Spaces:
Runtime error
Runtime error
# coding=utf-8 | |
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team. | |
# Copyright (c) 2018, NVIDIA CORPORATION. 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. | |
"""PyTorch RoBERTa model.""" | |
import math | |
from typing import List, Optional, Tuple, Union | |
import torch | |
import torch.utils.checkpoint | |
from packaging import version | |
from torch import nn | |
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss | |
from .decompx_utils import DecompXConfig, DecompXOutput | |
from transformers.activations import ACT2FN, gelu | |
from transformers.modeling_outputs import ( | |
BaseModelOutputWithPastAndCrossAttentions, | |
BaseModelOutputWithPoolingAndCrossAttentions, | |
CausalLMOutputWithCrossAttentions, | |
MaskedLMOutput, | |
MultipleChoiceModelOutput, | |
QuestionAnsweringModelOutput, | |
SequenceClassifierOutput, | |
TokenClassifierOutput, | |
) | |
from transformers.modeling_utils import ( | |
PreTrainedModel, | |
apply_chunking_to_forward, | |
find_pruneable_heads_and_indices, | |
prune_linear_layer, | |
) | |
from transformers.utils import ( | |
add_code_sample_docstrings, | |
add_start_docstrings, | |
add_start_docstrings_to_model_forward, | |
logging, | |
replace_return_docstrings, | |
) | |
from transformers.models.roberta.configuration_roberta import RobertaConfig | |
logger = logging.get_logger(__name__) | |
_CHECKPOINT_FOR_DOC = "roberta-base" | |
_CONFIG_FOR_DOC = "RobertaConfig" | |
_TOKENIZER_FOR_DOC = "RobertaTokenizer" | |
ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST = [ | |
"roberta-base", | |
"roberta-large", | |
"roberta-large-mnli", | |
"distilroberta-base", | |
"roberta-base-openai-detector", | |
"roberta-large-openai-detector", | |
# See all RoBERTa models at https://huggingface.co/models?filter=roberta | |
] | |
def output_builder(input_vector, output_mode): | |
if output_mode is None: | |
return None | |
elif output_mode == "vector": | |
return (input_vector,) | |
elif output_mode == "norm": | |
return (torch.norm(input_vector, dim=-1),) | |
elif output_mode == "both": | |
return ((torch.norm(input_vector, dim=-1), input_vector),) | |
elif output_mode == "distance_based": | |
recomposed_vectors = torch.sum(input_vector, dim=-2, keepdim=True) | |
importance_matrix = -torch.nn.functional.pairwise_distance(input_vector, recomposed_vectors, p=1) | |
norm_y = torch.norm(recomposed_vectors, dim=-1, p=1) | |
maxed = torch.maximum(torch.zeros(1, device=norm_y.device), norm_y + importance_matrix) | |
return (maxed / (torch.sum(maxed, dim=-2, keepdim=True) + 1e-12),) | |
class RobertaEmbeddings(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))) | |
if version.parse(torch.__version__) > version.parse("1.6.0"): | |
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 | |
else: | |
return inputs_embeds | |
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.bert.modeling_bert.BertSelfAttention with Bert->Roberta | |
class RobertaSelfAttention(nn.Module): | |
def __init__(self, config, position_embedding_type=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 | |
def transpose_for_scores(self, x): | |
new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size) | |
x = x.view(new_x_shape) | |
return x.permute(0, 2, 1, 3) | |
def transpose_for_scores_for_decomposed(self, x): | |
# x: (B, N, N, H*V) | |
new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size) | |
# x: (B, N, N, H, V) | |
x = x.view(new_x_shape) | |
# x: (B, H, N, N, V) | |
return x.permute(0, 3, 1, 2, 4) | |
def forward( | |
self, | |
hidden_states: torch.Tensor, | |
attribution_vectors: Optional[torch.FloatTensor] = None, | |
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[Tuple[Tuple[torch.FloatTensor]]] = None, | |
output_attentions: Optional[bool] = False, | |
decompx_ready: Optional[bool] = None, # added by Fayyaz / Modarressi | |
) -> Tuple[torch.Tensor]: | |
mixed_query_layer = self.query(hidden_states) | |
# 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 | |
decomposed_value_layer = None | |
if is_cross_attention and past_key_value is not None: | |
# reuse k,v, cross_attentions | |
key_layer = past_key_value[0] | |
value_layer = past_key_value[1] | |
attention_mask = encoder_attention_mask | |
elif is_cross_attention: | |
key_layer = self.transpose_for_scores(self.key(encoder_hidden_states)) | |
value_layer = self.transpose_for_scores(self.value(encoder_hidden_states)) | |
attention_mask = encoder_attention_mask | |
elif past_key_value is not None: | |
key_layer = self.transpose_for_scores(self.key(hidden_states)) | |
value_layer = self.transpose_for_scores(self.value(hidden_states)) | |
key_layer = torch.cat([past_key_value[0], key_layer], dim=2) | |
value_layer = torch.cat([past_key_value[1], value_layer], dim=2) | |
else: | |
key_layer = self.transpose_for_scores(self.key(hidden_states)) | |
value_layer = self.transpose_for_scores(self.value(hidden_states)) | |
if attribution_vectors is not None: | |
decomposed_value_layer = torch.einsum("bijd,vd->bijv", attribution_vectors, self.value.weight) | |
decomposed_value_layer = self.transpose_for_scores_for_decomposed(decomposed_value_layer) | |
query_layer = self.transpose_for_scores(mixed_query_layer) | |
if self.is_decoder: | |
# if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states. | |
# Further calls to cross_attention layer can then reuse all cross-attention | |
# key/value_states (first "if" case) | |
# if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of | |
# all previous decoder key/value_states. Further calls to uni-directional self-attention | |
# can concat previous decoder key/value_states to current projected key/value_states (third "elif" case) | |
# if encoder bi-directional self-attention `past_key_value` is always `None` | |
past_key_value = (key_layer, value_layer) | |
# 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": | |
seq_length = hidden_states.size()[1] | |
position_ids_l = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(-1, 1) | |
position_ids_r = torch.arange(seq_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 RobertaModel 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) | |
