-
Notifications
You must be signed in to change notification settings - Fork 26.3k
/
modeling_jamba.py
executable file
·1882 lines (1592 loc) · 86.5 KB
/
modeling_jamba.py
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
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# coding=utf-8
# Copyright 2024 AI21 Labs Ltd. and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# 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
#
# http://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.
""" PyTorch Jamba model."""
import inspect
import math
from typing import Any, Dict, List, Optional, Tuple, Union
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
from ...activations import ACT2FN
from ...cache_utils import DynamicCache # we need __iter__ and __len__ of pkv
from ...modeling_attn_mask_utils import (
AttentionMaskConverter,
)
from ...modeling_outputs import (
MoeCausalLMOutputWithPast,
MoeModelOutputWithPast,
SequenceClassifierOutputWithPast,
)
from ...modeling_utils import PreTrainedModel
from ...utils import (
add_start_docstrings,
add_start_docstrings_to_model_forward,
is_flash_attn_greater_or_equal_2_10,
logging,
replace_return_docstrings,
)
from ...utils.import_utils import (
is_causal_conv1d_available,
is_flash_attn_2_available,
is_mamba_ssm_available,
)
from .configuration_jamba import JambaConfig
if is_flash_attn_2_available():
from flash_attn import flash_attn_func, flash_attn_varlen_func
from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa
_flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
if is_mamba_ssm_available():
from mamba_ssm.ops.selective_scan_interface import mamba_inner_fn, selective_scan_fn
from mamba_ssm.ops.triton.selective_state_update import selective_state_update
else:
selective_state_update, selective_scan_fn, mamba_inner_fn = None, None, None
if is_causal_conv1d_available():
from causal_conv1d import causal_conv1d_fn, causal_conv1d_update
else:
causal_conv1d_update, causal_conv1d_fn = None, None
is_fast_path_available = all(
(selective_state_update, selective_scan_fn, causal_conv1d_fn, causal_conv1d_update, mamba_inner_fn)
)
logger = logging.get_logger(__name__)
_CONFIG_FOR_DOC = "JambaConfig"
# Copied from transformers.models.mixtral.modeling_mixtral.load_balancing_loss_func with gate->router
def load_balancing_loss_func(
router_logits: torch.Tensor,
num_experts: torch.Tensor = None,
top_k=2,
attention_mask: Optional[torch.Tensor] = None,
) -> float:
r"""
Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.
See Switch Transformer (https://arxiv.org/abs/2101.03961) for more details. This function implements the loss
function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
experts is too unbalanced.
Args:
router_logits (Union[`torch.Tensor`, Tuple[torch.Tensor]):
Logits from the `router`, should be a tuple of model.config.num_hidden_layers tensors of
shape [batch_size X sequence_length, num_experts].
attention_mask (`torch.Tensor`, None):
The attention_mask used in forward function
shape [batch_size X sequence_length] if not None.
num_experts (`int`, *optional*):
Number of experts
Returns:
The auxiliary loss.
"""
if router_logits is None or not isinstance(router_logits, tuple):
return 0
if isinstance(router_logits, tuple):
compute_device = router_logits[0].device
concatenated_router_logits = torch.cat(
[layer_router.to(compute_device) for layer_router in router_logits], dim=0
)
routing_weights = torch.nn.functional.softmax(concatenated_router_logits, dim=-1)
_, selected_experts = torch.topk(routing_weights, top_k, dim=-1)
expert_mask = torch.nn.functional.one_hot(selected_experts, num_experts)
if attention_mask is None:
# Compute the percentage of tokens routed to each experts
tokens_per_expert = torch.mean(expert_mask.float(), dim=0)
# Compute the average probability of routing to these experts
router_prob_per_expert = torch.mean(routing_weights, dim=0)
else:
batch_size, sequence_length = attention_mask.shape
num_hidden_layers = concatenated_router_logits.shape[0] // (batch_size * sequence_length)
# Compute the mask that masks all padding tokens as 0 with the same shape of expert_mask
expert_attention_mask = (
attention_mask[None, :, :, None, None]
.expand((num_hidden_layers, batch_size, sequence_length, top_k, num_experts))
.reshape(-1, top_k, num_experts)
.to(compute_device)
)
# Compute the percentage of tokens routed to each experts
tokens_per_expert = torch.sum(expert_mask.float() * expert_attention_mask, dim=0) / torch.sum(
expert_attention_mask, dim=0
)
