File size: 9,053 Bytes
e0609f8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import warnings
from typing import List, Optional, Union

import torch
import torch.distributed as dist
from torch import nn
from transformers import BatchEncoding
from transformers.generation.logits_process import (
    LogitsProcessorList,
)
from transformers.generation.stopping_criteria import (
    StoppingCriteriaList,
    validate_stopping_criteria,
)

from transformers.generation.utils import SampleOutput, SampleEncoderDecoderOutput, SampleDecoderOnlyOutput

def sample(
    self,
    input_ids: torch.LongTensor,
    logits_processor: Optional[LogitsProcessorList] = None,
    stopping_criteria: Optional[StoppingCriteriaList] = None,
    logits_warper: Optional[LogitsProcessorList] = None,
    max_length: Optional[int] = None,
    pad_token_id: Optional[int] = None,
    eos_token_id: Optional[Union[int, List[int]]] = None,
    output_attentions: Optional[bool] = None,
    output_hidden_states: Optional[bool] = None,
    output_scores: Optional[bool] = None,
    return_dict_in_generate: Optional[bool] = None,
    synced_gpus: Optional[bool] = False,
    **model_kwargs,
) -> Union[SampleOutput, torch.LongTensor]:

    if type(input_ids) in [dict, BatchEncoding]:
        input_ids, ngram_sequences = input_ids["input_ids"], input_ids
        del ngram_sequences["input_ids"]
        del ngram_sequences["attention_mask"]
    else:
        ngram_sequences = {}

    # init values
    logits_processor = logits_processor if logits_processor is not None else LogitsProcessorList()
    stopping_criteria = stopping_criteria if stopping_criteria is not None else StoppingCriteriaList()
    if max_length is not None:
        warnings.warn(
            "`max_length` is deprecated in this function, use"
            " `stopping_criteria=StoppingCriteriaList(MaxLengthCriteria(max_length=max_length))` instead.",
            UserWarning,
        )
        stopping_criteria = validate_stopping_criteria(stopping_criteria, max_length)
    logits_warper = logits_warper if logits_warper is not None else LogitsProcessorList()
    pad_token_id = pad_token_id if pad_token_id is not None else self.generation_config.pad_token_id
    eos_token_id = eos_token_id if eos_token_id is not None else self.generation_config.eos_token_id
    if isinstance(eos_token_id, int):
        eos_token_id = [eos_token_id]

    eos_token_id_tensor = torch.tensor(eos_token_id).to(input_ids.device) if eos_token_id is not None else None
    output_scores = output_scores if output_scores is not None else self.generation_config.output_scores
    output_attentions = (
        output_attentions if output_attentions is not None else self.generation_config.output_attentions
    )
    output_hidden_states = (
        output_hidden_states if output_hidden_states is not None else self.generation_config.output_hidden_states
    )
    return_dict_in_generate = (
        return_dict_in_generate
        if return_dict_in_generate is not None
        else self.generation_config.return_dict_in_generate
    )

    # init attention / hidden states / scores tuples
    scores = () if (return_dict_in_generate and output_scores) else None
    decoder_attentions = () if (return_dict_in_generate and output_attentions) else None
    cross_attentions = () if (return_dict_in_generate and output_attentions) else None
    decoder_hidden_states = () if (return_dict_in_generate and output_hidden_states) else None

    # if model is an encoder-decoder, retrieve encoder attention weights and hidden states
    if return_dict_in_generate and self.config.is_encoder_decoder:
        encoder_attentions = model_kwargs["encoder_outputs"].get("attentions") if output_attentions else None
        encoder_hidden_states = (
            model_kwargs["encoder_outputs"].get("hidden_states") if output_hidden_states else None
        )

    # keep track of which sequences are already finished
    unfinished_sequences = torch.ones(input_ids.shape[0], dtype=torch.long, device=input_ids.device)

    this_peer_finished = False  # used by synced_gpus only
    # auto-regressive generation
    while True:
        if synced_gpus:
            # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
            # The following logic allows an early break if all peers finished generating their sequence
            this_peer_finished_flag = torch.tensor(0.0 if this_peer_finished else 1.0).to(input_ids.device)
            # send 0.0 if we finished, 1.0 otherwise
            dist.all_reduce(this_peer_finished_flag, op=dist.ReduceOp.SUM)
            # did all peers finish? the reduced sum will be 0.0 then
            if this_peer_finished_flag.item() == 0.0:
                break

