File size: 19,016 Bytes
1ce5e18
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# coding=utf-8
# Copyright 2021 The Facebook, Inc. and The HuggingFace Inc. team. All rights reserved.
#
# 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.
""" Blenderbot model configuration"""

from collections import OrderedDict
from typing import Any, Mapping, Optional

from ... import PreTrainedTokenizer
from ...configuration_utils import PretrainedConfig
from ...file_utils import TensorType, is_torch_available
from ...onnx import OnnxConfig, OnnxConfigWithPast, OnnxSeq2SeqConfigWithPast
from ...onnx.utils import compute_effective_axis_dimension
from ...utils import logging


logger = logging.get_logger(__name__)

BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP = {
    "facebook/blenderbot-3B": "https://huggingface.co/facebook/blenderbot-3B/resolve/main/config.json",
    # See all Blenderbot models at https://huggingface.co/models?filter=blenderbot
}


class BlenderbotConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`BlenderbotModel`]. It is used to instantiate an
    Blenderbot model according to the specified arguments, defining the model architecture. Instantiating a
    configuration with the defaults will yield a similar configuration to that of the Blenderbot
    [facebook/blenderbot-3B](https://huggingface.co/facebook/blenderbot-3B) architecture.

    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
    documentation from [`PretrainedConfig`] for more information.


    Args:
        vocab_size (`int`, *optional*, defaults to 50265):
            Vocabulary size of the Blenderbot model. Defines the number of different tokens that can be represented by
            the `inputs_ids` passed when calling [`BlenderbotModel`] or [`TFBlenderbotModel`].
        d_model (`int`, *optional*, defaults to 1024):
            Dimensionality of the layers and the pooler layer.
        encoder_layers (`int`, *optional*, defaults to 12):
            Number of encoder layers.
        decoder_layers (`int`, *optional*, defaults to 12):
            Number of decoder layers.
        encoder_attention_heads (`int`, *optional*, defaults to 16):
            Number of attention heads for each attention layer in the Transformer encoder.
        decoder_attention_heads (`int`, *optional*, defaults to 16):
            Number of attention heads for each attention layer in the Transformer decoder.
        decoder_ffn_dim (`int`, *optional*, defaults to 4096):
            Dimensionality of the "intermediate" (often named feed-forward) layer in decoder.
        encoder_ffn_dim (`int`, *optional*, defaults to 4096):
            Dimensionality of the "intermediate" (often named feed-forward) layer in decoder.
        activation_function (`str` or `function`, *optional*, defaults to `"gelu"`):
            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
            `"relu"`, `"silu"` and `"gelu_new"` are supported.
        dropout (`float`, *optional*, defaults to 0.1):
            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            The dropout ratio for the attention probabilities.
        activation_dropout (`float`, *optional*, defaults to 0.0):
            The dropout ratio for activations inside the fully connected layer.
        max_position_embeddings (`int`, *optional*, defaults to 128):
            The maximum sequence length that this model might ever be used with. Typically set this to something large
            just in case (e.g., 512 or 1024 or 2048).
        init_std (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        encoder_layerdrop (`float`, *optional*, defaults to 0.0):
            The LayerDrop probability for the encoder. See the [LayerDrop paper](see https://arxiv.org/abs/1909.11556)
            for more details.
        decoder_layerdrop (`float`, *optional*, defaults to 0.0):
            The LayerDrop probability for the decoder. See the [LayerDrop paper](see https://arxiv.org/abs/1909.11556)
            for more details.
        scale_embedding (`bool`, *optional*, defaults to `False`):
            Scale embeddings by diving by sqrt(d_model).
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether or not the model should return the last key/values attentions (not used by all models)
        forced_eos_token_id (`int`, *optional*, defaults to 2):
            The id of the token to force as the last generated token when `max_length` is reached. Usually set to
            `eos_token_id`.

