1528 lines
67 KiB
Python
1528 lines
67 KiB
Python
# coding=utf-8
|
|
# Copyright 2023 Meta AI Team and the HuggingFace Inc. team.
|
|
#
|
|
# 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 X-MOD model."""
|
|
|
|
import math
|
|
from typing import Optional, Union
|
|
|
|
import torch
|
|
import torch.utils.checkpoint
|
|
from torch import nn
|
|
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
|
|
|
|
from ...activations import ACT2FN, gelu
|
|
from ...cache_utils import Cache, EncoderDecoderCache
|
|
from ...generation import GenerationMixin
|
|
from ...modeling_layers import GradientCheckpointingLayer
|
|
from ...modeling_outputs import (
|
|
BaseModelOutputWithPastAndCrossAttentions,
|
|
BaseModelOutputWithPoolingAndCrossAttentions,
|
|
CausalLMOutputWithCrossAttentions,
|
|
MaskedLMOutput,
|
|
MultipleChoiceModelOutput,
|
|
QuestionAnsweringModelOutput,
|
|
SequenceClassifierOutput,
|
|
TokenClassifierOutput,
|
|
)
|
|
from ...modeling_utils import PreTrainedModel
|
|
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
|
|
from ...utils import auto_docstring, logging
|
|
from ...utils.deprecation import deprecate_kwarg
|
|
from .configuration_xmod import XmodConfig
|
|
|
|
|
|
logger = logging.get_logger(__name__)
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaEmbeddings with Roberta->Xmod
|
|
class XmodEmbeddings(nn.Module):
|
|
"""
|
|
Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
|
|
"""
|
|
|
|
# Copied from transformers.models.bert.modeling_bert.BertEmbeddings.__init__
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
|
|
self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
|
|
self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
|
|
|
|
# self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
|
|
# any TensorFlow checkpoint file
|
|
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
|
|
self.dropout = nn.Dropout(config.hidden_dropout_prob)
|
|
# position_ids (1, len position emb) is contiguous in memory and exported when serialized
|
|
self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
|
|
self.register_buffer(
|
|
"position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
|
|
)
|
|
self.register_buffer(
|
|
"token_type_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
|
|
)
|
|
|
|
# End copy
|
|
self.padding_idx = config.pad_token_id
|
|
self.position_embeddings = nn.Embedding(
|
|
config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
|
|
)
|
|
|
|
def forward(
|
|
self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, past_key_values_length=0
|
|
):
|
|
if position_ids is None:
|
|
if input_ids is not None:
|
|
# Create the position ids from the input token ids. Any padded tokens remain padded.
|
|
position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
|
|
else:
|
|
position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)
|
|
|
|
if input_ids is not None:
|
|
input_shape = input_ids.size()
|
|
else:
|
|
input_shape = inputs_embeds.size()[:-1]
|
|
|
|
seq_length = input_shape[1]
|
|
|
|
# Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
|
|
# when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
|
|
# issue #5664
|
|
if token_type_ids is None:
|
|
if hasattr(self, "token_type_ids"):
|
|
buffered_token_type_ids = self.token_type_ids[:, :seq_length]
|
|
buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
|
|
token_type_ids = buffered_token_type_ids_expanded
|
|
else:
|
|
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)
|
|
|
|
if inputs_embeds is None:
|
|
inputs_embeds = self.word_embeddings(input_ids)
|
|
token_type_embeddings = self.token_type_embeddings(token_type_ids)
|
|
|
|
embeddings = inputs_embeds + token_type_embeddings
|
|
if self.position_embedding_type == "absolute":
|
|
position_embeddings = self.position_embeddings(position_ids)
|
|
embeddings += position_embeddings
|
|
embeddings = self.LayerNorm(embeddings)
|
|
embeddings = self.dropout(embeddings)
|
|
return embeddings
|
|
|
|
def create_position_ids_from_inputs_embeds(self, inputs_embeds):
|
|
"""
|
|
We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
|
|
|
|
Args:
|
|
inputs_embeds: torch.Tensor
|
|
|
|
Returns: torch.Tensor
|
|
"""
|
|
input_shape = inputs_embeds.size()[:-1]
|
|
sequence_length = input_shape[1]
|
|
|
|
position_ids = torch.arange(
|
|
self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
|
|
)
|
|
return position_ids.unsqueeze(0).expand(input_shape)
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaSelfAttention with Roberta->Xmod
|
|
class XmodSelfAttention(nn.Module):
|
|
def __init__(self, config, position_embedding_type=None, layer_idx=None):
|
|
super().__init__()
|
|
if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
|
|
raise ValueError(
|
|
f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
|
|
f"heads ({config.num_attention_heads})"
|
|
)
|
|
|
|
self.num_attention_heads = config.num_attention_heads
|
|
self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
|
|
self.all_head_size = self.num_attention_heads * self.attention_head_size
|
|
|
|
self.query = nn.Linear(config.hidden_size, self.all_head_size)
|
|
self.key = nn.Linear(config.hidden_size, self.all_head_size)
|
|
self.value = nn.Linear(config.hidden_size, self.all_head_size)
|
|
|
|
self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
|
|
self.position_embedding_type = position_embedding_type or getattr(
|
|
config, "position_embedding_type", "absolute"
|
|
)
|
|
if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
|
|
self.max_position_embeddings = config.max_position_embeddings
|
|
self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size)
|
|
|
|
self.is_decoder = config.is_decoder
|
|
self.layer_idx = layer_idx
|
|
|
|
@deprecate_kwarg("encoder_attention_mask", version="4.55.0")
|
|
def forward(
|
|
self,
|
|
hidden_states: torch.Tensor,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
encoder_hidden_states: Optional[torch.FloatTensor] = None,
|
|
encoder_attention_mask: Optional[torch.FloatTensor] = None,
|
|
past_key_value: Optional[Cache] = None,
|
|
output_attentions: Optional[bool] = False,
|
|
cache_position: Optional[torch.Tensor] = None,
|
|
) -> tuple[torch.Tensor]:
|
|
batch_size, seq_length, _ = hidden_states.shape
|
|
query_layer = self.query(hidden_states)
|
|
query_layer = query_layer.view(batch_size, -1, self.num_attention_heads, self.attention_head_size).transpose(
|
|
1, 2
|
|
)
