr(op, "type", "Not an Operation")
op_id = getattr(op, "op_index", "-")
lines.append(f" {idx} = {op_type} ({op_id})")
return lines
lines = [f"Invalid {self.name} tensor. {msg}"]
lines += [" Driving operators:"]
lines += _print_operators(self.ops)
lines += [" Consuming operators:"]
lines += _print_operators(self.consumer_list)
raise VelaError("\n".join(lines))
def check_quantized_tens_scaling_equal(tens_a: Tensor, tens_b: Tensor) -> bool:
# checks that the scaling of two quantized tensors are equal
return tens_a.is_quantized() and tens_b.is_quantized() and tens_a.quantization.is_scaling_equal(tens_b.quantization)
|