# added by Fayyaz / Modarressi | |
# ------------------------------- | |
if decompx_ready: | |
outputs = (context_layer, attention_probs, value_layer, decomposed_value_layer) | |
return outputs | |
# ------------------------------- | |
outputs = (context_layer, attention_probs) if output_attentions else (context_layer,) | |
if self.is_decoder: | |
outputs = outputs + (past_key_value,) | |
return outputs | |
# Copied from transformers.models.bert.modeling_bert.BertSelfOutput | |
class RobertaSelfOutput(nn.Module): | |
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, | |
decompx_ready=False): # added by Fayyaz / Modarressi | |
hidden_states = self.dense(hidden_states) | |
hidden_states = self.dropout(hidden_states) | |
# hidden_states = self.LayerNorm(hidden_states + input_tensor) | |
pre_ln_states = hidden_states + input_tensor # added by Fayyaz / Modarressi | |
post_ln_states = self.LayerNorm(pre_ln_states) # added by Fayyaz / Modarressi | |
# added by Fayyaz / Modarressi | |
if decompx_ready: | |
return post_ln_states, pre_ln_states | |
else: | |
return post_ln_states | |
# Copied from transformers.models.bert.modeling_bert.BertAttention with Bert->Roberta | |
class RobertaAttention(nn.Module): | |
def __init__(self, config, position_embedding_type=None): | |
super().__init__() | |
self.self = RobertaSelfAttention(config, position_embedding_type=position_embedding_type) | |
self.output = RobertaSelfOutput(config) | |
self.pruned_heads = set() | |
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) | |
def forward( | |
self, | |
hidden_states: torch.Tensor, | |
attribution_vectors: Optional[torch.FloatTensor] = None, | |
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[Tuple[Tuple[torch.FloatTensor]]] = None, | |
output_attentions: Optional[bool] = False, | |
decompx_ready: Optional[bool] = None, # added by Fayyaz / Modarressi | |
) -> Tuple[torch.Tensor]: | |
self_outputs = self.self( | |
hidden_states, | |
attribution_vectors, | |
attention_mask, | |
head_mask, | |
encoder_hidden_states, | |
encoder_attention_mask, | |
past_key_value, | |
output_attentions, | |
decompx_ready=decompx_ready, # added by Fayyaz / Modarressi | |
) | |
attention_output = self.output( | |
self_outputs[0], | |
hidden_states, | |
decompx_ready=decompx_ready, # added by Goro Kobayashi (Edited by Fayyaz / Modarressi) | |
) | |
# Added by Fayyaz / Modarressi | |
# ------------------------------- | |
if decompx_ready: | |
_, attention_probs, value_layer, decomposed_value_layer = self_outputs | |
attention_output, pre_ln_states = attention_output | |
outputs = (attention_output, attention_probs,) + ( | |
value_layer, decomposed_value_layer, pre_ln_states) # add attentions and norms if we output them | |
return outputs | |
# ------------------------------- | |
outputs = (attention_output,) + self_outputs[1:] # add attentions if we output them | |
return outputs | |
# Copied from transformers.models.bert.modeling_bert.BertIntermediate | |
class RobertaIntermediate(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, decompx_ready: Optional[bool] = False) -> torch.Tensor: | |
pre_act_hidden_states = self.dense(hidden_states) | |
hidden_states = self.intermediate_act_fn(pre_act_hidden_states) | |
if decompx_ready: | |
return hidden_states, pre_act_hidden_states | |
return hidden_states, None | |
# Copied from transformers.models.bert.modeling_bert.BertOutput | |
class RobertaOutput(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.dropout = nn.Dropout(config.hidden_dropout_prob) | |
def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor, decompx_ready: Optional[bool] = False): | |
hidden_states = self.dense(hidden_states) | |
hidden_states = self.dropout(hidden_states) | |
# hidden_states = self.LayerNorm(hidden_states + input_tensor) | |
# return hidden_states | |
# Added by Fayyaz / Modarressi | |
# ------------------------------- | |
pre_ln_states = hidden_states + input_tensor | |
hidden_states = self.LayerNorm(pre_ln_states) | |
if decompx_ready: | |
return hidden_states, pre_ln_states | |
return hidden_states, None | |
# ------------------------------- | |
# Copied from transformers.models.bert.modeling_bert.BertLayer with Bert->Roberta | |
class RobertaLayer(nn.Module): | |
def __init__(self, config): | |
super().__init__() | |
self.chunk_size_feed_forward = config.chunk_size_feed_forward | |
self.seq_len_dim = 1 | |
self.attention = RobertaAttention(config) | |
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 = RobertaAttention(config, position_embedding_type="absolute") | |
self.intermediate = RobertaIntermediate(config) | |
self.output = RobertaOutput(config) | |
self.similarity_fn = torch.nn.CosineSimilarity(dim=-1) | |
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 | |
def bias_decomposer(self, bias, attribution_vectors, bias_decomp_type="absdot"): | |
# Decomposes the input bias based on similarity to the attribution vectors | |
# Args: | |
# bias: a bias vector (all_head_size) | |
# attribution_vectors: the attribution vectors from token j to i (b, i, j, all_head_size) :: (batch, seq_length, seq_length, all_head_size) | |
if bias_decomp_type == "absdot": | |
weights = torch.abs(torch.einsum("bskd,d->bsk", attribution_vectors, bias)) | |
elif bias_decomp_type == "abssim": | |
weights = torch.abs(torch.nn.functional.cosine_similarity(attribution_vectors, bias, dim=-1)) | |
weights = (torch.norm(attribution_vectors, dim=-1) != 0) * weights | |
elif bias_decomp_type == "norm": | |
weights = torch.norm(attribution_vectors, dim=-1) | |
elif bias_decomp_type == "equal": | |
weights = (torch.norm(attribution_vectors, dim=-1) != 0) * 1.0 | |
elif bias_decomp_type == "cls": | |
weights = torch.zeros(attribution_vectors.shape[:-1], device=attribution_vectors.device) | |
weights[:, :, 0] = 1.0 | |
elif bias_decomp_type == "dot": | |
weights = torch.einsum("bskd,d->bsk", attribution_vectors, bias) | |
elif bias_decomp_type == "biastoken": | |
attrib_shape = attribution_vectors.shape | |
if attrib_shape[1] == attrib_shape[2]: | |
attribution_vectors = torch.concat([attribution_vectors, | |
torch.zeros((attrib_shape[0], attrib_shape[1], 1, attrib_shape[3]), | |
device=attribution_vectors.device)], dim=-2) | |
attribution_vectors[:, :, -1] = attribution_vectors[:, :, -1] + bias | |