    Example:

    ```python
    >>> from transformers import BlenderbotConfig, BlenderbotModel

    >>> # Initializing a Blenderbot facebook/blenderbot-3B style configuration
    >>> configuration = BlenderbotConfig()

    >>> # Initializing a model (with random weights) from the facebook/blenderbot-3B style configuration
    >>> model = BlenderbotModel(configuration)

    >>> # Accessing the model configuration
    >>> configuration = model.config
    ```"""
    model_type = "blenderbot"
    keys_to_ignore_at_inference = ["past_key_values"]
    attribute_map = {"num_attention_heads": "encoder_attention_heads", "hidden_size": "d_model"}

    def __init__(
        self,
        vocab_size=8008,
        max_position_embeddings=128,
        encoder_layers=2,
        encoder_ffn_dim=10240,
        encoder_attention_heads=32,
        decoder_layers=24,
        decoder_ffn_dim=10240,
        decoder_attention_heads=32,
        encoder_layerdrop=0.0,
        decoder_layerdrop=0.0,
        use_cache=True,
        is_encoder_decoder=True,
        activation_function="gelu",
        d_model=2560,
        dropout=0.1,
        attention_dropout=0.0,
        activation_dropout=0.0,
        init_std=0.02,
        decoder_start_token_id=1,
        scale_embedding=False,
        pad_token_id=0,
        bos_token_id=1,
        eos_token_id=2,
        encoder_no_repeat_ngram_size=3,
        forced_eos_token_id=2,
        **kwargs,
    ):
        self.vocab_size = vocab_size
        self.max_position_embeddings = max_position_embeddings
        self.d_model = d_model
        self.encoder_ffn_dim = encoder_ffn_dim
        self.encoder_layers = encoder_layers
        self.encoder_attention_heads = encoder_attention_heads
        self.decoder_ffn_dim = decoder_ffn_dim
        self.decoder_layers = decoder_layers
        self.decoder_attention_heads = decoder_attention_heads
        self.dropout = dropout
        self.attention_dropout = attention_dropout
        self.activation_dropout = activation_dropout
        self.activation_function = activation_function
        self.init_std = init_std
        self.encoder_layerdrop = encoder_layerdrop
        self.decoder_layerdrop = decoder_layerdrop
        self.use_cache = use_cache
        self.num_hidden_layers = encoder_layers
        self.scale_embedding = scale_embedding  # scale factor will be sqrt(d_model) if True

        super().__init__(
            pad_token_id=pad_token_id,
            bos_token_id=bos_token_id,
            eos_token_id=eos_token_id,
            is_encoder_decoder=is_encoder_decoder,
            decoder_start_token_id=decoder_start_token_id,
            encoder_no_repeat_ngram_size=encoder_no_repeat_ngram_size,
            forced_eos_token_id=forced_eos_token_id,
            **kwargs,
        )


class BlenderbotOnnxConfig(OnnxSeq2SeqConfigWithPast):
    @property
    def inputs(self) -> Mapping[str, Mapping[int, str]]:
        if self.task in ["default", "seq2seq-lm"]:
            common_inputs = OrderedDict(
                [
                    ("input_ids", {0: "batch", 1: "encoder_sequence"}),
                    ("attention_mask", {0: "batch", 1: "encoder_sequence"}),
                ]
            )
            if self.use_past:
                common_inputs["decoder_input_ids"] = {0: "batch"}
                common_inputs["decoder_attention_mask"] = {0: "batch", 1: "past_decoder_sequence + sequence"}
            else:
                common_inputs["decoder_input_ids"] = {0: "batch", 1: "decoder_sequence"}
                common_inputs["decoder_attention_mask"] = {0: "batch", 1: "decoder_sequence"}
            if self.use_past:
                self.fill_with_past_key_values_(common_inputs, direction="inputs")
        elif self.task == "causal-lm":
            common_inputs = OrderedDict(
                [
                    ("input_ids", {0: "batch", 1: "encoder_sequence"}),
                    ("attention_mask", {0: "batch", 1: "encoder_sequence"}),
                ]
            )
            if self.use_past:
                _, num_decoder_layers = self.num_layers
                for i in range(num_decoder_layers):
                    common_inputs[f"past_key_values.{i}.key"] = {0: "batch", 2: "past_sequence + sequence"}
                    common_inputs[f"past_key_values.{i}.value"] = {0: "batch", 2: "past_sequence + sequence"}
        else:
            common_inputs = OrderedDict(
                [
                    ("input_ids", {0: "batch", 1: "encoder_sequence"}),
                    ("attention_mask", {0: "batch", 1: "encoder_sequence"}),
                    ("decoder_input_ids", {0: "batch", 1: "decoder_sequence"}),
                    ("decoder_attention_mask", {0: "batch", 1: "decoder_sequence"}),
                ]
            )