|
|
|
|
# If this is instantiated as a cross-attention module, the keys
|
|
# and values come from an encoder; the attention mask needs to be
|
|
# such that the encoder's padding tokens are not attended to.
|
|
is_cross_attention = encoder_hidden_states is not None
|
|
if is_cross_attention and encoder_attention_mask is not None:
|
|
attention_mask = encoder_attention_mask
|
|
|
|
if past_key_value is not None:
|
|
if isinstance(past_key_value, EncoderDecoderCache):
|
|
is_updated = past_key_value.is_updated.get(self.layer_idx)
|
|
if is_cross_attention:
|
|
# after the first generated id, we can subsequently re-use all key/value_layer from cache
|
|
curr_past_key_value = past_key_value.cross_attention_cache
|
|
else:
|
|
curr_past_key_value = past_key_value.self_attention_cache
|
|
else:
|
|
curr_past_key_value = past_key_value
|
|
|
|
current_states = encoder_hidden_states if is_cross_attention else hidden_states
|
|
if is_cross_attention and past_key_value is not None and is_updated:
|
|
# reuse k,v, cross_attentions
|
|
key_layer = curr_past_key_value.key_cache[self.layer_idx]
|
|
value_layer = curr_past_key_value.value_cache[self.layer_idx]
|
|
else:
|
|
key_layer = self.key(current_states)
|
|
key_layer = key_layer.view(batch_size, -1, self.num_attention_heads, self.attention_head_size).transpose(
|
|
1, 2
|
|
)
|
|
value_layer = self.value(current_states)
|
|
value_layer = value_layer.view(
|
|
batch_size, -1, self.num_attention_heads, self.attention_head_size
|
|
).transpose(1, 2)
|
|
|
|
if past_key_value is not None:
|
|
# save all key/value_layer to cache to be re-used for fast auto-regressive generation
|
|
cache_position = cache_position if not is_cross_attention else None
|
|
key_layer, value_layer = curr_past_key_value.update(
|
|
key_layer, value_layer, self.layer_idx, {"cache_position": cache_position}
|
|
)
|
|
# set flag that curr layer for cross-attn is already updated so we can re-use in subsequent calls
|
|
if is_cross_attention:
|
|
past_key_value.is_updated[self.layer_idx] = True
|
|
|
|
# Take the dot product between "query" and "key" to get the raw attention scores.
|
|
attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
|
|
|
|
if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
|
|
query_length, key_length = query_layer.shape[2], key_layer.shape[2]
|
|
if past_key_value is not None:
|
|
position_ids_l = torch.tensor(key_length - 1, dtype=torch.long, device=hidden_states.device).view(
|
|
-1, 1
|
|
)
|
|
else:
|
|
position_ids_l = torch.arange(query_length, dtype=torch.long, device=hidden_states.device).view(-1, 1)
|
|
position_ids_r = torch.arange(key_length, dtype=torch.long, device=hidden_states.device).view(1, -1)
|
|
distance = position_ids_l - position_ids_r
|
|
|
|
positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
|
|
positional_embedding = positional_embedding.to(dtype=query_layer.dtype) # fp16 compatibility
|
|
|
|
if self.position_embedding_type == "relative_key":
|
|
relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
|
|
attention_scores = attention_scores + relative_position_scores
|
|
elif self.position_embedding_type == "relative_key_query":
|
|
relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
|
|
relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
|
|
attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key
|
|
|
|
attention_scores = attention_scores / math.sqrt(self.attention_head_size)
|
|
if attention_mask is not None:
|
|
# Apply the attention mask is (precomputed for all layers in XmodModel forward() function)
|
|
attention_scores = attention_scores + attention_mask
|
|
|
|
# Normalize the attention scores to probabilities.
|
|
attention_probs = nn.functional.softmax(attention_scores, dim=-1)
|
|
|
|
# This is actually dropping out entire tokens to attend to, which might
|
|
# seem a bit unusual, but is taken from the original Transformer paper.
|
|
attention_probs = self.dropout(attention_probs)
|
|
|
|
# Mask heads if we want to
|
|
if head_mask is not None:
|
|
attention_probs = attention_probs * head_mask
|
|
|
|
context_layer = torch.matmul(attention_probs, value_layer)
|
|
|
|
context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
|
|
new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
|
|
context_layer = context_layer.view(new_context_layer_shape)
|
|
|
|
return context_layer, attention_probs
|
|
|
|
|
|
class XmodSelfOutput(nn.Module):
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaSelfOutput.__init__
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
|
|
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
|
|
self.dropout = nn.Dropout(config.hidden_dropout_prob)
|
|
|
|
def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
|
|
hidden_states = self.dense(hidden_states)
|
|
hidden_states = self.dropout(hidden_states)
|
|
hidden_states = hidden_states + input_tensor
|
|
return hidden_states
|
|
|
|
|
|
class XmodAttention(nn.Module):
|
|
def __init__(self, config, position_embedding_type=None, layer_idx=None):
|
|
super().__init__()
|
|
self.self = XmodSelfAttention(config, position_embedding_type=position_embedding_type, layer_idx=layer_idx)
|
|
self.output = XmodSelfOutput(config)
|
|
self.pruned_heads = set()
|
|
self.pre_norm = config.pre_norm
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaAttention.prune_heads
|
|
def prune_heads(self, heads):
|
|
if len(heads) == 0:
|
|
return
|
|
heads, index = find_pruneable_heads_and_indices(
|
|
heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
|
|
)
|
|
|
|
# Prune linear layers
|
|
self.self.query = prune_linear_layer(self.self.query, index)
|
|
self.self.key = prune_linear_layer(self.self.key, index)
|
|
self.self.value = prune_linear_layer(self.self.value, index)
|
|
self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)
|
|
|
|
# Update hyper params and store pruned heads
|
|
self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
|
|
self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
|
|
self.pruned_heads = self.pruned_heads.union(heads)
|
|
|
|
@deprecate_kwarg("encoder_attention_mask", version="4.55.0")
|
|
def forward(
|
|
self,
|
|
hidden_states: torch.Tensor,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
encoder_hidden_states: Optional[torch.FloatTensor] = None,
|
|
encoder_attention_mask: Optional[torch.FloatTensor] = None,
|
|
past_key_value: Optional[Cache] = None,
|
|
output_attentions: Optional[bool] = False,
|
|
cache_position: Optional[torch.Tensor] = None,
|
|
) -> tuple[torch.Tensor]:
|
|
residual = hidden_states
|
|
if self.pre_norm:
|
|
hidden_states = self.output.LayerNorm(hidden_states)
|
|
self_outputs = self.self(
|
|
hidden_states,
|
|
attention_mask,
|
|
head_mask,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
past_key_value,
|
|
output_attentions,
|
|
cache_position,
|
|
)
|
|
attention_output = self.output(self_outputs[0], residual)
|
|
if not self.pre_norm:
|
|
attention_output = self.output.LayerNorm(attention_output)
|
|
outputs = (attention_output,) + self_outputs[1:] # add attentions if we output them
|
|
return outputs
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaIntermediate
|
|
class XmodIntermediate(nn.Module):
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
|
|
if isinstance(config.hidden_act, str):
|
|
self.intermediate_act_fn = ACT2FN[config.hidden_act]
|
|
else:
|
|