return attribution_vectors | |
weights = weights / (weights.sum(dim=-1, keepdim=True) + 1e-12) | |
weighted_bias = torch.matmul(weights.unsqueeze(dim=-1), bias.unsqueeze(dim=0)) | |
return attribution_vectors + weighted_bias | |
def ln_decomposer(self, attribution_vectors, pre_ln_states, gamma, beta, eps, include_biases=True, | |
bias_decomp_type="absdot"): | |
mean = pre_ln_states.mean(-1, keepdim=True) # (batch, seq_len, 1) m(y=Σy_j) | |
var = (pre_ln_states - mean).pow(2).mean(-1, keepdim=True).unsqueeze(dim=2) # (batch, seq_len, 1, 1) s(y) | |
each_mean = attribution_vectors.mean(-1, keepdim=True) # (batch, seq_len, seq_len, 1) m(y_j) | |
normalized_layer = torch.div(attribution_vectors - each_mean, | |
(var + eps) ** (1 / 2)) # (batch, seq_len, seq_len, all_head_size) | |
post_ln_layer = torch.einsum('bskd,d->bskd', normalized_layer, | |
gamma) # (batch, seq_len, seq_len, all_head_size) | |
if include_biases: | |
return self.bias_decomposer(beta, post_ln_layer, bias_decomp_type=bias_decomp_type) | |
else: | |
return post_ln_layer | |
def gelu_linear_approximation(self, intermediate_hidden_states, intermediate_output): | |
def phi(x): | |
return (1 + torch.erf(x / math.sqrt(2))) / 2. | |
def normal_pdf(x): | |
return torch.exp(-(x ** 2) / 2) / math.sqrt(2. * math.pi) | |
def gelu_deriv(x): | |
return phi(x) + x * normal_pdf(x) | |
m = gelu_deriv(intermediate_hidden_states) | |
b = intermediate_output - m * intermediate_hidden_states | |
return m, b | |
def gelu_decomposition(self, attribution_vectors, intermediate_hidden_states, intermediate_output, | |
bias_decomp_type): | |
m, b = self.gelu_linear_approximation(intermediate_hidden_states, intermediate_output) | |
mx = attribution_vectors * m.unsqueeze(dim=-2) | |
if bias_decomp_type == "absdot": | |
weights = torch.abs(torch.einsum("bskl,bsl->bsk", mx, b)) | |
elif bias_decomp_type == "abssim": | |
weights = torch.abs(torch.nn.functional.cosine_similarity(mx, b)) | |
weights = (torch.norm(mx, dim=-1) != 0) * weights | |
elif bias_decomp_type == "norm": | |
weights = torch.norm(mx, dim=-1) | |
elif bias_decomp_type == "equal": | |
weights = (torch.norm(mx, dim=-1) != 0) * 1.0 | |
elif bias_decomp_type == "cls": | |
weights = torch.zeros(mx.shape[:-1], device=mx.device) | |
weights[:, :, 0] = 1.0 | |
weights = weights / (weights.sum(dim=-1, keepdim=True) + 1e-12) | |
weighted_bias = torch.einsum("bsl,bsk->bskl", b, weights) | |
return mx + weighted_bias | |
def gelu_zo_decomposition(self, attribution_vectors, intermediate_hidden_states, intermediate_output): | |
m = intermediate_output / (intermediate_hidden_states + 1e-12) | |
mx = attribution_vectors * m.unsqueeze(dim=-2) | |
return mx | |
def ffn_decomposer(self, attribution_vectors, intermediate_hidden_states, intermediate_output, include_biases=True, | |
approximation_type="GeLU_LA", bias_decomp_type="absdot"): | |
post_first_layer = torch.einsum("ld,bskd->bskl", self.intermediate.dense.weight, attribution_vectors) | |
if include_biases: | |
post_first_layer = self.bias_decomposer(self.intermediate.dense.bias, post_first_layer, | |
bias_decomp_type=bias_decomp_type) | |
if approximation_type == "ReLU": | |
mask_for_gelu_approx = (intermediate_hidden_states > 0) | |
post_act_first_layer = torch.einsum("bskl, bsl->bskl", post_first_layer, mask_for_gelu_approx) | |
post_act_first_layer = post_first_layer * mask_for_gelu_approx.unsqueeze(dim=-2) | |
elif approximation_type == "GeLU_LA": | |
post_act_first_layer = self.gelu_decomposition(post_first_layer, intermediate_hidden_states, | |
intermediate_output, bias_decomp_type=bias_decomp_type) | |
elif approximation_type == "GeLU_ZO": | |
post_act_first_layer = self.gelu_zo_decomposition(post_first_layer, intermediate_hidden_states, | |
intermediate_output) | |
post_second_layer = torch.einsum("bskl, dl->bskd", post_act_first_layer, self.output.dense.weight) | |
if include_biases: | |
post_second_layer = self.bias_decomposer(self.output.dense.bias, post_second_layer, | |
bias_decomp_type=bias_decomp_type) | |
return post_second_layer | |
def ffn_decomposer_fast(self, attribution_vectors, intermediate_hidden_states, intermediate_output, | |
include_biases=True, approximation_type="GeLU_LA", bias_decomp_type="absdot"): | |
if approximation_type == "ReLU": | |
theta = (intermediate_hidden_states > 0) | |
elif approximation_type == "GeLU_ZO": | |
theta = intermediate_output / (intermediate_hidden_states + 1e-12) | |
scaled_W1 = torch.einsum("bsl,ld->bsld", theta, self.intermediate.dense.weight) | |
W_equiv = torch.einsum("bsld, zl->bszd", scaled_W1, self.output.dense.weight) | |
post_ffn_layer = torch.einsum("bszd,bskd->bskz", W_equiv, attribution_vectors) | |
if include_biases: | |
scaled_b1 = torch.einsum("bsl,l->bsl", theta, self.intermediate.dense.bias) | |
b_equiv = torch.einsum("bsl, dl->bsd", scaled_b1, self.output.dense.weight) | |
b_equiv = b_equiv + self.output.dense.bias | |
if bias_decomp_type == "absdot": | |
weights = torch.abs(torch.einsum("bskd,bsd->bsk", post_ffn_layer, b_equiv)) | |
elif bias_decomp_type == "abssim": | |
weights = torch.abs(torch.nn.functional.cosine_similarity(post_ffn_layer, b_equiv)) | |
weights = (torch.norm(post_ffn_layer, dim=-1) != 0) * weights | |
elif bias_decomp_type == "norm": | |
weights = torch.norm(post_ffn_layer, dim=-1) | |
elif bias_decomp_type == "equal": | |
weights = (torch.norm(post_ffn_layer, dim=-1) != 0) * 1.0 | |
weights = weights / (weights.sum(dim=-1, keepdim=True) + 1e-12) | |
weighted_bias = torch.einsum("bsd,bsk->bskd", b_equiv, weights) | |
post_ffn_layer = post_ffn_layer + weighted_bias | |
return post_ffn_layer | |
def forward( | |
self, | |
hidden_states: torch.Tensor, | |
attribution_vectors: Optional[torch.FloatTensor] = None, | |
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[Tuple[Tuple[torch.FloatTensor]]] = None, | |
output_attentions: Optional[bool] = False, | |
decompx_config: Optional[DecompXConfig] = None, # added by Fayyaz / Modarressi | |
) -> Tuple[torch.Tensor]: | |
decompx_ready = decompx_config is not None | |
