# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
# Copyright 2018 The OpenAI Team Authors and 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.
from __future__ import annotations
import collections
from functools import partial
from typing import Any, Dict, List
import numpy as np
import paddle
import paddle.nn as nn
import paddle.nn.functional as F
import paddle.tensor as tensor
from paddle.distributed import fleet
from paddle.nn import Layer
from paddle.nn.layer.transformer import _convert_param_attr_to_list
from paddlenlp.transformers.conversion_utils import StateDictNameMapping
from paddlenlp.transformers.model_utils import PretrainedModel, register_base_model
from paddlenlp.utils.log import logger
from ..model_outputs import (
BaseModelOutputWithPastAndCrossAttentions,
CausalLMOutputWithCrossAttentions,
)
from .configuration import (
OPT_PRETRAINED_INIT_CONFIGURATION,
OPT_PRETRAINED_RESOURCE_FILES_MAP,
OPTConfig,
)
__all__ = ["OPTModel", "OPTPretrainedModel", "OPTForCausalLM", "OPTForConditionalGeneration"]
def finfo(dtype):
if dtype == "float32":
return np.finfo(np.float32)
if dtype == "float16":
return np.finfo(np.float16)
if dtype == "float64":
return np.finfo(np.float64)
def _make_causal_mask(input_ids_shape, past_key_values_length, dtype):
"""
Make causal mask used for self-attention.
"""
batch_size, target_length = input_ids_shape
mask = paddle.full((target_length, target_length), float(finfo(paddle.get_default_dtype()).min))
mask_cond = paddle.arange(mask.shape[-1])
mask_cond = mask_cond < (mask_cond + 1).reshape([mask.shape[-1], 1])
mask = paddle.where(mask_cond, paddle.full(mask_cond.shape, 0), mask)
if past_key_values_length > 0:
mask = paddle.concat([paddle.zeros([target_length, past_key_values_length], dtype=mask.dtype), mask], axis=-1)
expanded_mask = mask.unsqueeze(0).expand([batch_size, 1, target_length, target_length + past_key_values_length])
return expanded_mask
def _expand_mask(mask, tgt_length):
"""
Expands attention_mask from `[batch_size, src_length]` to `[batch_size, 1, tgt_length, src_length]`.
"""
batch_size, src_length = mask.shape[0], mask.shape[-1]
tgt_length = tgt_length if tgt_length is not None else src_length
expanded_mask = ~(paddle.cast(mask[:, None, None, :], "bool"))
expanded_mask = paddle.cast(expanded_mask, dtype=paddle.get_default_dtype())
expanded_mask = expanded_mask.expand([batch_size, 1, tgt_length, src_length])
expanded_mask = expanded_mask * float(finfo(paddle.get_default_dtype()).min)
return expanded_mask
class MultiHeadAttention(nn.Layer):
"""
Attention mapps queries and a set of key-value pairs to outputs, and
Multi-Head Attention performs multiple parallel attention to jointly attending
to information from different representation subspaces.