        # prepare model inputs
        model_inputs = {"input_ids": input_ids}

        # forward pass to get next token
        outputs = self(
            **model_inputs,
            return_dict=True,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            **ngram_sequences
        )

        if synced_gpus and this_peer_finished:
            continue  # don't waste resources running the code we don't need

        next_token_logits = outputs.logits[:, -1, :]

        # pre-process distribution
        next_token_scores = logits_processor(input_ids, next_token_logits)
        next_token_scores = logits_warper(input_ids, next_token_scores)

        # Store scores, attentions and hidden_states when required
        if return_dict_in_generate:
            if output_scores:
                scores += (next_token_scores,)
            if output_attentions:
                decoder_attentions += (
                    (outputs.decoder_attentions,) if self.config.is_encoder_decoder else (outputs.attentions,)
                )
                if self.config.is_encoder_decoder:
                    cross_attentions += (outputs.cross_attentions,)

            if output_hidden_states:
                decoder_hidden_states += (
                    (outputs.decoder_hidden_states,)
                    if self.config.is_encoder_decoder
                    else (outputs.hidden_states,)
                )

        # sample
        probs = nn.functional.softmax(next_token_scores, dim=-1)
        next_tokens = torch.multinomial(probs, num_samples=1).squeeze(1)

        # finished sentences should have their next token be a padding token
        if eos_token_id is not None:
            if pad_token_id is None:
                raise ValueError("If `eos_token_id` is defined, make sure that `pad_token_id` is defined.")
            next_tokens = next_tokens * unfinished_sequences + pad_token_id * (1 - unfinished_sequences)

        # update generated ids, model inputs, and length for next step
        input_ids = torch.cat([input_ids, next_tokens[:, None]], dim=-1)
        decoded = self.tokenizer.batch_decode(input_ids)[0]
        encoded = self.tokenizer(
            decoded, return_tensors="pt", return_ngram_sequences=True
        )
        input_ids = encoded.input_ids.to(self.device)

        ngram_sequences = {}

        if "label_gram_2_sequence" in encoded:
            ngram_sequences["label_gram_2_sequence"] = encoded["label_gram_2_sequence"].to(self.device)

        if "label_gram_3_sequence" in encoded:
            ngram_sequences["label_gram_3_sequence"] = encoded["label_gram_3_sequence"].to(self.device)

        if "label_gram_4_sequence" in encoded:
            ngram_sequences["label_gram_4_sequence"] = encoded["label_gram_4_sequence"].to(self.device)

        model_kwargs = self._update_model_kwargs_for_generation(
            outputs, model_kwargs, is_encoder_decoder=self.config.is_encoder_decoder
        )

        # if eos_token was found in one sentence, set sentence to finished
        if eos_token_id_tensor is not None:
            unfinished_sequences = unfinished_sequences.mul(
                next_tokens.tile(eos_token_id_tensor.shape[0], 1).ne(eos_token_id_tensor.unsqueeze(1)).prod(dim=0)
            )

        # stop when each sentence is finished, or if we exceed the maximum length
        if unfinished_sequences.max() == 0 or stopping_criteria(input_ids, scores):
            if not synced_gpus:
                break
            else:
                this_peer_finished = True

    if return_dict_in_generate:
        if self.config.is_encoder_decoder:
            return SampleEncoderDecoderOutput(
                sequences=input_ids,
                scores=scores,
                encoder_attentions=encoder_attentions,
                encoder_hidden_states=encoder_hidden_states,
                decoder_attentions=decoder_attentions,
                cross_attentions=cross_attentions,
                decoder_hidden_states=decoder_hidden_states,
            )
        else:
            return SampleDecoderOnlyOutput(
                sequences=input_ids,
                scores=scores,
                attentions=decoder_attentions,
                hidden_states=decoder_hidden_states,
            )
    else:
        return input_ids