# decoder uni-directional self-attention cached key/values tuple is at positions 1,2 | |
# self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None | |
# self_attention_outputs = self.attention( | |
# hidden_states, | |
# attribution_vectors, | |
# attention_mask, | |
# head_mask, | |
# output_attentions=output_attentions, | |
# past_key_value=self_attn_past_key_value, | |
# decompx_ready=decompx_ready, | |
# ) | |
self_attention_outputs = self.attention( | |
hidden_states, | |
attribution_vectors, | |
attention_mask, | |
head_mask, | |
output_attentions=output_attentions, | |
decompx_ready=decompx_ready, | |
) # changed by Goro Kobayashi | |
attention_output = self_attention_outputs[0] | |
# if decoder, the last output is tuple of self-attn cache | |
if self.is_decoder: | |
outputs = self_attention_outputs[1:-1] | |
present_key_value = self_attention_outputs[-1] | |
else: | |
outputs = self_attention_outputs[1:] # add self attentions if we output attention weights | |
cross_attn_present_key_value = None | |
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_attn cached key/values tuple is at positions 3,4 of past_key_value tuple | |
cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None | |
cross_attention_outputs = self.crossattention( | |
attention_output, | |
attention_mask, | |
head_mask, | |
encoder_hidden_states, | |
encoder_attention_mask, | |
cross_attn_past_key_value, | |
output_attentions, | |
) | |
attention_output = cross_attention_outputs[0] | |
outputs = outputs + cross_attention_outputs[1:-1] # add cross attentions if we output attention weights | |
# add cross-attn cache to positions 3,4 of present_key_value tuple | |
cross_attn_present_key_value = cross_attention_outputs[-1] | |
present_key_value = present_key_value + cross_attn_present_key_value | |
# layer_output = apply_chunking_to_forward( | |
# self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output | |
# ) | |
# Added by Fayyaz / Modarressi | |
# ------------------------------- | |
bias_decomp_type = "biastoken" if decompx_config.include_bias_token else decompx_config.bias_decomp_type | |
intermediate_output, pre_act_hidden_states = self.intermediate(attention_output, decompx_ready=decompx_ready) | |
layer_output, pre_ln2_states = self.output(intermediate_output, attention_output, decompx_ready=decompx_ready) | |
if decompx_ready: | |
attention_probs, value_layer, decomposed_value_layer, pre_ln_states = outputs | |
headmixing_weight = self.attention.output.dense.weight.view(self.all_head_size, self.num_attention_heads, | |
self.attention_head_size) | |
if decomposed_value_layer is None or decompx_config.aggregation != "vector": | |
transformed_layer = torch.einsum('bhsv,dhv->bhsd', value_layer, headmixing_weight) # V * W^o (z=(qk)v) | |
# Make weighted vectors αf(x) from transformed vectors (transformed_layer) | |
# and attention weights (attentions): | |
# (batch, num_heads, seq_length, seq_length, all_head_size) | |
weighted_layer = torch.einsum('bhks,bhsd->bhksd', attention_probs, | |
transformed_layer) # attention_probs(Q*K^t) * V * W^o | |
# Sum each weighted vectors αf(x) over all heads: | |
# (batch, seq_length, seq_length, all_head_size) | |
summed_weighted_layer = weighted_layer.sum(dim=1) # sum over heads | |
# Make residual matrix (batch, seq_length, seq_length, all_head_size) | |
hidden_shape = hidden_states.size() # (batch, seq_length, all_head_size) | |
device = hidden_states.device | |
residual = torch.einsum('sk,bsd->bskd', torch.eye(hidden_shape[1]).to(device), | |
hidden_states) # diagonal representations (hidden states) | |
# Make matrix of summed weighted vector + residual vectors | |
residual_weighted_layer = summed_weighted_layer + residual | |
accumulated_bias = self.attention.output.dense.bias | |
else: | |
transformed_layer = torch.einsum('bhsqv,dhv->bhsqd', decomposed_value_layer, headmixing_weight) | |
weighted_layer = torch.einsum('bhks,bhsqd->bhkqd', attention_probs, | |
transformed_layer) # attention_probs(Q*K^t) * V * W^o | |
summed_weighted_layer = weighted_layer.sum(dim=1) # sum over heads | |
residual_weighted_layer = summed_weighted_layer + attribution_vectors | |
accumulated_bias = torch.matmul(self.attention.output.dense.weight, | |
self.attention.self.value.bias) + self.attention.output.dense.bias | |
if decompx_config.include_biases: | |
residual_weighted_layer = self.bias_decomposer(accumulated_bias, residual_weighted_layer, | |
bias_decomp_type) | |
if decompx_config.include_LN1: | |
post_ln_layer = self.ln_decomposer( | |
attribution_vectors=residual_weighted_layer, | |
pre_ln_states=pre_ln_states, | |
gamma=self.attention.output.LayerNorm.weight.data, | |
beta=self.attention.output.LayerNorm.bias.data, | |
eps=self.attention.output.LayerNorm.eps, | |
include_biases=decompx_config.include_biases, | |
bias_decomp_type=bias_decomp_type | |
) | |
else: | |
post_ln_layer = residual_weighted_layer | |
if decompx_config.include_FFN: | |
post_ffn_layer = self.ffn_decomposer_fast if decompx_config.FFN_fast_mode else self.ffn_decomposer( | |
attribution_vectors=post_ln_layer, | |
intermediate_hidden_states=pre_act_hidden_states, | |
intermediate_output=intermediate_output, | |
approximation_type=decompx_config.FFN_approx_type, | |
include_biases=decompx_config.include_biases, | |
bias_decomp_type=bias_decomp_type | |
) | |
pre_ln2_layer = post_ln_layer + post_ffn_layer | |
else: | |
pre_ln2_layer = post_ln_layer | |
post_ffn_layer = None | |
if decompx_config.include_LN2: | |
post_ln2_layer = self.ln_decomposer( | |
attribution_vectors=pre_ln2_layer, | |
pre_ln_states=pre_ln2_states, | |
gamma=self.output.LayerNorm.weight.data, | |
beta=self.output.LayerNorm.bias.data, | |
eps=self.output.LayerNorm.eps, | |
include_biases=decompx_config.include_biases, | |
bias_decomp_type=bias_decomp_type | |
) | |
else: | |
post_ln2_layer = pre_ln2_layer | |
new_outputs = DecompXOutput( | |
attention=output_builder(summed_weighted_layer, decompx_config.output_attention), | |
res1=output_builder(residual_weighted_layer, decompx_config.output_res1), | |
LN1=output_builder(post_ln_layer, decompx_config.output_res2), | |
FFN=output_builder(post_ffn_layer, decompx_config.output_FFN), | |
res2=output_builder(pre_ln2_layer, decompx_config.output_res2), | |
encoder=output_builder(post_ln2_layer, "both") | |
) | |
return (layer_output,) + (new_outputs,) | |