self.intermediate_act_fn = config.hidden_act
|
|
|
|
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
|
|
hidden_states = self.dense(hidden_states)
|
|
hidden_states = self.intermediate_act_fn(hidden_states)
|
|
return hidden_states
|
|
|
|
|
|
class XmodAdapter(nn.Module):
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.bottleneck_size = config.hidden_size // config.adapter_reduction_factor
|
|
self.dense1 = nn.Linear(config.hidden_size, self.bottleneck_size)
|
|
self.dense2 = nn.Linear(self.bottleneck_size, config.hidden_size)
|
|
if isinstance(config.hidden_act, str):
|
|
self.adapter_act_fn = ACT2FN[config.hidden_act]
|
|
else:
|
|
self.adapter_act_fn = config.hidden_act
|
|
|
|
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
|
|
hidden_states = self.dense1(hidden_states)
|
|
hidden_states = self.adapter_act_fn(hidden_states)
|
|
hidden_states = self.dense2(hidden_states)
|
|
return hidden_states
|
|
|
|
|
|
class XmodOutput(nn.Module):
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
|
|
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
|
|
self.ln_before_adapter = config.ln_before_adapter
|
|
self.dropout = nn.Dropout(config.hidden_dropout_prob)
|
|
if config.adapter_layer_norm:
|
|
self.adapter_layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
|
|
else:
|
|
self.adapter_layer_norm = None
|
|
self.adapter_reuse_layer_norm = config.adapter_reuse_layer_norm
|
|
self.adapter_modules = nn.ModuleDict({})
|
|
for language in config.languages:
|
|
self.adapter_modules[str(language)] = XmodAdapter(config)
|
|
|
|
def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor, lang_ids: torch.Tensor) -> torch.Tensor:
|
|
hidden_states = self.dense(hidden_states)
|
|
hidden_states = self.dropout(hidden_states)
|
|
hidden_states = hidden_states + input_tensor
|
|
hidden_states = self.lang_adapter(lang_ids, hidden_states)
|
|
return hidden_states
|
|
|
|
def lang_adapter(self, lang_ids: torch.Tensor, hidden_states: torch.Tensor):
|
|
# Process subsequent samples with the same lang_id in parallel
|
|
lang_ids, lang_lengths = torch.unique_consecutive(lang_ids, return_counts=True)
|
|
|
|
if not self.ln_before_adapter:
|
|
residual = hidden_states
|
|
|
|
if self.adapter_layer_norm is not None:
|
|
hidden_states = self.adapter_layer_norm(hidden_states)
|
|
elif self.adapter_reuse_layer_norm:
|
|
hidden_states = self.LayerNorm(hidden_states)
|
|
|
|
if self.ln_before_adapter:
|
|
residual = hidden_states
|
|
|
|
split_hidden_states = torch.split(hidden_states, lang_lengths.tolist(), 0)
|
|
lang_wise_outputs = []
|
|
for i, (lang_id, split_hidden_state) in enumerate(zip(lang_ids, split_hidden_states)):
|
|
lang = list(self.adapter_modules.keys())[int(lang_id.item())]
|
|
lang_wise_outputs.append(self.adapter_modules[lang](split_hidden_state))
|
|
hidden_states = torch.cat(lang_wise_outputs, 0)
|
|
|
|
hidden_states = self.dropout(hidden_states)
|
|
hidden_states += residual
|
|
return hidden_states
|
|
|
|
|
|
class XmodLayer(GradientCheckpointingLayer):
|
|
def __init__(self, config, layer_idx=None):
|
|
super().__init__()
|
|
self.chunk_size_feed_forward = config.chunk_size_feed_forward
|
|
self.seq_len_dim = 1
|
|
self.attention = XmodAttention(config, layer_idx=layer_idx)
|
|
self.is_decoder = config.is_decoder
|
|
self.add_cross_attention = config.add_cross_attention
|
|
if self.add_cross_attention:
|
|
if not self.is_decoder:
|
|
raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
|
|
self.crossattention = XmodAttention(config, position_embedding_type="absolute", layer_idx=layer_idx)
|
|
self.intermediate = XmodIntermediate(config)
|
|
self.output = XmodOutput(config)
|
|
self.pre_norm = config.pre_norm
|
|
|
|
def forward(
|
|
self,
|
|
hidden_states: torch.Tensor,
|
|
lang_ids: torch.Tensor,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
encoder_hidden_states: Optional[torch.FloatTensor] = None,
|
|
encoder_attention_mask: Optional[torch.FloatTensor] = None,
|
|
past_key_value: Optional[Cache] = None,
|
|
output_attentions: Optional[bool] = False,
|
|
cache_position: Optional[torch.Tensor] = None,
|
|
) -> tuple[torch.Tensor]:
|
|
self_attention_outputs = self.attention(
|
|
hidden_states,
|
|
attention_mask=attention_mask,
|
|
head_mask=head_mask,
|
|
output_attentions=output_attentions,
|
|
past_key_value=past_key_value,
|
|
cache_position=cache_position,
|
|
)
|
|
attention_output = self_attention_outputs[0]
|
|
outputs = self_attention_outputs[1:] # add self attentions if we output attention weights
|
|
|
|
if self.is_decoder and encoder_hidden_states is not None:
|
|
if not hasattr(self, "crossattention"):
|
|
raise ValueError(
|
|
f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
|
|
" by setting `config.add_cross_attention=True`"
|
|
)
|
|
|
|
cross_attention_outputs = self.crossattention(
|
|
attention_output,
|
|
attention_mask=encoder_attention_mask,
|
|
head_mask=head_mask,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
past_key_value=past_key_value,
|
|
output_attentions=output_attentions,
|
|
cache_position=cache_position,
|
|
)
|
|
attention_output = cross_attention_outputs[0]
|
|
outputs = outputs + cross_attention_outputs[1:] # add cross attentions if we output attention weights
|
|
|
|
residual = attention_output
|
|
if self.pre_norm:
|
|
attention_output = self.output.LayerNorm(attention_output)
|
|
intermediate_output = apply_chunking_to_forward(
|
|
self.feed_forward_chunk,
|
|
self.chunk_size_feed_forward,
|
|
self.seq_len_dim,
|
|
attention_output,
|
|
)
|
|
layer_output = self.output(intermediate_output, residual, lang_ids)
|
|
if not self.pre_norm:
|
|
layer_output = self.output.LayerNorm(layer_output)
|
|
return (layer_output,) + outputs
|
|
|
|
def feed_forward_chunk(self, attention_output):
|
|
return self.intermediate(attention_output)
|
|
|
|
|
|
class XmodEncoder(nn.Module):
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.config = config
|
|
self.layer = nn.ModuleList([XmodLayer(config, layer_idx=i) for i in range(config.num_hidden_layers)])
|
|
self.is_pre_norm = config.pre_norm
|
|
if self.is_pre_norm:
|
|
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
|
|
self.gradient_checkpointing = False
|
|
|
|
def forward(
|
|
self,
|
|
hidden_states: torch.Tensor,
|
|
lang_ids: torch.Tensor,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
encoder_hidden_states: Optional[torch.FloatTensor] = None,
|
|
encoder_attention_mask: Optional[torch.FloatTensor] = None,
|
|
past_key_values: Optional[tuple[tuple[torch.FloatTensor]]] = None,
|
|
use_cache: Optional[bool] = None,
|
|
output_attentions: Optional[bool] = False,
|
|
output_hidden_states: Optional[bool] = False,
|
|
return_dict: Optional[bool] = True,
|
|
cache_position: Optional[torch.Tensor] = None,
|
|
) -> Union[tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
|
|
if self.gradient_checkpointing and self.training:
|
|
if use_cache:
|
|
logger.warning_once(
|
|
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
|
|
)
|
|
use_cache = False
|
|
|
|
return_legacy_cache = False
|
|
if use_cache and not isinstance(past_key_values, Cache):
|
|
logger.warning_once(
|
|
"Passing a tuple of `past_key_values` is deprecated and will be removed in Transformers v4.58.0. "
|
|
"You should pass an instance of `EncoderDecoderCache` instead, e.g. "
|
|
"`past_key_values=EncoderDecoderCache.from_legacy_cache(past_key_values)`."