# Compute the mask that masks all padding tokens as 0 with the same shape of tokens_per_expert
router_per_expert_attention_mask = (
attention_mask[None, :, :, None]
.expand((num_hidden_layers, batch_size, sequence_length, num_experts))
.reshape(-1, num_experts)
.to(compute_device)
)
# Compute the average probability of routing to these experts
router_prob_per_expert = torch.sum(routing_weights * router_per_expert_attention_mask, dim=0) / torch.sum(
router_per_expert_attention_mask, dim=0
)
overall_loss = torch.sum(tokens_per_expert * router_prob_per_expert.unsqueeze(0))
return overall_loss * num_experts
# Copied from transformers.models.llama.modeling_llama._get_unpad_data
def _get_unpad_data(attention_mask):
seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
max_seqlen_in_batch = seqlens_in_batch.max().item()
cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
return (
indices,
cu_seqlens,
max_seqlen_in_batch,
)
# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Jamba
class JambaRMSNorm(nn.Module):
def __init__(self, hidden_size, eps=1e-6):
"""
JambaRMSNorm is equivalent to T5LayerNorm
"""
super().__init__()
self.weight = nn.Parameter(torch.ones(hidden_size))
self.variance_epsilon = eps
def forward(self, hidden_states):
input_dtype = hidden_states.dtype
hidden_states = hidden_states.to(torch.float32)
variance = hidden_states.pow(2).mean(-1, keepdim=True)
hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
return self.weight * hidden_states.to(input_dtype)
# Copied from transformers.models.llama.modeling_llama.repeat_kv
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
"""
This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
"""
batch, num_key_value_heads, slen, head_dim = hidden_states.shape
if n_rep == 1:
return hidden_states
hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
class HybridMambaAttentionDynamicCache(DynamicCache):
"""
A dynamic cache that can handle both the attention cache (which has a seq_len dimension) and the mamba cache
(which has a constant shape regardless of seq_len).
This cache has two sets of lists of tensors: `key_cache` and `value_cache` for attention cache and `conv_states`
and `ssm_states` for mamba cache. Each of these lists has `num_layers` tensors. The expected shape for each tensor
For attention layers, `key_cache` and `value_cache` have a shape of `(batch_size, num_heads, seq_len, head_dim)`,
while `conv_states` and `ssm_states` have a shape of `(batch_size, 0)` (empty tensors).
For mamba layers, `key_cache` and `value_cache` have a shape of `(batch_size, 0)` (empty tensors),
while `conv_states` represents the convolution state and has a shape of `(batch_size, d_inner, d_conv)`,
and `ssm_states` represents the ssm state and has a shape of `(batch_size, d_inner, d_state)`.
"""
def __init__(self, config, batch_size, dtype=torch.float16, device=None):
self.dtype = dtype
self.layers_block_type = config.layers_block_type
self.has_previous_state = False # only used by mamba
intermediate_size = config.mamba_expand * config.hidden_size
ssm_state_size = config.mamba_d_state
conv_kernel_size = config.mamba_d_conv
self.conv_states = []
self.ssm_states = []
for i in range(config.num_hidden_layers):
if self.layers_block_type[i] == "mamba":
self.conv_states += [
torch.zeros(batch_size, intermediate_size, conv_kernel_size, device=device, dtype=dtype)
]
self.ssm_states += [
torch.zeros(batch_size, intermediate_size, ssm_state_size, device=device, dtype=dtype)
]
else:
self.conv_states += [torch.tensor([[]] * batch_size, device=device)]
self.ssm_states += [torch.tensor([[]] * batch_size, device=device)]
self.key_cache = [torch.tensor([[]] * batch_size, device=device) for _ in range(config.num_hidden_layers)]
self.value_cache = [torch.tensor([[]] * batch_size, device=device) for _ in range(config.num_hidden_layers)]
def update(
self,
key_states: torch.Tensor,
value_states: torch.Tensor,
layer_idx: int,
cache_kwargs: Optional[Dict[str, Any]] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
# Update the cache
if self.key_cache[layer_idx].shape[-1] == 0:
self.key_cache[layer_idx] = key_states
self.value_cache[layer_idx] = value_states
else:
self.key_cache[layer_idx] = torch.cat([self.key_cache[layer_idx], key_states], dim=2)
self.value_cache[layer_idx] = torch.cat([self.value_cache[layer_idx], value_states], dim=2)
return self.key_cache[layer_idx], self.value_cache[layer_idx]
def reorder_cache(self, beam_idx: torch.LongTensor):
"""Reorders the cache for beam search, given the selected beam indices."""