# ------------------------------- | |
outputs = (layer_output,) + outputs | |
# if decoder, return the attn key/values as the last output | |
if self.is_decoder: | |
outputs = outputs + (present_key_value,) | |
return outputs | |
# Copied from transformers.models.bert.modeling_bert.BertEncoder with Bert->Roberta | |
class RobertaEncoder(nn.Module): | |
def __init__(self, config): | |
super().__init__() | |
self.config = config | |
self.layer = nn.ModuleList([RobertaLayer(config) for _ in range(config.num_hidden_layers)]) | |
self.gradient_checkpointing = False | |
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_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, | |
decompx_config: Optional[DecompXConfig] = None, # added by Fayyaz / Modarressi | |
) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]: | |
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 | |
next_decoder_cache = () if use_cache else None | |
aggregated_encoder_norms = None # added by Fayyaz / Modarressi | |
aggregated_encoder_vectors = None # added by Fayyaz / Modarressi | |
# -- added by Fayyaz / Modarressi | |
if decompx_config and decompx_config.output_all_layers: | |
all_decompx_outputs = DecompXOutput( | |
attention=() if decompx_config.output_attention else None, | |
res1=() if decompx_config.output_res1 else None, | |
LN1=() if decompx_config.output_LN1 else None, | |
FFN=() if decompx_config.output_LN1 else None, | |
res2=() if decompx_config.output_res2 else None, | |
encoder=() if decompx_config.output_encoder else None, | |
aggregated=() if decompx_config.output_aggregated and decompx_config.aggregation else None, | |
) | |
else: | |
all_decompx_outputs = None | |
# -- added by Fayyaz / Modarressi | |
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 | |
past_key_value = past_key_values[i] if past_key_values is not None else None | |
if self.gradient_checkpointing and self.training: | |
if use_cache: | |
logger.warning( | |
"`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..." | |
) | |
use_cache = False | |
def create_custom_forward(module): | |
def custom_forward(*inputs): | |
return module(*inputs, past_key_value, output_attentions) | |
return custom_forward | |
layer_outputs = torch.utils.checkpoint.checkpoint( | |
create_custom_forward(layer_module), | |
hidden_states, | |
attention_mask, | |
layer_head_mask, | |
encoder_hidden_states, | |
encoder_attention_mask, | |
) | |
else: | |
layer_outputs = layer_module( | |
hidden_states, | |
aggregated_encoder_vectors, | |
attention_mask, | |
layer_head_mask, | |
encoder_hidden_states, | |
encoder_attention_mask, | |
past_key_value, | |
output_attentions, | |
decompx_config # added by Fayyaz / Modarressi | |
) | |
hidden_states = layer_outputs[0] | |
if use_cache: | |
next_decoder_cache += (layer_outputs[-1],) | |
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],) | |
# added by Fayyaz / Modarressi | |
if decompx_config: | |
decompx_output = layer_outputs[1] | |
if decompx_config.aggregation == "rollout": | |
if decompx_config.include_classifier_w_pooler: | |
raise Exception("Classifier and pooler could be included in vector aggregation mode") | |
encoder_norms = decompx_output.encoder[0][0] | |
if aggregated_encoder_norms is None: | |
aggregated_encoder_norms = encoder_norms * torch.exp(attention_mask).view( | |
(-1, attention_mask.shape[-1], 1)) | |
else: | |
aggregated_encoder_norms = torch.einsum("ijk,ikm->ijm", encoder_norms, aggregated_encoder_norms) | |
if decompx_config.output_aggregated == "norm": | |
decompx_output.aggregated = (aggregated_encoder_norms,) | |
elif decompx_config.output_aggregated is not None: | |
raise Exception( | |
"Rollout aggregated values are only available in norms. Set output_aggregated to 'norm'.") | |
elif decompx_config.aggregation == "vector": | |
aggregated_encoder_vectors = decompx_output.encoder[0][1] | |
if decompx_config.include_classifier_w_pooler: | |
decompx_output.aggregated = (aggregated_encoder_vectors,) | |
else: | |
decompx_output.aggregated = output_builder(aggregated_encoder_vectors, | |
decompx_config.output_aggregated) | |
decompx_output.encoder = output_builder(decompx_output.encoder[0][1], decompx_config.output_encoder) | |
if decompx_config.output_all_layers: | |
all_decompx_outputs.attention = all_decompx_outputs.attention + decompx_output.attention if decompx_config.output_attention else None | |
all_decompx_outputs.res1 = all_decompx_outputs.res1 + decompx_output.res1 if decompx_config.output_res1 else None | |
all_decompx_outputs.LN1 = all_decompx_outputs.LN1 + decompx_output.LN1 if decompx_config.output_LN1 else None | |
all_decompx_outputs.FFN = all_decompx_outputs.FFN + decompx_output.FFN if decompx_config.output_FFN else None | |
all_decompx_outputs.res2 = all_decompx_outputs.res2 + decompx_output.res2 if decompx_config.output_res2 else None | |
all_decompx_outputs.encoder = all_decompx_outputs.encoder + decompx_output.encoder if decompx_config.output_encoder else None | |
if decompx_config.include_classifier_w_pooler and decompx_config.aggregation == "vector": | |
all_decompx_outputs.aggregated = all_decompx_outputs.aggregated + output_builder( | |
aggregated_encoder_vectors, | |
decompx_config.output_aggregated) if decompx_config.output_aggregated else None | |
else: | |
all_decompx_outputs.aggregated = all_decompx_outputs.aggregated + decompx_output.aggregated if decompx_config.output_aggregated else None | |
if output_hidden_states: | |
all_hidden_states = all_hidden_states + (hidden_states,) | |
if not return_dict: | |
return tuple( | |
v | |
for v in [ | |
hidden_states, | |
next_decoder_cache, | |
all_hidden_states, | |
all_self_attentions, | |
all_cross_attentions, | |
decompx_output if decompx_config else None, | |
all_decompx_outputs | |
] | |
if v is not None | |
) | |
return BaseModelOutputWithPastAndCrossAttentions( | |
last_hidden_state=hidden_states, | |
past_key_values=next_decoder_cache, | |
hidden_states=all_hidden_states, | |
attentions=all_self_attentions, | |
cross_attentions=all_cross_attentions, | |
) | |
# Copied from transformers.models.bert.modeling_bert.BertPooler | |
class RobertaPooler(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] | |
pre_pooled_output = self.dense(first_token_tensor) | |