|
|
)
|
|
return_legacy_cache = True
|
|
past_key_values = EncoderDecoderCache.from_legacy_cache(past_key_values)
|
|
|
|
all_hidden_states = () if output_hidden_states else None
|
|
all_self_attentions = () if output_attentions else None
|
|
all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
|
|
|
|
for i, layer_module in enumerate(self.layer):
|
|
if output_hidden_states:
|
|
all_hidden_states = all_hidden_states + (hidden_states,)
|
|
|
|
layer_head_mask = head_mask[i] if head_mask is not None else None
|
|
|
|
layer_outputs = layer_module(
|
|
hidden_states,
|
|
lang_ids,
|
|
attention_mask,
|
|
layer_head_mask,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
past_key_values,
|
|
output_attentions,
|
|
cache_position,
|
|
)
|
|
|
|
hidden_states = layer_outputs[0]
|
|
if output_attentions:
|
|
all_self_attentions = all_self_attentions + (layer_outputs[1],)
|
|
if self.config.add_cross_attention:
|
|
all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
|
|
|
|
if self.is_pre_norm:
|
|
hidden_states = self.LayerNorm(hidden_states)
|
|
|
|
if output_hidden_states:
|
|
all_hidden_states = all_hidden_states + (hidden_states,)
|
|
|
|
if return_legacy_cache:
|
|
past_key_values = past_key_values.to_legacy_cache()
|
|
|
|
if not return_dict:
|
|
return tuple(
|
|
v
|
|
for v in [
|
|
hidden_states,
|
|
past_key_values,
|
|
all_hidden_states,
|
|
all_self_attentions,
|
|
all_cross_attentions,
|
|
]
|
|
if v is not None
|
|
)
|
|
return BaseModelOutputWithPastAndCrossAttentions(
|
|
last_hidden_state=hidden_states,
|
|
past_key_values=past_key_values,
|
|
hidden_states=all_hidden_states,
|
|
attentions=all_self_attentions,
|
|
cross_attentions=all_cross_attentions,
|
|
)
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaPooler
|
|
class XmodPooler(nn.Module):
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
|
|
self.activation = nn.Tanh()
|
|
|
|
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
|
|
# We "pool" the model by simply taking the hidden state corresponding
|
|
# to the first token.
|
|
first_token_tensor = hidden_states[:, 0]
|
|
pooled_output = self.dense(first_token_tensor)
|
|
pooled_output = self.activation(pooled_output)
|
|
return pooled_output
|
|
|
|
|
|
@auto_docstring
|
|
class XmodPreTrainedModel(PreTrainedModel):
|
|
config: XmodConfig
|
|
base_model_prefix = "roberta"
|
|
supports_gradient_checkpointing = True
|
|
|
|
# Copied from transformers.models.bert.modeling_bert.BertPreTrainedModel._init_weights with BertLMPredictionHead->XmodLMHead
|
|
def _init_weights(self, module):
|
|
"""Initialize the weights"""
|
|
if isinstance(module, nn.Linear):
|
|
# Slightly different from the TF version which uses truncated_normal for initialization
|
|
# cf https://github.com/pytorch/pytorch/pull/5617
|
|
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
|
|
if module.bias is not None:
|
|
module.bias.data.zero_()
|
|
elif isinstance(module, nn.Embedding):
|
|
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
|
|
if module.padding_idx is not None:
|
|
module.weight.data[module.padding_idx].zero_()
|
|
elif isinstance(module, nn.LayerNorm):
|
|
module.bias.data.zero_()
|
|
module.weight.data.fill_(1.0)
|
|
elif isinstance(module, XmodLMHead):
|
|
module.bias.data.zero_()
|
|
|
|
def set_default_language(self, language: str):
|
|
"""
|
|
Set the default language code for the model. This is used when the language is not specified in the input.
|
|
|
|
Args:
|
|
language (`str`): The language code, such as `"en_XX"` or `"de_DE"`.
|
|
"""
|
|
if language not in self.config.languages:
|
|
raise ValueError(
|
|
f"{self} does not have an adapter for {language}. Supported languages: {list(self.config.languages)}"
|
|
)
|
|
self.config.default_language = language
|
|
|
|
def freeze_embeddings_and_language_adapters(self):
|
|
"""
|
|
Freeze the embeddings and language adapters of the model. Usually, this is applied before the model is
|
|
fine-tuned on a downstream task.
|
|
"""
|
|
logger.info("Freezing embeddings")
|
|
for parameter in self.roberta.embeddings.parameters():
|
|
parameter.requires_grad = False
|
|
logger.info("Freezing adapters")
|
|
for layer in self.roberta.encoder.layer:
|
|
if layer.output.adapter_layer_norm is not None:
|
|
for parameter in layer.output.adapter_layer_norm.parameters():
|
|
parameter.requires_grad = False
|
|
for parameter in layer.output.adapter_modules.parameters():
|
|
parameter.requires_grad = False
|
|
|
|
|
|
@auto_docstring(
|
|
custom_intro="""
|
|
The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
|
|
cross-attention is added between the self-attention layers, following the architecture described in *Attention is
|
|
all you need*_ by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz
|
|
Kaiser and Illia Polosukhin.