        return common_inputs

    @property
    # Copied from transformers.models.bart.configuration_bart.BartOnnxConfig.outputs
    def outputs(self) -> Mapping[str, Mapping[int, str]]:
        if self.task in ["default", "seq2seq-lm"]:
            common_outputs = super().outputs
        else:
            common_outputs = super(OnnxConfigWithPast, self).outputs
            if self.use_past:
                num_encoder_layers, _ = self.num_layers
                for i in range(num_encoder_layers):
                    common_outputs[f"present.{i}.key"] = {0: "batch", 2: "past_sequence + sequence"}
                    common_outputs[f"present.{i}.value"] = {0: "batch", 2: "past_sequence + sequence"}
        return common_outputs

    def _generate_dummy_inputs_for_default_and_seq2seq_lm(
        self,
        tokenizer: PreTrainedTokenizer,
        batch_size: int = -1,
        seq_length: int = -1,
        is_pair: bool = False,
        framework: Optional[TensorType] = None,
    ) -> Mapping[str, Any]:
        encoder_inputs = self._generate_dummy_inputs_for_sequence_classification_and_question_answering(
            tokenizer, batch_size, seq_length, is_pair, framework
        )
        # Generate decoder inputs
        decoder_seq_length = seq_length if not self.use_past else 1
        decoder_inputs = self._generate_dummy_inputs_for_sequence_classification_and_question_answering(
            tokenizer, batch_size, decoder_seq_length, is_pair, framework
        )
        decoder_inputs = {f"decoder_{name}": tensor for name, tensor in decoder_inputs.items()}
        common_inputs = dict(**encoder_inputs, **decoder_inputs)

        if self.use_past:
            if not is_torch_available():
                raise ValueError("Cannot generate dummy past_keys inputs without PyTorch installed.")
            else:
                import torch
            batch, encoder_seq_length = common_inputs["input_ids"].shape
            decoder_seq_length = common_inputs["decoder_input_ids"].shape[1]
            num_encoder_attention_heads, num_decoder_attention_heads = self.num_attention_heads
            encoder_shape = (
                batch,
                num_encoder_attention_heads,
                encoder_seq_length,
                self._config.hidden_size // num_encoder_attention_heads,
            )
            decoder_past_length = decoder_seq_length
            decoder_shape = (
                batch,
                num_decoder_attention_heads,
                decoder_past_length,
                self._config.hidden_size // num_decoder_attention_heads,
            )
            common_inputs["decoder_attention_mask"] = torch.cat(
                [common_inputs["decoder_attention_mask"], torch.ones(batch, decoder_past_length)], dim=1
            )
            common_inputs["past_key_values"] = []
            _, num_decoder_layers = self.num_layers

            for _ in range(num_decoder_layers):
                common_inputs["past_key_values"].append(
                    (
                        torch.zeros(decoder_shape),
                        torch.zeros(decoder_shape),
                        torch.zeros(encoder_shape),
                        torch.zeros(encoder_shape),
                    )
                )
        return common_inputs

    def _generate_dummy_inputs_for_causal_lm(
        self,
        tokenizer: PreTrainedTokenizer,
        batch_size: int = -1,
        seq_length: int = -1,
        is_pair: bool = False,
        framework: Optional[TensorType] = None,
    ) -> Mapping[str, Any]:
        common_inputs = self._generate_dummy_inputs_for_sequence_classification_and_question_answering(
            tokenizer, batch_size, seq_length, is_pair, framework
        )

        if self.use_past:
            if not is_torch_available():
                raise ValueError("Cannot generate dummy past_keys inputs without PyTorch installed.")
            else:
                import torch
            batch, seqlen = common_inputs["input_ids"].shape
            past_key_values_length = seqlen
            _, num_decoder_layers = self.num_layers
            num_encoder_attention_heads, _ = self.num_attention_heads
            past_shape = (
                batch,
                num_encoder_attention_heads,
                past_key_values_length,
                self._config.hidden_size // num_encoder_attention_heads,
            )
            mask_dtype = common_inputs["attention_mask"].dtype
            common_inputs["attention_mask"] = torch.cat(
                [common_inputs["attention_mask"], torch.ones(batch, past_key_values_length, dtype=mask_dtype)], dim=1
            )
            common_inputs["past_key_values"] = [
                (torch.zeros(past_shape), torch.zeros(past_shape)) for _ in range(num_decoder_layers)
            ]
        return common_inputs