"""
Cache = collections.namedtuple("Cache", ["k", "v"])
StaticCache = collections.namedtuple("StaticCache", ["k", "v"])
def __init__(
self,
config: OPTConfig,
need_weights=False,
):
super(MultiHeadAttention, self).__init__()
self.num_heads = config.num_attention_heads
self.head_dim = config.hidden_size // self.num_heads
# get the `num_heads`
assert self.num_heads % config.tensor_parallel_degree == 0
if config.tensor_parallel_degree > 0:
self.num_heads = self.num_heads // config.tensor_parallel_degree
assert (
self.head_dim * self.num_heads * config.tensor_parallel_degree == config.hidden_size
), "hidden_size must be divisible by num_heads"
self.dropout = config.attention_probs_dropout_prob
self.need_weights = need_weights
self.fuse_attention_qkv = config.fuse_attention_qkv
if config.tensor_parallel_degree > 1:
if self.fuse_attention_qkv:
self.qkv_proj = fleet.meta_parallel.ColumnParallelLinear(
config.hidden_size,
config.hidden_size * 3,
has_bias=True,
input_is_parallel=True,
)
else:
self.q_proj = fleet.meta_parallel.ColumnParallelLinear(
config.hidden_size,
config.hidden_size,
has_bias=True,
gather_output=False,
)
self.k_proj = fleet.meta_parallel.ColumnParallelLinear(
config.hidden_size,
config.hidden_size,
has_bias=True,
gather_output=False,
)
self.v_proj = fleet.meta_parallel.ColumnParallelLinear(
config.hidden_size,
config.hidden_size,
has_bias=True,
gather_output=False,
)
self.out_proj = fleet.meta_parallel.RowParallelLinear(
config.hidden_size, config.hidden_size, input_is_parallel=True, has_bias=True
)
else:
if self.fuse_attention_qkv:
self.qkv_proj = nn.Linear(config.hidden_size, 3 * config.hidden_size)
else:
self.q_proj = nn.Linear(config.hidden_size, config.hidden_size)
self.k_proj = nn.Linear(config.hidden_size, config.hidden_size)
self.v_proj = nn.Linear(config.hidden_size, config.hidden_size)
self.out_proj = nn.Linear(config.hidden_size, config.hidden_size)
def _fuse_prepare_qkv(self, query, use_cache=False, cache=None):
mix_layer = self.qkv_proj(query)
mix_layer = paddle.reshape_(mix_layer, [0, 0, self.num_heads, 3 * self.head_dim])
mix_layer = paddle.transpose(mix_layer, [0, 2, 1, 3])
q, k, v = paddle.split(mix_layer, num_or_sections=3, axis=-1)
assert not isinstance(cache, self.StaticCache), "cache currently does not support the StaticCache type"
if isinstance(cache, self.Cache):
# for decoder self-attention in inference
k = paddle.concat([cache.k, k], axis=2)
v = paddle.concat([cache.v, v], axis=2)
if use_cache is True:
cache = self.Cache(k, v)
return (q, k, v, cache) if use_cache else (q, k, v, None)
def _prepare_qkv(self, query, key, value, use_cache=False, cache=None):
r"""
Prapares linear projected queries, keys and values for usage of subsequnt
multiple parallel attention. If `cache` is not None, using cached results
to reduce redundant calculations.
"""
q = self.q_proj(query)
q = paddle.reshape(x=q, shape=[0, 0, self.num_heads, self.head_dim])
q = paddle.transpose(x=q, perm=[0, 2, 1, 3])
if isinstance(cache, self.StaticCache):
# for encoder-decoder attention in inference and has cached
k, v = cache.k, cache.v
else:
k, v = self.compute_kv(key, value)
if isinstance(cache, self.Cache):
# for decoder self-attention in inference
k = paddle.concat([cache.k, k], axis=2)
v = paddle.concat([cache.v, v], axis=2)
if use_cache is True:
cache = self.Cache(k, v)
return (q, k, v, None) if use_cache is False else (q, k, v, cache)
def compute_kv(self, key, value):
r"""
Applies linear projection on input keys and values, then splits heads
(reshape and transpose) to get keys and values from different representation
subspaces. The results are used as key-values pairs for subsequent multiple
parallel attention.
It is part of calculations in multi-head attention, and is provided as
a method to pre-compute and prefetch these results, thus we can use them
to construct cache for inference.
"""
k = self.k_proj(key)
v = self.v_proj(value)
k = tensor.reshape(x=k, shape=[0, 0, self.num_heads, self.head_dim])
k = tensor.transpose(x=k, perm=[0, 2, 1, 3])
v = tensor.reshape(x=v, shape=[0, 0, self.num_heads, self.head_dim])
v = tensor.transpose(x=v, perm=[0, 2, 1, 3])
return k, v
def gen_cache(self, key, value=None, type=Cache):
"""
Generates cache for `forward` usage in inference accroding to arguments.
The generated cache is an instance of `MultiHeadAttention.Cache` or an
instance of `MultiHeadAttention.StaticCache`.