for layer_idx in range(len(self.key_cache)):
device = self.key_cache[layer_idx].device
self.key_cache[layer_idx] = self.key_cache[layer_idx].index_select(0, beam_idx.to(device))
device = self.value_cache[layer_idx].device
self.value_cache[layer_idx] = self.value_cache[layer_idx].index_select(0, beam_idx.to(device))
device = self.conv_states[layer_idx].device
self.conv_states[layer_idx] = self.conv_states[layer_idx].index_select(0, beam_idx.to(device))
device = self.ssm_states[layer_idx].device
self.ssm_states[layer_idx] = self.ssm_states[layer_idx].index_select(0, beam_idx.to(device))
def to_legacy_cache(self) -> Tuple[Tuple[torch.Tensor], Tuple[torch.Tensor]]:
raise NotImplementedError("HybridMambaAttentionDynamicCache does not have a legacy cache equivalent.")
@classmethod
def from_legacy_cache(cls, past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None) -> "DynamicCache":
raise NotImplementedError("HybridMambaAttentionDynamicCache does not have a legacy cache equivalent.")
# Adapted from transformers.models.mistral.modeling_mistral.MistralAttention with Mistral->Jamba
class JambaAttention(nn.Module):
"""
Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
and "Generating Long Sequences with Sparse Transformers".
"""
def __init__(self, config: JambaConfig, layer_idx: Optional[int] = None):
super().__init__()
self.config = config
self.layer_idx = layer_idx
if layer_idx is None:
logger.warning_once(
f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
"lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
"when creating this class."
)
self.hidden_size = config.hidden_size
self.num_heads = config.num_attention_heads
self.head_dim = self.hidden_size // self.num_heads
self.num_key_value_heads = config.num_key_value_heads
self.num_key_value_groups = self.num_heads // self.num_key_value_heads
self.is_causal = True
self.attention_dropout = config.attention_dropout
if (self.head_dim * self.num_heads) != self.hidden_size:
raise ValueError(
f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
f" and `num_heads`: {self.num_heads})."
)
self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False)
self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_value: Optional[HybridMambaAttentionDynamicCache] = None,
output_attentions: bool = False,
use_cache: bool = False,
cache_position: Optional[torch.LongTensor] = None,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
bsz, q_len, _ = hidden_states.size()
query_states = self.q_proj(hidden_states)
key_states = self.k_proj(hidden_states)
value_states = self.v_proj(hidden_states)
query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
if past_key_value is not None:
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx)
# repeat k/v heads if n_kv_heads < n_heads
key_states = repeat_kv(key_states, self.num_key_value_groups)
value_states = repeat_kv(value_states, self.num_key_value_groups)
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
if attention_mask is not None: # no matter the length, we just slice it
causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
attn_weights = attn_weights + causal_mask
# upcast attention to fp32
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
attn_output = torch.matmul(attn_weights, value_states)
if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
raise ValueError(
f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
f" {attn_output.size()}"
)
attn_output = attn_output.transpose(1, 2).contiguous()
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
attn_output = self.o_proj(attn_output)
if not output_attentions:
attn_weights = None
return attn_output, attn_weights, past_key_value
# Adapted from transformers.models.mistral.modeling_mistral.MistralFlashAttention2 with Mistral->Jamba
class JambaFlashAttention2(JambaAttention):
"""
Jamba flash attention module. This module inherits from `JambaAttention` as the weights of the module stays
untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
flash attention and deal with padding tokens in case the input contains any of them.