|
|
|
|
To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
|
|
to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
|
|
`add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
|
|
|
|
.. _*Attention is all you need*: https://huggingface.co/papers/1706.03762
|
|
"""
|
|
)
|
|
class XmodModel(XmodPreTrainedModel):
|
|
# Copied from transformers.models.clap.modeling_clap.ClapTextModel.__init__ with ClapText->Xmod
|
|
def __init__(self, config, add_pooling_layer=True):
|
|
r"""
|
|
add_pooling_layer (bool, *optional*, defaults to `True`):
|
|
Whether to add a pooling layer
|
|
"""
|
|
super().__init__(config)
|
|
self.config = config
|
|
|
|
self.embeddings = XmodEmbeddings(config)
|
|
self.encoder = XmodEncoder(config)
|
|
|
|
self.pooler = XmodPooler(config) if add_pooling_layer else None
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaModel.get_input_embeddings
|
|
def get_input_embeddings(self):
|
|
return self.embeddings.word_embeddings
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaModel.set_input_embeddings
|
|
def set_input_embeddings(self, value):
|
|
self.embeddings.word_embeddings = value
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaModel._prune_heads
|
|
def _prune_heads(self, heads_to_prune):
|
|
"""
|
|
Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
|
|
class PreTrainedModel
|
|
"""
|
|
for layer, heads in heads_to_prune.items():
|
|
self.encoder.layer[layer].attention.prune_heads(heads)
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.Tensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.Tensor] = None,
|
|
token_type_ids: Optional[torch.Tensor] = None,
|
|
position_ids: Optional[torch.Tensor] = None,
|
|
head_mask: Optional[torch.Tensor] = None,
|
|
inputs_embeds: Optional[torch.Tensor] = None,
|
|
encoder_hidden_states: Optional[torch.Tensor] = None,
|
|
encoder_attention_mask: Optional[torch.Tensor] = None,
|
|
past_key_values: Optional[list[torch.FloatTensor]] = None,
|
|
use_cache: Optional[bool] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
cache_position: Optional[torch.Tensor] = None,
|
|
) -> Union[tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
|
|
r"""
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
"""
|
|
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
|
output_hidden_states = (
|
|
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
|
)
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
|
|
if self.config.is_decoder:
|
|
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
|
else:
|
|
use_cache = False
|
|
|
|
if input_ids is not None and inputs_embeds is not None:
|
|
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
|
|
elif input_ids is not None:
|
|
self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
|
|
input_shape = input_ids.size()
|
|
elif inputs_embeds is not None:
|
|
input_shape = inputs_embeds.size()[:-1]
|
|
else:
|
|
raise ValueError("You have to specify either input_ids or inputs_embeds")
|
|
|
|
batch_size, seq_length = input_shape
|
|
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
|
|
|
past_key_values_length = 0
|
|
if past_key_values is not None:
|
|
past_key_values_length = (
|
|
past_key_values[0][0].shape[-2]
|
|
if not isinstance(past_key_values, Cache)
|
|
else past_key_values.get_seq_length()
|
|
)
|
|
|
|
if lang_ids is None:
|
|
if self.config.default_language is None:
|
|
raise ValueError("Input language unknown. Please call `XmodPreTrainedModel.set_default_language()`")
|
|
adapter_languages = list(self.encoder.layer[0].output.adapter_modules.keys())
|
|
default_lang_id = adapter_languages.index(self.config.default_language)
|
|
lang_ids = default_lang_id * torch.ones(batch_size, device=device)
|
|
|
|
if attention_mask is None:
|
|
attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device)
|
|
|
|
if token_type_ids is None:
|
|
if hasattr(self.embeddings, "token_type_ids"):
|
|
buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
|
|
buffered_token_type_ids_expanded = buffered_token_type_ids.expand(batch_size, seq_length)
|
|
token_type_ids = buffered_token_type_ids_expanded
|
|
else:
|
|
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
|
|
|
|
# We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
|
|
# ourselves in which case we just need to make it broadcastable to all heads.
|
|
extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape)
|
|
|
|
# If a 2D or 3D attention mask is provided for the cross-attention
|
|
# we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
|
|
if self.config.is_decoder and encoder_hidden_states is not None:
|
|
encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
|
|
encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
|
|
if encoder_attention_mask is None:
|
|
encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
|
|
encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
|
|
else:
|
|
encoder_extended_attention_mask = None
|
|
|
|
# Prepare head mask if needed
|
|
# 1.0 in head_mask indicate we keep the head
|
|
# attention_probs has shape bsz x n_heads x N x N
|
|
# input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
|
|
# and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
|
|
head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
|
|
|
|
embedding_output = self.embeddings(
|
|
input_ids=input_ids,
|
|
position_ids=position_ids,
|
|
token_type_ids=token_type_ids,
|
|
inputs_embeds=inputs_embeds,
|
|
past_key_values_length=past_key_values_length,
|
|
)
|
|
encoder_outputs = self.encoder(
|
|
embedding_output,
|
|
lang_ids=lang_ids,
|
|
attention_mask=extended_attention_mask,
|
|
head_mask=head_mask,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_extended_attention_mask,
|
|
past_key_values=past_key_values,
|
|
use_cache=use_cache,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
cache_position=cache_position,
|
|
)
|
|
sequence_output = encoder_outputs[0]
|
|
pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
|
|
|
|
if not return_dict:
|
|
return (sequence_output, pooled_output) + encoder_outputs[1:]
|
|
|
|
return BaseModelOutputWithPoolingAndCrossAttentions(
|
|
last_hidden_state=sequence_output,
|
|
pooler_output=pooled_output,
|
|
past_key_values=encoder_outputs.past_key_values,
|
|
hidden_states=encoder_outputs.hidden_states,
|
|
attentions=encoder_outputs.attentions,
|
|
cross_attentions=encoder_outputs.cross_attentions,
|
|
)
|
|
|
|
|
|
@auto_docstring(
|
|
custom_intro="""
|
|
X-MOD Model with a `language modeling` head on top for CLM fine-tuning.