pooled_output = self.activation(pre_pooled_output) | |
return pooled_output | |
class RobertaPreTrainedModel(PreTrainedModel): | |
""" | |
An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained | |
models. | |
""" | |
config_class = RobertaConfig | |
base_model_prefix = "roberta" | |
supports_gradient_checkpointing = True | |
# Copied from transformers.models.bert.modeling_bert.BertPreTrainedModel._init_weights | |
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) | |
def _set_gradient_checkpointing(self, module, value=False): | |
if isinstance(module, RobertaEncoder): | |
module.gradient_checkpointing = value | |
def update_keys_to_ignore(self, config, del_keys_to_ignore): | |
"""Remove some keys from ignore list""" | |
if not config.tie_word_embeddings: | |
# must make a new list, or the class variable gets modified! | |
self._keys_to_ignore_on_save = [k for k in self._keys_to_ignore_on_save if k not in del_keys_to_ignore] | |
self._keys_to_ignore_on_load_missing = [ | |
k for k in self._keys_to_ignore_on_load_missing if k not in del_keys_to_ignore | |
] | |
ROBERTA_START_DOCSTRING = r""" | |
This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the | |
library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads | |
etc.) | |
This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. | |
Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage | |
and behavior. | |
Parameters: | |
config ([`RobertaConfig`]): Model configuration class with all the parameters of the | |
model. Initializing with a config file does not load the weights associated with the model, only the | |
configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights. | |
""" | |
ROBERTA_INPUTS_DOCSTRING = r""" | |
Args: | |
input_ids (`torch.LongTensor` of shape `({0})`): | |
Indices of input sequence tokens in the vocabulary. | |
Indices can be obtained using [`RobertaTokenizer`]. See [`PreTrainedTokenizer.encode`] and | |
[`PreTrainedTokenizer.__call__`] for details. | |
[What are input IDs?](../glossary#input-ids) | |
attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*): | |
Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: | |
- 1 for tokens that are **not masked**, | |
- 0 for tokens that are **masked**. | |
[What are attention masks?](../glossary#attention-mask) | |
token_type_ids (`torch.LongTensor` of shape `({0})`, *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) | |
position_ids (`torch.LongTensor` of shape `({0})`, *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) | |
head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*): | |
Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`: | |
- 1 indicates the head is **not masked**, | |
- 0 indicates the head is **masked**. | |
inputs_embeds (`torch.FloatTensor` of shape `({0}, 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. | |
output_attentions (`bool`, *optional*): | |
Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned | |
tensors for more detail. | |
output_hidden_states (`bool`, *optional*): | |
Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for | |
more detail. | |
return_dict (`bool`, *optional*): | |
Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. | |
""" | |
class RobertaModel(RobertaPreTrainedModel): | |
""" | |
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://arxiv.org/abs/1706.03762 | |
""" | |
_keys_to_ignore_on_load_missing = [r"position_ids"] | |
# Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->Roberta | |
def __init__(self, config, add_pooling_layer=True): | |
super().__init__(config) | |
self.config = config | |
self.embeddings = RobertaEmbeddings(config) | |
self.encoder = RobertaEncoder(config) | |
self.pooler = RobertaPooler(config) if add_pooling_layer else None | |
# Initialize weights and apply final processing | |
self.post_init() | |
def get_input_embeddings(self): | |
return self.embeddings.word_embeddings | |
def set_input_embeddings(self, value): | |
self.embeddings.word_embeddings = value | |
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) | |
# Copied from transformers.models.bert.modeling_bert.BertModel.forward | |
def forward( | |
self, | |
input_ids: Optional[torch.Tensor] = 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, | |
decompx_config: Optional[DecompXConfig] = None, # added by Fayyaz / Modarressi | |
) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]: | |
r""" | |
encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): | |
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if | |
the model is configured as a decoder. | |
encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): | |
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in | |
the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`: | |
- 1 for tokens that are **not masked**, | |
- 0 for tokens that are **masked**. | |
past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): | |
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. | |
If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that | |
don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all | |
`decoder_input_ids` of shape `(batch_size, sequence_length)`. | |
use_cache (`bool`, *optional*): | |
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see | |
`past_key_values`). | |
""" | |
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: | |
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 | |
past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0 | |
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, device) | |
# 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, | |
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, | |
decompx_config=decompx_config, # added by Fayyaz / Modarressi | |
) | |
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, | |
) | |
class RobertaForCausalLM(RobertaPreTrainedModel): | |
_keys_to_ignore_on_save = [r"lm_head.decoder.weight", r"lm_head.decoder.bias"] | |
_keys_to_ignore_on_load_missing = [r"position_ids", r"lm_head.decoder.weight", r"lm_head.decoder.bias"] | |
_keys_to_ignore_on_load_unexpected = [r"pooler"] | |
def __init__(self, config): | |
super().__init__(config) | |
if not config.is_decoder: | |