"""
# Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2.__init__
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
# TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
# flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
# Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_value: Optional[HybridMambaAttentionDynamicCache] = None,
output_attentions: bool = False,
use_cache: bool = False,
cache_position: Optional[torch.LongTensor] = None,
**kwargs,
):
bsz, q_len, _ = hidden_states.size()
query_states = self.q_proj(hidden_states)
key_states = self.k_proj(hidden_states)
value_states = self.v_proj(hidden_states)
# Flash attention requires the input to have the shape
# batch_size x seq_length x head_dim x hidden_dim
# therefore we just need to keep the original shape
query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
kv_seq_len = cache_position[-1]
use_sliding_windows = (
_flash_supports_window_size
and getattr(self.config, "sliding_window", None) is not None
and kv_seq_len > self.config.sliding_window
)
if not _flash_supports_window_size:
logger.warning_once(
"The current flash attention version does not support sliding window attention, for a more memory efficient implementation"
" make sure to upgrade flash-attn library."
)
if past_key_value is not None:
# Activate slicing cache only if the config has a value `sliding_windows` attribute
cache_has_contents = cache_position[0] > 0
if (
getattr(self.config, "sliding_window", None) is not None
and kv_seq_len > self.config.sliding_window
and cache_has_contents
):
slicing_tokens = 1 - self.config.sliding_window
past_key = past_key_value[self.layer_idx][0]
past_value = past_key_value[self.layer_idx][1]
past_key = past_key[:, :, slicing_tokens:, :].contiguous()
past_value = past_value[:, :, slicing_tokens:, :].contiguous()
if past_key.shape[-2] != self.config.sliding_window - 1:
raise ValueError(
f"past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got"
f" {past_key.shape}"
)
if attention_mask is not None:
attention_mask = attention_mask[:, slicing_tokens:]
attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx)
# repeat k/v heads if n_kv_heads < n_heads
key_states = repeat_kv(key_states, self.num_key_value_groups)
value_states = repeat_kv(value_states, self.num_key_value_groups)
dropout_rate = 0.0 if not self.training else self.attention_dropout
# In PEFT, usually we cast the layer norms in float32 for training stability reasons
# therefore the input hidden states gets silently casted in float32. Hence, we need
# cast them back in float16 just to be sure everything works as expected.
input_dtype = query_states.dtype
if input_dtype == torch.float32:
if torch.is_autocast_enabled():
target_dtype = torch.get_autocast_gpu_dtype()
# Handle the case where the model is quantized
elif hasattr(self.config, "_pre_quantization_dtype"):
target_dtype = self.config._pre_quantization_dtype
else:
target_dtype = self.q_proj.weight.dtype
logger.warning_once(
f"The input hidden states seems to be silently casted in float32, this might be related to"
f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
f" {target_dtype}."
)
query_states = query_states.to(target_dtype)
key_states = key_states.to(target_dtype)
value_states = value_states.to(target_dtype)
# Reashape to the expected shape for Flash Attention
query_states = query_states.transpose(1, 2)
key_states = key_states.transpose(1, 2)
value_states = value_states.transpose(1, 2)
attn_output = self._flash_attention_forward(
query_states,
key_states,
value_states,
attention_mask,
q_len,
dropout=dropout_rate,
use_sliding_windows=use_sliding_windows,
)
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
attn_output = self.o_proj(attn_output)
if not output_attentions:
attn_weights = None
return attn_output, attn_weights, past_key_value
def _flash_attention_forward(
self,
query_states,
key_states,
value_states,
attention_mask,
query_length,
dropout=0.0,
softmax_scale=None,
use_sliding_windows=False,
):
"""
Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
first unpad the input, then computes the attention scores and pad the final attention scores.
Args:
query_states (`torch.Tensor`):
Input query states to be passed to Flash Attention API
key_states (`torch.Tensor`):
Input key states to be passed to Flash Attention API
value_states (`torch.Tensor`):
Input value states to be passed to Flash Attention API
attention_mask (`torch.Tensor`):
The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
position of padding tokens and 1 for the position of non-padding tokens.
dropout (`float`, *optional*):
Attention dropout
softmax_scale (`float`, *optional*):
The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
use_sliding_windows (`bool`, *optional*):
Whether to activate sliding window attention.