|
|
"""
|
|
)
|
|
class XmodForCausalLM(XmodPreTrainedModel, GenerationMixin):
|
|
_tied_weights_keys = ["lm_head.decoder.weight", "lm_head.decoder.bias"]
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForCausalLM.__init__ with Roberta->Xmod
|
|
def __init__(self, config):
|
|
super().__init__(config)
|
|
|
|
if not config.is_decoder:
|
|
logger.warning("If you want to use `XmodLMHeadModel` as a standalone, add `is_decoder=True.`")
|
|
|
|
self.roberta = XmodModel(config, add_pooling_layer=False)
|
|
self.lm_head = XmodLMHead(config)
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForCausalLM.get_output_embeddings
|
|
def get_output_embeddings(self):
|
|
return self.lm_head.decoder
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForCausalLM.set_output_embeddings
|
|
def set_output_embeddings(self, new_embeddings):
|
|
self.lm_head.decoder = new_embeddings
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.LongTensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
token_type_ids: Optional[torch.LongTensor] = None,
|
|
position_ids: Optional[torch.LongTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
inputs_embeds: Optional[torch.FloatTensor] = None,
|
|
encoder_hidden_states: Optional[torch.FloatTensor] = None,
|
|
encoder_attention_mask: Optional[torch.FloatTensor] = None,
|
|
labels: Optional[torch.LongTensor] = None,
|
|
past_key_values: Optional[tuple[tuple[torch.FloatTensor]]] = None,
|
|
use_cache: Optional[bool] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
cache_position: Optional[torch.Tensor] = None,
|
|
**kwargs,
|
|
) -> Union[tuple[torch.Tensor], CausalLMOutputWithCrossAttentions]:
|
|
r"""
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Labels for computing the left-to-right language modeling loss (next word prediction). Indices should be in
|
|
`[-100, 0, ..., config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are
|
|
ignored (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
|
|
|
|
Example:
|
|
|
|
```python
|
|
>>> from transformers import AutoTokenizer, XmodForCausalLM, AutoConfig
|
|
>>> import torch
|
|
|
|
>>> tokenizer = AutoTokenizer.from_pretrained("FacebookAI/xlm-roberta-base")
|
|
>>> config = AutoConfig.from_pretrained("facebook/xmod-base")
|
|
>>> config.is_decoder = True
|
|
>>> model = XmodForCausalLM.from_pretrained("facebook/xmod-base", config=config)
|
|
>>> model.set_default_language("en_XX")
|
|
|
|
>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
|
|
>>> outputs = model(**inputs)
|
|
|
|
>>> prediction_logits = outputs.logits
|
|
```"""
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
if labels is not None:
|
|
use_cache = False
|
|
|
|
outputs = self.roberta(
|
|
input_ids,
|
|
lang_ids=lang_ids,
|
|
attention_mask=attention_mask,
|
|
token_type_ids=token_type_ids,
|
|
position_ids=position_ids,
|
|
head_mask=head_mask,
|
|
inputs_embeds=inputs_embeds,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_attention_mask,
|
|
past_key_values=past_key_values,
|
|
use_cache=use_cache,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
cache_position=cache_position,
|
|
)
|
|
|
|
sequence_output = outputs[0]
|
|
prediction_scores = self.lm_head(sequence_output)
|
|
|
|
lm_loss = None
|
|
if labels is not None:
|
|
lm_loss = self.loss_function(
|
|
prediction_scores,
|
|
labels,
|
|
vocab_size=self.config.vocab_size,
|
|
**kwargs,
|
|
)
|
|
|
|
if not return_dict:
|
|
output = (prediction_scores,) + outputs[2:]
|
|
return ((lm_loss,) + output) if lm_loss is not None else output
|
|
|
|
return CausalLMOutputWithCrossAttentions(
|
|
loss=lm_loss,
|
|
logits=prediction_scores,
|
|
past_key_values=outputs.past_key_values,
|
|
hidden_states=outputs.hidden_states,
|
|
attentions=outputs.attentions,
|
|
cross_attentions=outputs.cross_attentions,
|
|
)
|
|
|
|
|
|
@auto_docstring
|
|
class XmodForMaskedLM(XmodPreTrainedModel):
|
|
_tied_weights_keys = ["lm_head.decoder.weight", "lm_head.decoder.bias"]
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForMaskedLM.__init__ with Roberta->Xmod
|
|
def __init__(self, config):
|
|
super().__init__(config)
|
|
|
|
if config.is_decoder:
|
|
logger.warning(
|
|
"If you want to use `XmodForMaskedLM` make sure `config.is_decoder=False` for "
|
|
"bi-directional self-attention."
|
|
)
|
|
|
|
self.roberta = XmodModel(config, add_pooling_layer=False)
|
|
self.lm_head = XmodLMHead(config)
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForMaskedLM.get_output_embeddings
|
|
def get_output_embeddings(self):
|
|
return self.lm_head.decoder
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForMaskedLM.set_output_embeddings
|
|
def set_output_embeddings(self, new_embeddings):
|
|
self.lm_head.decoder = new_embeddings
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.LongTensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
token_type_ids: Optional[torch.LongTensor] = None,
|
|
position_ids: Optional[torch.LongTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
inputs_embeds: Optional[torch.FloatTensor] = None,
|
|
encoder_hidden_states: Optional[torch.FloatTensor] = None,
|
|
encoder_attention_mask: Optional[torch.FloatTensor] = None,
|
|
labels: Optional[torch.LongTensor] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
) -> Union[tuple[torch.Tensor], MaskedLMOutput]:
|
|
r"""
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
|
|
config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
|
|
loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
|
|
"""
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
|
|
outputs = self.roberta(
|
|
input_ids,
|
|
lang_ids=lang_ids,
|
|
attention_mask=attention_mask,
|
|
token_type_ids=token_type_ids,
|
|
position_ids=position_ids,
|
|
head_mask=head_mask,
|
|
inputs_embeds=inputs_embeds,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_attention_mask,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
)
|
|
sequence_output = outputs[0]
|
|
prediction_scores = self.lm_head(sequence_output)
|
|
|
|
masked_lm_loss = None
|
|
if labels is not None:
|
|
loss_fct = CrossEntropyLoss()
|
|
masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
|
|
|
|
if not return_dict:
|
|
output = (prediction_scores,) + outputs[2:]
|
|
return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output
|
|
|
|
return MaskedLMOutput(
|
|
loss=masked_lm_loss,
|
|
logits=prediction_scores,
|
|
hidden_states=outputs.hidden_states,
|
|
attentions=outputs.attentions,
|
|
)
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaLMHead
|
|
class XmodLMHead(nn.Module):
|
|
"""Roberta Head for masked language modeling."""
|
|
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
|
|
self.layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
|
|
|
|
self.decoder = nn.Linear(config.hidden_size, config.vocab_size)
|
|
self.bias = nn.Parameter(torch.zeros(config.vocab_size))
|
|
self.decoder.bias = self.bias
|
|
|
|
def forward(self, features, **kwargs):
|
|
x = self.dense(features)
|
|
x = gelu(x)
|
|
x = self.layer_norm(x)
|
|
|
|
# project back to size of vocabulary with bias
|
|
x = self.decoder(x)
|
|
|
|
return x
|
|
|
|
def _tie_weights(self):
|
|
# To tie those two weights if they get disconnected (on TPU or when the bias is resized)
|
|
# For accelerate compatibility and to not break backward compatibility
|
|
if self.decoder.bias.device.type == "meta":
|
|
self.decoder.bias = self.bias
|
|
else:
|
|
self.bias = self.decoder.bias
|
|
|
|
|
|
@auto_docstring(
|
|
custom_intro="""
|
|
X-MOD Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
|
|
output) e.g. for GLUE tasks.