logger.warning("If you want to use `RobertaLMHeadModel` as a standalone, add `is_decoder=True.`") | |
self.roberta = RobertaModel(config, add_pooling_layer=False) | |
self.lm_head = RobertaLMHead(config) | |
# The LM head weights require special treatment only when they are tied with the word embeddings | |
self.update_keys_to_ignore(config, ["lm_head.decoder.weight"]) | |
# Initialize weights and apply final processing | |
self.post_init() | |
def get_output_embeddings(self): | |
return self.lm_head.decoder | |
def set_output_embeddings(self, new_embeddings): | |
self.lm_head.decoder = new_embeddings | |
def forward( | |
self, | |
input_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: 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, | |
) -> Union[Tuple, CausalLMOutputWithCrossAttentions]: | |
r""" | |
encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): | |
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if | |
the model is configured as a decoder. | |
encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): | |
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in | |
the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`: | |
- 1 for tokens that are **not masked**, | |
- 0 for tokens that are **masked**. | |
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]` | |
past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): | |
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. | |
If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that | |
don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all | |
`decoder_input_ids` of shape `(batch_size, sequence_length)`. | |
use_cache (`bool`, *optional*): | |
If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see | |
`past_key_values`). | |
Returns: | |
Example: | |
```python | |
>>> from transformers import RobertaTokenizer, RobertaForCausalLM, RobertaConfig | |
>>> import torch | |
>>> tokenizer = RobertaTokenizer.from_pretrained("roberta-base") | |
>>> config = RobertaConfig.from_pretrained("roberta-base") | |
>>> config.is_decoder = True | |
>>> model = RobertaForCausalLM.from_pretrained("roberta-base", config=config) | |
>>> 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, | |
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, | |
) | |
sequence_output = outputs[0] | |
prediction_scores = self.lm_head(sequence_output) | |
lm_loss = None | |
if labels is not None: | |
# we are doing next-token prediction; shift prediction scores and input ids by one | |
shifted_prediction_scores = prediction_scores[:, :-1, :].contiguous() | |
labels = labels[:, 1:].contiguous() | |
loss_fct = CrossEntropyLoss() | |
lm_loss = loss_fct(shifted_prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) | |
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, | |
) | |
def prepare_inputs_for_generation(self, input_ids, past=None, attention_mask=None, **model_kwargs): | |
input_shape = input_ids.shape | |
# if model is used as a decoder in encoder-decoder model, the decoder attention mask is created on the fly | |
if attention_mask is None: | |
attention_mask = input_ids.new_ones(input_shape) | |
# cut decoder_input_ids if past is used | |
if past is not None: | |
input_ids = input_ids[:, -1:] | |
return {"input_ids": input_ids, "attention_mask": attention_mask, "past_key_values": past} | |
def _reorder_cache(self, past, beam_idx): | |
reordered_past = () | |
for layer_past in past: | |
reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),) | |
return reordered_past | |
class RobertaForMaskedLM(RobertaPreTrainedModel): | |
_keys_to_ignore_on_save = [r"lm_head.decoder.weight", r"lm_head.decoder.bias"] | |
_keys_to_ignore_on_load_missing = [r"position_ids", r"lm_head.decoder.weight", r"lm_head.decoder.bias"] | |
_keys_to_ignore_on_load_unexpected = [r"pooler"] | |
def __init__(self, config): | |
super().__init__(config) | |
if config.is_decoder: | |
logger.warning( | |
"If you want to use `RobertaForMaskedLM` make sure `config.is_decoder=False` for " | |
"bi-directional self-attention." | |
) | |
self.roberta = RobertaModel(config, add_pooling_layer=False) | |
self.lm_head = RobertaLMHead(config) | |
# The LM head weights require special treatment only when they are tied with the word embeddings | |
self.update_keys_to_ignore(config, ["lm_head.decoder.weight"]) | |
# Initialize weights and apply final processing | |
self.post_init() | |
def get_output_embeddings(self): | |
return self.lm_head.decoder | |
def set_output_embeddings(self, new_embeddings): | |
self.lm_head.decoder = new_embeddings | |
def forward( | |
self, | |
input_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, MaskedLMOutput]: | |
r""" | |
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]` | |
kwargs (`Dict[str, any]`, optional, defaults to *{}*): | |
Used to hide legacy arguments that have been deprecated. | |
""" | |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict | |
outputs = self.roberta( | |
input_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, | |
) | |
class RobertaLMHead(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) | |
self.bias = self.decoder.bias | |
class RobertaForSequenceClassification(RobertaPreTrainedModel): | |
_keys_to_ignore_on_load_missing = [r"position_ids"] | |
def __init__(self, config): | |
super().__init__(config) | |
self.num_labels = config.num_labels | |
self.config = config | |
self.roberta = RobertaModel(config, add_pooling_layer=False) | |
self.classifier = RobertaClassificationHead(config) | |
# Initialize weights and apply final processing | |
self.post_init() | |
def tanh_linear_approximation(self, pre_act_pooled, post_act_pooled): | |
def tanh_deriv(x): | |
return 1 - torch.tanh(x) ** 2.0 | |
m = tanh_deriv(pre_act_pooled) | |
b = post_act_pooled - m * pre_act_pooled | |
return m, b | |
def tanh_la_decomposition(self, attribution_vectors, pre_act_pooled, post_act_pooled, bias_decomp_type): | |
m, b = self.tanh_linear_approximation(pre_act_pooled, post_act_pooled) | |
mx = attribution_vectors * m.unsqueeze(dim=-2) | |
if bias_decomp_type == "absdot": | |
weights = torch.abs(torch.einsum("bkd,bd->bk", mx, b)) | |
elif bias_decomp_type == "abssim": | |
weights = torch.abs(torch.nn.functional.cosine_similarity(mx, b, dim=-1)) | |