"""
if not self._flash_attn_uses_top_left_mask:
causal = self.is_causal
else:
# TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
causal = self.is_causal and query_length != 1
# Contains at least one padding token in the sequence
if attention_mask is not None:
batch_size = query_states.shape[0]
query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
query_states, key_states, value_states, attention_mask, query_length
)
cu_seqlens_q, cu_seqlens_k = cu_seq_lens
max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
if not use_sliding_windows:
attn_output_unpad = flash_attn_varlen_func(
query_states,
key_states,
value_states,
cu_seqlens_q=cu_seqlens_q,
cu_seqlens_k=cu_seqlens_k,
max_seqlen_q=max_seqlen_in_batch_q,
max_seqlen_k=max_seqlen_in_batch_k,
dropout_p=dropout,
softmax_scale=softmax_scale,
causal=causal,
)
else:
attn_output_unpad = flash_attn_varlen_func(
query_states,
key_states,
value_states,
cu_seqlens_q=cu_seqlens_q,
cu_seqlens_k=cu_seqlens_k,
max_seqlen_q=max_seqlen_in_batch_q,
max_seqlen_k=max_seqlen_in_batch_k,
dropout_p=dropout,
softmax_scale=softmax_scale,
causal=causal,
window_size=(self.config.sliding_window, self.config.sliding_window),
)
attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
else:
if not use_sliding_windows:
attn_output = flash_attn_func(
query_states,
key_states,
value_states,
dropout,
softmax_scale=softmax_scale,
causal=causal,
)
else:
attn_output = flash_attn_func(
query_states,
key_states,
value_states,
dropout,
softmax_scale=softmax_scale,
causal=causal,
window_size=(self.config.sliding_window, self.config.sliding_window),
)
return attn_output
# Copied from transformers.models.mixtral.modeling_mixtral.MixtralFlashAttention2._upad_input
def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape
# On the first iteration we need to properly re-create the padding mask
# by slicing it on the proper place
if kv_seq_len != attention_mask.shape[-1]:
attention_mask_num_tokens = attention_mask.shape[-1]
attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len :]
indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
if query_length == kv_seq_len:
query_layer = index_first_axis(
query_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k
)
cu_seqlens_q = cu_seqlens_k
max_seqlen_in_batch_q = max_seqlen_in_batch_k
indices_q = indices_k
elif query_length == 1:
max_seqlen_in_batch_q = 1
cu_seqlens_q = torch.arange(
batch_size + 1, dtype=torch.int32, device=query_layer.device
) # There is a memcpy here, that is very bad.
indices_q = cu_seqlens_q[:-1]
query_layer = query_layer.squeeze(1)
else:
# The -q_len: slice assumes left padding.
attention_mask = attention_mask[:, -query_length:]
query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
return (
query_layer,
key_layer,
value_layer,
indices_q,
(cu_seqlens_q, cu_seqlens_k),
(max_seqlen_in_batch_q, max_seqlen_in_batch_k),
)
# Adapted from transformers.models.mistral.modeling_mistral.MistralSdpaAttention with Mistral->Jamba
class JambaSdpaAttention(JambaAttention):
"""
Jamba attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
`JambaAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
SDPA API.
"""
# Adapted from JambaAttention.forward
def forward(
self,
hidden_states: torch.Tensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_value: Optional[HybridMambaAttentionDynamicCache] = None,
output_attentions: bool = False,
use_cache: bool = False,
cache_position: Optional[torch.LongTensor] = None,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
if output_attentions:
# TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
logger.warning_once(
"JambaModel is using JambaSdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
)
return super().forward(
hidden_states=hidden_states,
attention_mask=attention_mask,
position_ids=position_ids,
past_key_value=past_key_value,
output_attentions=output_attentions,
use_cache=use_cache,
)
bsz, q_len, _ = hidden_states.size()
query_states = self.q_proj(hidden_states)
key_states = self.k_proj(hidden_states)
value_states = self.v_proj(hidden_states)
query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
if past_key_value is not None:
key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx)
key_states = repeat_kv(key_states, self.num_key_value_groups)
value_states = repeat_kv(value_states, self.num_key_value_groups)
causal_mask = attention_mask
if attention_mask is not None:
causal_mask = causal_mask[:, :, :, : key_states.shape[-2]]
# SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
# Reference: https://github.com/pytorch/pytorch/issues/112577.
if query_states.device.type == "cuda" and attention_mask is not None:
query_states = query_states.contiguous()
key_states = key_states.contiguous()
value_states = value_states.contiguous()
attn_output = torch.nn.functional.scaled_dot_product_attention(
query_states,
key_states,
value_states,
attn_mask=causal_mask,
dropout_p=self.attention_dropout if self.training else 0.0,
# The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
is_causal=self.is_causal and attention_mask is None and q_len > 1,
)
attn_output = attn_output.transpose(1, 2).contiguous()
attn_output = attn_output.view(bsz, q_len, self.hidden_size)
attn_output = self.o_proj(attn_output)
return attn_output, None, past_key_value
JAMBA_ATTENTION_CLASSES = {
"eager": JambaAttention,
"flash_attention_2": JambaFlashAttention2,
"sdpa": JambaSdpaAttention,
}
# Adapted from transformers.models.mamba.modeling_mamba.MambaMixer
class JambaMambaMixer(nn.Module):
"""
Compute ∆, A, B, C, and D the state space parameters and compute the `contextualized_states`.
A, D are input independent (see Mamba paper [1] Section 3.5.2 "Interpretation of A" for why A isn't selective)
∆, B, C are input-dependent (this is a key difference between Mamba and the linear time invariant S4,
and is why Mamba is called **selective** state spaces)
"""
def __init__(self, config: JambaConfig, layer_idx):
super().__init__()
self.config = config
self.layer_idx = layer_idx
self.hidden_size = config.hidden_size
self.ssm_state_size = config.mamba_d_state
self.conv_kernel_size = config.mamba_d_conv
self.intermediate_size = config.mamba_expand * config.hidden_size
self.time_step_rank = config.mamba_dt_rank
self.use_conv_bias = config.mamba_conv_bias
self.use_bias = config.mamba_proj_bias
self.conv1d = nn.Conv1d(
in_channels=self.intermediate_size,
out_channels=self.intermediate_size,
bias=self.use_conv_bias,
kernel_size=self.conv_kernel_size,
groups=self.intermediate_size,
padding=self.conv_kernel_size - 1,
)
self.activation = config.hidden_act
self.act = ACT2FN[config.hidden_act]
self.use_fast_kernels = config.use_mamba_kernels
# projection of the input hidden states
self.in_proj = nn.Linear(self.hidden_size, self.intermediate_size * 2, bias=self.use_bias)
# selective projection used to make dt, B and C input dependant
self.x_proj = nn.Linear(self.intermediate_size, self.time_step_rank + self.ssm_state_size * 2, bias=False)
# time step projection (discretization)
self.dt_proj = nn.Linear(self.time_step_rank, self.intermediate_size, bias=True)
# S4D real initialization. These are not discretized!
# The core is to load them, compute the discrete states, then write the updated state. Keeps the memory bounded
A = torch.arange(1, self.ssm_state_size + 1, dtype=torch.float32)[None, :]
A = A.expand(self.intermediate_size, -1).contiguous()
self.A_log = nn.Parameter(torch.log(A))
self.D = nn.Parameter(torch.ones(self.intermediate_size))
self.out_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=self.use_bias)
self.dt_layernorm = JambaRMSNorm(self.time_step_rank, eps=config.rms_norm_eps)
self.b_layernorm = JambaRMSNorm(self.ssm_state_size, eps=config.rms_norm_eps)
self.c_layernorm = JambaRMSNorm(self.ssm_state_size, eps=config.rms_norm_eps)
if not is_fast_path_available:
logger.warning_once(
"The fast path is not available because on of `(selective_state_update, selective_scan_fn, causal_conv1d_fn, causal_conv1d_update, mamba_inner_fn)`"
" is None. To install follow https://github.com/state-spaces/mamba/#installation and"
" https://github.com/Dao-AILab/causal-conv1d. If you want to use the naive implementation, set `use_mamba_kernels=False` in the model config"
)
def cuda_kernels_forward(self, hidden_states: torch.Tensor, cache_params: HybridMambaAttentionDynamicCache = None):
batch_size, seq_len, _ = hidden_states.shape
use_precomputed_states = (