    # Copied from transformers.models.bart.configuration_bart.BartOnnxConfig._generate_dummy_inputs_for_sequence_classification_and_question_answering
    def _generate_dummy_inputs_for_sequence_classification_and_question_answering(
        self,
        tokenizer: PreTrainedTokenizer,
        batch_size: int = -1,
        seq_length: int = -1,
        is_pair: bool = False,
        framework: Optional[TensorType] = None,
    ) -> Mapping[str, Any]:
        # Copied from OnnxConfig.generate_dummy_inputs
        # Did not use super(OnnxConfigWithPast, self).generate_dummy_inputs for code clarity.
        # If dynamic axis (-1) we forward with a fixed dimension of 2 samples to avoid optimizations made by ONNX
        batch_size = compute_effective_axis_dimension(
            batch_size, fixed_dimension=OnnxConfig.default_fixed_batch, num_token_to_add=0
        )

        # If dynamic axis (-1) we forward with a fixed dimension of 8 tokens to avoid optimizations made by ONNX
        token_to_add = tokenizer.num_special_tokens_to_add(is_pair)
        seq_length = compute_effective_axis_dimension(
            seq_length, fixed_dimension=OnnxConfig.default_fixed_sequence, num_token_to_add=token_to_add
        )

        # Generate dummy inputs according to compute batch and sequence
        dummy_input = [" ".join([tokenizer.unk_token]) * seq_length] * batch_size
        common_inputs = dict(tokenizer(dummy_input, return_tensors=framework))
        return common_inputs

    # Copied from transformers.models.bart.configuration_bart.BartOnnxConfig.generate_dummy_inputs
    def generate_dummy_inputs(
        self,
        tokenizer: PreTrainedTokenizer,
        batch_size: int = -1,
        seq_length: int = -1,
        is_pair: bool = False,
        framework: Optional[TensorType] = None,
    ) -> Mapping[str, Any]:
        if self.task in ["default", "seq2seq-lm"]:
            common_inputs = self._generate_dummy_inputs_for_default_and_seq2seq_lm(
                tokenizer, batch_size=batch_size, seq_length=seq_length, is_pair=is_pair, framework=framework
            )

        elif self.task == "causal-lm":
            common_inputs = self._generate_dummy_inputs_for_causal_lm(
                tokenizer, batch_size=batch_size, seq_length=seq_length, is_pair=is_pair, framework=framework
            )
        else:
            common_inputs = self._generate_dummy_inputs_for_sequence_classification_and_question_answering(
                tokenizer, batch_size=batch_size, seq_length=seq_length, is_pair=is_pair, framework=framework
            )

        return common_inputs

    # Copied from transformers.models.bart.configuration_bart.BartOnnxConfig._flatten_past_key_values_
    def _flatten_past_key_values_(self, flattened_output, name, idx, t):
        if self.task in ["default", "seq2seq-lm"]:
            flattened_output = super()._flatten_past_key_values_(flattened_output, name, idx, t)
        else:
            flattened_output = super(OnnxSeq2SeqConfigWithPast, self)._flatten_past_key_values_(
                flattened_output, name, idx, t
            )

    def fill_with_past_key_values_(self, inputs_or_outputs: Mapping[str, Mapping[int, str]], direction: str):
        if direction not in ["inputs", "outputs"]:
            raise ValueError(f'direction must either be "inputs" or "outputs", but {direction} was given')

        name = "past_key_values" if direction == "inputs" else "present"
        _, num_decoder_layers = self.num_layers

        encoder_sequence = "past_encoder_sequence"
        decoder_sequence = "past_decoder_sequence" if direction == "inputs" else "past_decoder_sequence + sequence"

        for i in range(num_decoder_layers):
            inputs_or_outputs[f"{name}.{i}.decoder.key"] = {0: "batch", 2: decoder_sequence}
            inputs_or_outputs[f"{name}.{i}.decoder.value"] = {0: "batch", 2: decoder_sequence}
            inputs_or_outputs[f"{name}.{i}.encoder.key"] = {0: "batch", 2: encoder_sequence}
            inputs_or_outputs[f"{name}.{i}.encoder.value"] = {0: "batch", 2: encoder_sequence}