weights = (torch.norm(mx, dim=-1) != 0) * weights | |
elif bias_decomp_type == "norm": | |
weights = torch.norm(mx, dim=-1) | |
elif bias_decomp_type == "equal": | |
weights = (torch.norm(mx, dim=-1) != 0) * 1.0 | |
elif bias_decomp_type == "cls": | |
weights = torch.zeros(mx.shape[:-1], device=mx.device) | |
weights[:, 0] = 1.0 | |
weights = weights / (weights.sum(dim=-1, keepdim=True) + 1e-12) | |
weighted_bias = torch.einsum("bd,bk->bkd", b, weights) | |
return mx + weighted_bias | |
def tanh_zo_decomposition(self, attribution_vectors, pre_act_pooled, post_act_pooled): | |
m = post_act_pooled / (pre_act_pooled + 1e-12) | |
mx = attribution_vectors * m.unsqueeze(dim=-2) | |
return mx | |
def pooler_decomposer(self, attribution_vectors, pre_act_pooled, post_act_pooled, include_biases=True, | |
bias_decomp_type="absdot", tanh_approx_type="LA"): | |
post_pool = torch.einsum("ld,bsd->bsl", self.classifier.dense.weight, attribution_vectors) | |
if include_biases: | |
post_pool = self.bias_decomposer(self.classifier.dense.bias, post_pool, bias_decomp_type=bias_decomp_type) | |
if tanh_approx_type == "LA": | |
post_act_pool = self.tanh_la_decomposition(post_pool, pre_act_pooled, post_act_pooled, | |
bias_decomp_type=bias_decomp_type) | |
else: | |
post_act_pool = self.tanh_zo_decomposition(post_pool, pre_act_pooled, post_act_pooled) | |
return post_act_pool | |
def bias_decomposer(self, bias, attribution_vectors, bias_decomp_type="absdot"): | |
# Decomposes the input bias based on similarity to the attribution vectors | |
# Args: | |
# bias: a bias vector (all_head_size) | |
# attribution_vectors: the attribution vectors from token j to i (b, i, j, all_head_size) :: (batch, seq_length, seq_length, all_head_size) | |
if bias_decomp_type == "absdot": | |
weights = torch.abs(torch.einsum("bkd,d->bk", attribution_vectors, bias)) | |
elif bias_decomp_type == "abssim": | |
weights = torch.abs(torch.nn.functional.cosine_similarity(attribution_vectors, bias, dim=-1)) | |
weights = (torch.norm(attribution_vectors, dim=-1) != 0) * weights | |
elif bias_decomp_type == "norm": | |
weights = torch.norm(attribution_vectors, dim=-1) | |
elif bias_decomp_type == "equal": | |
weights = (torch.norm(attribution_vectors, dim=-1) != 0) * 1.0 | |
elif bias_decomp_type == "cls": | |
weights = torch.zeros(attribution_vectors.shape[:-1], device=attribution_vectors.device) | |
weights[:, 0] = 1.0 | |
elif bias_decomp_type == "dot": | |
weights = torch.einsum("bkd,d->bk", attribution_vectors, bias) | |
elif bias_decomp_type == "biastoken": | |
attribution_vectors[:, -1] = attribution_vectors[:, -1] + bias | |
return attribution_vectors | |
weights = weights / (weights.sum(dim=-1, keepdim=True) + 1e-12) | |
weighted_bias = torch.matmul(weights.unsqueeze(dim=-1), bias.unsqueeze(dim=0)) | |
return attribution_vectors + weighted_bias | |
def biastoken_decomposer(self, biastoken, attribution_vectors, bias_decomp_type="absdot"): | |
# Decomposes the input bias based on similarity to the attribution vectors | |
# Args: | |
# bias: a bias vector (all_head_size) | |
# attribution_vectors: the attribution vectors from token j to i (b, i, j, all_head_size) :: (batch, seq_length, seq_length, all_head_size) | |
if bias_decomp_type == "absdot": | |
weights = torch.abs(torch.einsum("bkd,bd->bk", attribution_vectors, biastoken)) | |
elif bias_decomp_type == "abssim": | |
weights = torch.abs(torch.nn.functional.cosine_similarity(attribution_vectors, biastoken, dim=-1)) | |
weights = (torch.norm(attribution_vectors, dim=-1) != 0) * weights | |
elif bias_decomp_type == "norm": | |
weights = torch.norm(attribution_vectors, dim=-1) | |
elif bias_decomp_type == "equal": | |
weights = (torch.norm(attribution_vectors, dim=-1) != 0) * 1.0 | |
elif bias_decomp_type == "cls": | |
weights = torch.zeros(attribution_vectors.shape[:-1], device=attribution_vectors.device) | |
weights[:, 0] = 1.0 | |
elif bias_decomp_type == "dot": | |
weights = torch.einsum("bkd,d->bk", attribution_vectors, biastoken) | |
weights = weights / (weights.sum(dim=-1, keepdim=True) + 1e-12) | |
weighted_bias = torch.matmul(weights.unsqueeze(dim=-1), biastoken.unsqueeze(dim=1)) | |
return attribution_vectors + weighted_bias | |
def ffn_decomposer(self, attribution_vectors, include_biases=True, bias_decomp_type="absdot"): | |
post_classifier = torch.einsum("ld,bkd->bkl", self.classifier.out_proj.weight, attribution_vectors) | |
if include_biases: | |
post_classifier = self.bias_decomposer(self.classifier.out_proj.bias, post_classifier, | |
bias_decomp_type=bias_decomp_type) | |
return post_classifier | |
def forward( | |
self, | |
input_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, | |
decompx_config: Optional[DecompXConfig] = None, # added by Fayyaz / Modarressi | |
) -> Union[Tuple, SequenceClassifierOutput]: | |
r""" | |
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, | |
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, | |
decompx_config=decompx_config | |
) | |
sequence_output = outputs[0] | |
logits, mid_classifier_outputs = self.classifier(sequence_output, decompx_ready=decompx_config is not None) | |
if decompx_config is not None: | |
pre_act_pooled = mid_classifier_outputs[0] | |
pooled_output = mid_classifier_outputs[1] | |
if decompx_config.include_classifier_w_pooler: | |
decompx_idx = -2 if decompx_config.output_all_layers else -1 | |
aggregated_attribution_vectors = outputs[decompx_idx].aggregated[0] | |
outputs[decompx_idx].aggregated = output_builder(aggregated_attribution_vectors, | |
decompx_config.output_aggregated) | |
pooler_decomposed = self.pooler_decomposer( | |
attribution_vectors=aggregated_attribution_vectors[:, 0], | |
pre_act_pooled=pre_act_pooled, | |
post_act_pooled=pooled_output, | |
include_biases=decompx_config.include_biases, | |
bias_decomp_type="biastoken" if decompx_config.include_bias_token else decompx_config.bias_decomp_type, | |
tanh_approx_type=decompx_config.tanh_approx_type | |
) | |
aggregated_attribution_vectors = pooler_decomposed | |
outputs[decompx_idx].pooler = output_builder(pooler_decomposed, decompx_config.output_pooler) | |
classifier_decomposed = self.ffn_decomposer( | |
attribution_vectors=aggregated_attribution_vectors, | |