cache_params is not None
and cache_params.has_previous_state
and seq_len == 1
and cache_params.conv_states[self.layer_idx].shape[0]
== cache_params.ssm_states[self.layer_idx].shape[0]
== batch_size
)
# 1. Gated MLP's linear projection
projected_states = self.in_proj(hidden_states).transpose(1, 2)
# We can't use `mamba_inner_fn` even if in training and without cache params because we have the
# inner layernorms which isn't supported by this fused kernel
hidden_states, gate = projected_states.chunk(2, dim=1)
# 2. Convolution sequence transformation
conv_weights = self.conv1d.weight.view(self.conv1d.weight.size(0), self.conv1d.weight.size(2))
if use_precomputed_states:
hidden_states = causal_conv1d_update(
hidden_states.squeeze(-1),
cache_params.conv_states[self.layer_idx],
conv_weights,
self.conv1d.bias,
self.activation,
)
hidden_states = hidden_states.unsqueeze(-1)
else:
if cache_params is not None:
conv_states = nn.functional.pad(hidden_states, (self.conv_kernel_size - hidden_states.shape[-1], 0))
cache_params.conv_states[self.layer_idx].copy_(conv_states)
hidden_states = causal_conv1d_fn(hidden_states, conv_weights, self.conv1d.bias, activation=self.activation)
# 3. State Space Model sequence transformation
# 3.a. input varying initialization of time_step, B and C
ssm_parameters = self.x_proj(hidden_states.transpose(1, 2))
time_step, B, C = torch.split(
ssm_parameters, [self.time_step_rank, self.ssm_state_size, self.ssm_state_size], dim=-1
)
time_step = self.dt_layernorm(time_step)
B = self.b_layernorm(B)
C = self.c_layernorm(C)
# Here we need to apply dt_proj without the bias, as the bias is added in the selective scan kernel.
# This is a hack to apply dt_proj while still using the forward pass of `torch.nn.Linear`, which is needed
# in order to make quantization work. Quantization code replaces `torch.nn.Linear` layers with quantized
# linear layers, and requires to call the forward pass directly.
# The original code here was: ```discrete_time_step = self.dt_proj.weight @ time_step.transpose(1, 2)```
time_proj_bias = self.dt_proj.bias
self.dt_proj.bias = None
discrete_time_step = self.dt_proj(time_step).transpose(1, 2)
self.dt_proj.bias = time_proj_bias
A = -torch.exp(self.A_log.float())
# 3.c perform the recurrence y ← SSM(A, B, C)(x)
time_proj_bias = time_proj_bias.float() if time_proj_bias is not None else None
if use_precomputed_states:
scan_outputs = selective_state_update(
cache_params.ssm_states[self.layer_idx],
hidden_states[..., 0],
discrete_time_step[..., 0],
A,
B[:, 0],
C[:, 0],
self.D,
gate[..., 0],
time_proj_bias,
dt_softplus=True,
).unsqueeze(-1)
else:
scan_outputs, ssm_state = selective_scan_fn(
hidden_states,
discrete_time_step,
A,
B.transpose(1, 2),
C.transpose(1, 2),
self.D.float(),
gate,
time_proj_bias,
delta_softplus=True,
return_last_state=True,
)
if ssm_state is not None and cache_params is not None:
cache_params.ssm_states[self.layer_idx].copy_(ssm_state)
# 4. Final linear projection
contextualized_states = self.out_proj(scan_outputs.transpose(1, 2))
return contextualized_states
# fmt: off
def slow_forward(self, input_states, cache_params: HybridMambaAttentionDynamicCache = None):
batch_size, seq_len, _ = input_states.shape
dtype = input_states.dtype
# 1. Gated MLP's linear projection
projected_states = self.in_proj(input_states).transpose(1, 2) # [batch, 2 * intermediate_size, seq_len]
hidden_states, gate = projected_states.chunk(2, dim=1)
use_cache = isinstance(cache_params,HybridMambaAttentionDynamicCache)
# 2. Convolution sequence transformation
if use_cache and cache_params.ssm_states[self.layer_idx].shape[0] == batch_size:
if self.training:
# In training mode, we don't want to perform in-place operations on ssm_state so we can compute the backwards pass
ssm_state = cache_params.ssm_states[self.layer_idx].clone()
else:
ssm_state = cache_params.ssm_states[self.layer_idx]
if cache_params.has_previous_state and seq_len == 1 and \