|
|
"""
|
|
)
|
|
class XmodForSequenceClassification(XmodPreTrainedModel):
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForSequenceClassification.__init__ with Roberta->Xmod
|
|
def __init__(self, config):
|
|
super().__init__(config)
|
|
self.num_labels = config.num_labels
|
|
self.config = config
|
|
|
|
self.roberta = XmodModel(config, add_pooling_layer=False)
|
|
self.classifier = XmodClassificationHead(config)
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.LongTensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
token_type_ids: Optional[torch.LongTensor] = None,
|
|
position_ids: Optional[torch.LongTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
inputs_embeds: Optional[torch.FloatTensor] = None,
|
|
labels: Optional[torch.LongTensor] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
) -> Union[tuple[torch.Tensor], SequenceClassifierOutput]:
|
|
r"""
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
|
|
Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
|
|
config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
|
|
`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
|
|
"""
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
|
|
outputs = self.roberta(
|
|
input_ids,
|
|
lang_ids=lang_ids,
|
|
attention_mask=attention_mask,
|
|
token_type_ids=token_type_ids,
|
|
position_ids=position_ids,
|
|
head_mask=head_mask,
|
|
inputs_embeds=inputs_embeds,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
)
|
|
sequence_output = outputs[0]
|
|
logits = self.classifier(sequence_output)
|
|
|
|
loss = None
|
|
if labels is not None:
|
|
if self.config.problem_type is None:
|
|
if self.num_labels == 1:
|
|
self.config.problem_type = "regression"
|
|
elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
|
|
self.config.problem_type = "single_label_classification"
|
|
else:
|
|
self.config.problem_type = "multi_label_classification"
|
|
|
|
if self.config.problem_type == "regression":
|
|
loss_fct = MSELoss()
|
|
if self.num_labels == 1:
|
|
loss = loss_fct(logits.squeeze(), labels.squeeze())
|
|
else:
|
|
loss = loss_fct(logits, labels)
|
|
elif self.config.problem_type == "single_label_classification":
|
|
loss_fct = CrossEntropyLoss()
|
|
loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
|
|
elif self.config.problem_type == "multi_label_classification":
|
|
loss_fct = BCEWithLogitsLoss()
|
|
loss = loss_fct(logits, labels)
|
|
|
|
if not return_dict:
|
|
output = (logits,) + outputs[2:]
|
|
return ((loss,) + output) if loss is not None else output
|
|
|
|
return SequenceClassifierOutput(
|
|
loss=loss,
|
|
logits=logits,
|
|
hidden_states=outputs.hidden_states,
|
|
attentions=outputs.attentions,
|
|
)
|
|
|
|
|
|
@auto_docstring
|
|
class XmodForMultipleChoice(XmodPreTrainedModel):
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForMultipleChoice.__init__ with Roberta->Xmod
|
|
def __init__(self, config):
|
|
super().__init__(config)
|
|
|
|
self.roberta = XmodModel(config)
|
|
self.dropout = nn.Dropout(config.hidden_dropout_prob)
|
|
self.classifier = nn.Linear(config.hidden_size, 1)
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.LongTensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
token_type_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
labels: Optional[torch.LongTensor] = None,
|
|
position_ids: Optional[torch.LongTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
inputs_embeds: Optional[torch.FloatTensor] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
) -> Union[tuple[torch.Tensor], MultipleChoiceModelOutput]:
|
|
r"""
|
|
input_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`):
|
|
Indices of input sequence tokens in the vocabulary.
|
|
|
|
Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
|
|
[`PreTrainedTokenizer.__call__`] for details.
|
|
|
|
[What are input IDs?](../glossary#input-ids)
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
token_type_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`, *optional*):
|
|
Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0,
|
|
1]`:
|
|
|
|
- 0 corresponds to a *sentence A* token,
|
|
- 1 corresponds to a *sentence B* token.
|
|
|
|
[What are token type IDs?](../glossary#token-type-ids)
|
|
labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
|
|
Labels for computing the multiple choice classification loss. Indices should be in `[0, ...,
|
|
num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See
|
|
`input_ids` above)
|
|
position_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`, *optional*):
|
|
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
|
|
config.max_position_embeddings - 1]`.
|
|
|
|
[What are position IDs?](../glossary#position-ids)
|
|
inputs_embeds (`torch.FloatTensor` of shape `(batch_size, num_choices, sequence_length, hidden_size)`, *optional*):
|
|
Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
|
|
is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
|
|
model's internal embedding lookup matrix.
|
|
"""
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]
|
|
|
|
flat_input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None
|
|
flat_lang_ids = lang_ids.repeat(input_ids.size(0) * input_ids.size(1)) if lang_ids is not None else None
|
|
flat_position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None
|
|
flat_token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None
|
|
flat_attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None
|
|
flat_inputs_embeds = (
|
|
inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1))
|
|
if inputs_embeds is not None
|
|
else None
|
|
)
|
|
|
|
outputs = self.roberta(
|
|
flat_input_ids,
|
|
lang_ids=flat_lang_ids,
|
|
position_ids=flat_position_ids,
|
|
token_type_ids=flat_token_type_ids,
|
|
attention_mask=flat_attention_mask,
|
|
head_mask=head_mask,
|
|
inputs_embeds=flat_inputs_embeds,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
)
|
|
pooled_output = outputs[1]
|
|
|
|
pooled_output = self.dropout(pooled_output)
|
|
logits = self.classifier(pooled_output)
|
|
reshaped_logits = logits.view(-1, num_choices)
|
|
|
|
loss = None
|
|
if labels is not None:
|
|
loss_fct = CrossEntropyLoss()
|
|
loss = loss_fct(reshaped_logits, labels)
|
|
|
|
if not return_dict:
|
|
output = (reshaped_logits,) + outputs[2:]
|
|
return ((loss,) + output) if loss is not None else output
|
|
|
|
return MultipleChoiceModelOutput(
|
|
loss=loss,
|
|
logits=reshaped_logits,
|
|
hidden_states=outputs.hidden_states,
|
|
attentions=outputs.attentions,
|
|
)
|
|
|
|
|
|
@auto_docstring
|
|
class XmodForTokenClassification(XmodPreTrainedModel):
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForTokenClassification.__init__ with Roberta->Xmod
|
|
def __init__(self, config):
|
|
super().__init__(config)
|
|
self.num_labels = config.num_labels
|
|
|
|
self.roberta = XmodModel(config, add_pooling_layer=False)
|
|
classifier_dropout = (
|
|
config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
|
|
)
|
|
self.dropout = nn.Dropout(classifier_dropout)
|
|
self.classifier = nn.Linear(config.hidden_size, config.num_labels)
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.LongTensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
token_type_ids: Optional[torch.LongTensor] = None,
|
|
position_ids: Optional[torch.LongTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
inputs_embeds: Optional[torch.FloatTensor] = None,
|
|
labels: Optional[torch.LongTensor] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
) -> Union[tuple[torch.Tensor], TokenClassifierOutput]:
|
|
r"""
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
|
|
"""
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
|
|
outputs = self.roberta(
|
|
input_ids,
|
|
lang_ids=lang_ids,
|
|
attention_mask=attention_mask,
|
|
token_type_ids=token_type_ids,
|
|
position_ids=position_ids,
|
|
head_mask=head_mask,
|
|
inputs_embeds=inputs_embeds,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
)
|
|
|
|
sequence_output = outputs[0]
|
|
|
|
sequence_output = self.dropout(sequence_output)
|
|
logits = self.classifier(sequence_output)
|
|
|
|
loss = None
|
|
if labels is not None:
|
|
loss_fct = CrossEntropyLoss()
|
|
loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
|
|
|
|
if not return_dict:
|
|
output = (logits,) + outputs[2:]
|
|
return ((loss,) + output) if loss is not None else output
|
|
|
|
return TokenClassifierOutput(
|
|
loss=loss,
|
|
logits=logits,
|
|
hidden_states=outputs.hidden_states,
|
|
attentions=outputs.attentions,
|
|
)
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaClassificationHead
|
|
class XmodClassificationHead(nn.Module):
|
|
"""Head for sentence-level classification tasks."""
|
|
|
|
def __init__(self, config):
|
|
super().__init__()
|
|
self.dense = nn.Linear(config.hidden_size, config.hidden_size)
|
|
classifier_dropout = (
|
|
config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
|
|
)
|
|
self.dropout = nn.Dropout(classifier_dropout)
|
|
self.out_proj = nn.Linear(config.hidden_size, config.num_labels)
|
|
|
|
def forward(self, features, **kwargs):
|
|
x = features[:, 0, :] # take <s> token (equiv. to [CLS])
|
|
x = self.dropout(x)
|
|
x = self.dense(x)
|
|
x = torch.tanh(x)
|
|
x = self.dropout(x)
|
|
x = self.out_proj(x)
|
|
return x
|
|
|
|
|
|
@auto_docstring
|
|
class XmodForQuestionAnswering(XmodPreTrainedModel):
|
|
# Copied from transformers.models.roberta.modeling_roberta.RobertaForQuestionAnswering.__init__ with Roberta->Xmod
|
|
def __init__(self, config):
|
|
super().__init__(config)
|
|
self.num_labels = config.num_labels
|
|
|
|
self.roberta = XmodModel(config, add_pooling_layer=False)
|
|
self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
|
|
|
|
# Initialize weights and apply final processing
|
|
self.post_init()
|
|
|
|
@auto_docstring
|
|
def forward(
|
|
self,
|
|
input_ids: Optional[torch.LongTensor] = None,
|
|
lang_ids: Optional[torch.LongTensor] = None,
|
|
attention_mask: Optional[torch.FloatTensor] = None,
|
|
token_type_ids: Optional[torch.LongTensor] = None,
|
|
position_ids: Optional[torch.LongTensor] = None,
|
|
head_mask: Optional[torch.FloatTensor] = None,
|
|
inputs_embeds: Optional[torch.FloatTensor] = None,
|
|
start_positions: Optional[torch.LongTensor] = None,
|
|
end_positions: Optional[torch.LongTensor] = None,
|
|
output_attentions: Optional[bool] = None,
|
|
output_hidden_states: Optional[bool] = None,
|
|
return_dict: Optional[bool] = None,
|
|
) -> Union[tuple[torch.Tensor], QuestionAnsweringModelOutput]:
|
|
r"""
|
|
lang_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
|
|
Indices of the language adapters that should be activated for each sample, respectively. Default: the index
|
|
that corresponds to `self.config.default_language`.
|
|
"""
|
|
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
|
|
|
outputs = self.roberta(
|
|
input_ids,
|
|
lang_ids=lang_ids,
|
|
attention_mask=attention_mask,
|
|
token_type_ids=token_type_ids,
|
|
position_ids=position_ids,
|
|
head_mask=head_mask,
|
|
inputs_embeds=inputs_embeds,
|
|
output_attentions=output_attentions,
|
|
output_hidden_states=output_hidden_states,
|
|
return_dict=return_dict,
|
|
)
|
|
|
|
sequence_output = outputs[0]
|
|
|
|
logits = self.qa_outputs(sequence_output)
|
|
start_logits, end_logits = logits.split(1, dim=-1)
|
|
start_logits = start_logits.squeeze(-1).contiguous()
|
|
end_logits = end_logits.squeeze(-1).contiguous()
|
|
|
|
total_loss = None
|
|
if start_positions is not None and end_positions is not None:
|
|
# If we are on multi-GPU, split add a dimension
|
|
if len(start_positions.size()) > 1:
|
|
start_positions = start_positions.squeeze(-1)
|
|
if len(end_positions.size()) > 1:
|
|
end_positions = end_positions.squeeze(-1)
|
|
# sometimes the start/end positions are outside our model inputs, we ignore these terms
|
|
ignored_index = start_logits.size(1)
|
|
start_positions = start_positions.clamp(0, ignored_index)
|
|
end_positions = end_positions.clamp(0, ignored_index)
|
|
|
|
loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
|
|
start_loss = loss_fct(start_logits, start_positions)
|
|
end_loss = loss_fct(end_logits, end_positions)
|
|
total_loss = (start_loss + end_loss) / 2
|
|
|
|
if not return_dict:
|
|
output = (start_logits, end_logits) + outputs[2:]
|
|
return ((total_loss,) + output) if total_loss is not None else output
|
|
|
|
return QuestionAnsweringModelOutput(
|
|
loss=total_loss,
|
|
start_logits=start_logits,
|
|
end_logits=end_logits,
|
|
hidden_states=outputs.hidden_states,
|
|
attentions=outputs.attentions,
|
|
)
|
|
|
|
|
|
# Copied from transformers.models.roberta.modeling_roberta.create_position_ids_from_input_ids
|
|
def create_position_ids_from_input_ids(input_ids, padding_idx, past_key_values_length=0):
|
|
"""
|
|
Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
|
|
are ignored. This is modified from fairseq's `utils.make_positions`.
|
|
|
|
Args:
|
|
x: torch.Tensor x:
|
|
|
|
Returns: torch.Tensor
|
|
"""
|
|
# The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
|
|
mask = input_ids.ne(padding_idx).int()
|
|
incremental_indices = (torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length) * mask
|
|
return incremental_indices.long() + padding_idx
|
|
|
|
|
|
__all__ = [
|
|
"XmodForCausalLM",
|
|
"XmodForMaskedLM",
|
|
"XmodForMultipleChoice",
|
|
"XmodForQuestionAnswering",
|
|
"XmodForSequenceClassification",
|
|
"XmodForTokenClassification",
|
|
"XmodModel",
|
|
"XmodPreTrainedModel",
|
|
]
|