cache_params.conv_states[self.layer_idx].shape[0] == batch_size:
conv_state = cache_params.conv_states[self.layer_idx] # [batch, intermediate_size, conv_kernel_size]
conv_state = torch.roll(conv_state, shifts=-1, dims=-1)
conv_state[:, :, -1] = hidden_states[:, :, 0]
cache_params.conv_states[self.layer_idx] = conv_state
hidden_states = torch.sum(conv_state * self.conv1d.weight[:, 0, :], dim=-1)
if self.use_conv_bias:
hidden_states += self.conv1d.bias
hidden_states = self.act(hidden_states).to(dtype).unsqueeze(-1) # [batch, intermediate_size, 1] : decoding
else:
conv_state = nn.functional.pad(
hidden_states,
(self.conv_kernel_size - hidden_states.shape[-1], 0)
)
cache_params.conv_states[self.layer_idx] = conv_state
hidden_states = self.act(self.conv1d(hidden_states)[..., :seq_len]) # [batch, intermediate_size, seq_len]
else:
ssm_state = torch.zeros(
(batch_size, self.intermediate_size, self.ssm_state_size),
device=hidden_states.device, dtype=dtype
)
hidden_states = self.act(self.conv1d(hidden_states)[..., :seq_len]) # [batch, intermediate_size, seq_len]
# 3. State Space Model sequence transformation
# 3.a. Selection: [batch, seq_len, self.time_step_rank + self.ssm_state_size * 2]
ssm_parameters = self.x_proj(hidden_states.transpose(1, 2))
time_step, B, C = torch.split(
ssm_parameters, [self.time_step_rank, self.ssm_state_size, self.ssm_state_size], dim=-1
)
time_step = self.dt_layernorm(time_step)
B = self.b_layernorm(B)
C = self.c_layernorm(C)
discrete_time_step = self.dt_proj(time_step) # [batch, seq_len, intermediate_size]
discrete_time_step = nn.functional.softplus(discrete_time_step).transpose(1, 2) # [batch, intermediate_size, seq_len]
# 3.b. Discretization: B and C to [batch, seq_len, intermediate_size, ssm_state_size] (SRAM)
A = -torch.exp(self.A_log.float()) # [intermediate_size, ssm_state_size]
discrete_A = torch.exp(A[None, :, None, :] * discrete_time_step[:, :, :, None]) # [batch, intermediate_size, seq_len, ssm_state_size]
discrete_B = discrete_time_step[:, :, :, None] * B[:, None, :, :].float() # [batch, intermediade_size, seq_len, ssm_state_size]
deltaB_u = discrete_B * hidden_states[:, :, :, None].float()
# 3.c perform the recurrence y ← SSM(A, B, C)(x)
scan_outputs = []
for i in range(seq_len):
ssm_state = discrete_A[:, :, i, :] * ssm_state + deltaB_u[:, :, i, :] # [batch, intermediade_size, ssm_state]
scan_output = torch.matmul(ssm_state.to(dtype), C[:, i, :].unsqueeze(-1)) # [batch, intermediade_size, 1]
scan_outputs.append(scan_output[:, :, 0])
scan_output = torch.stack(scan_outputs, dim=-1) # [batch, intermediade_size, seq_len]
scan_output = scan_output + (hidden_states * self.D[None, :, None])
scan_output = (scan_output * self.act(gate))
if use_cache:
cache_params.ssm_states[self.layer_idx] = ssm_state
# 4. Final linear projection
contextualized_states = self.out_proj(scan_output.transpose(1, 2)) # [batch, seq_len, hidden_size]
return contextualized_states
# fmt: on
def forward(self, hidden_states, cache_params: HybridMambaAttentionDynamicCache = None):
if self.use_fast_kernels:
if not is_fast_path_available or "cuda" not in self.x_proj.weight.device.type:
raise ValueError(
"Fast Mamba kernels are not available. Make sure to they are installed and that the mamba module is on a CUDA device"
)
return self.cuda_kernels_forward(hidden_states, cache_params)
return self.slow_forward(hidden_states, cache_params)
# Copied from transformers.models.mistral.modeling_mistral.MistralMLP with Mistral->Jamba
class JambaMLP(nn.Module):
def __init__(self, config):
super().__init__()
self.config = config
self.hidden_size = config.hidden_size
self.intermediate_size = config.intermediate_size
self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
self.act_fn = ACT2FN[config.hidden_act]
def forward(self, x):
return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
# Adapted from transformers.models.mixtral.modeling_mixtral.MixtralSparseMoeBlock with Mistral->Jamba