"""
if type == MultiHeadAttention.StaticCache: # static_kv
k, v = self.compute_kv(key, value)
return self.StaticCache(k, v)
elif value is None: # incremental_state
k = paddle.full(shape=[key.shape[0], self.num_heads, 0, self.head_dim], dtype=key.dtype, fill_value=0)
v = paddle.full(shape=[key.shape[0], self.num_heads, 0, self.head_dim], dtype=key.dtype, fill_value=0)
return self.Cache(k, v)
else:
# incremental_state with initial value, mainly for usage like UniLM
return self.Cache(key, value)
def forward(self, query, key, value, attn_mask=None, use_cache=False, cache=None):
r"""
Applies multi-head attention to map queries and a set of key-value pairs
to outputs.
"""
key = query if key is None else key
value = query if value is None else value
if self.fuse_attention_qkv:
q, k, v, cache = self._fuse_prepare_qkv(query, use_cache, cache)
else:
q, k, v, cache = self._prepare_qkv(query, key, value, use_cache, cache)
# scale dot product attention
product = paddle.matmul(x=q * (self.head_dim**-0.5), y=k, transpose_y=True)
if attn_mask is not None:
product = product + attn_mask
weights = F.softmax(product)
if self.dropout:
weights = F.dropout(weights, self.dropout, training=self.training, mode="upscale_in_train")
out = tensor.matmul(weights, v)
# combine heads
out = tensor.transpose(out, perm=[0, 2, 1, 3])
out = tensor.reshape(x=out, shape=[0, 0, out.shape[2] * out.shape[3]])
# project to output
out = self.out_proj(out)
outs = [out]
if self.need_weights:
outs.append(weights)
if use_cache:
outs.append(cache)
return out if len(outs) == 1 else tuple(outs)
class TransformerDecoderLayer(nn.Layer):
"""
The transformer decoder layer.
It contains multiheadattention and some linear layers.
"""
def __init__(self, config):
d_model = config.hidden_size
dim_feedforward = config.intermediate_size
dropout = config.hidden_dropout_prob
activation = config.hidden_act
attn_dropout = config.attention_probs_dropout_prob
act_dropout = config.hidden_dropout_prob
normalize_before = getattr(config, "normalize_before", True)
weight_attr = paddle.ParamAttr(initializer=nn.initializer.Normal(mean=0.0, std=config.initializer_range))
bias_attr = None
self._config = locals()
self._config.pop("self")
self._config.pop("__class__", None) # py3
super(TransformerDecoderLayer, self).__init__()
attn_dropout = dropout if attn_dropout is None else attn_dropout
act_dropout = dropout if act_dropout is None else act_dropout
self.normalize_before = normalize_before
weight_attrs = _convert_param_attr_to_list(weight_attr, 3)
bias_attrs = _convert_param_attr_to_list(bias_attr, 3)
self.self_attn = MultiHeadAttention(config, need_weights=True)
if config.tensor_parallel_degree > 1:
self.linear1 = fleet.meta_parallel.ColumnParallelLinear(
d_model,
dim_feedforward,
gather_output=False,
has_bias=True,
)
else:
self.linear1 = nn.Linear(d_model, dim_feedforward, weight_attrs[2], bias_attr=bias_attrs[2])
if config.tensor_parallel_degree > 1:
self.linear2 = fleet.meta_parallel.RowParallelLinear(
dim_feedforward,
d_model,
input_is_parallel=True,
has_bias=True,
)
else:
self.linear2 = nn.Linear(dim_feedforward, d_model, weight_attrs[2], bias_attr=bias_attrs[2])
self.norm1 = nn.LayerNorm(d_model, epsilon=1e-5)
self.norm2 = nn.LayerNorm(d_model, epsilon=1e-5)
self.dropout1 = nn.Dropout(dropout, mode="upscale_in_train")
self.dropout2 = nn.Dropout(act_dropout, mode="upscale_in_train")
if activation == "gelu":
self.activation = nn.GELU(approximate=True)
else:
self.activation = getattr(F, activation)
def forward(self, tgt, memory, tgt_mask=None, use_cache=False, cache=None, output_attentions=False):
residual = tgt
if self.normalize_before:
tgt = self.norm1(tgt)
# self.self_attn(...) --> hidden_states, weights, (cache)
if use_cache is False:
tgt, attn_weights = self.self_attn(tgt, tgt, tgt, tgt_mask, use_cache, cache)
else:
tgt, attn_weights, incremental_cache = self.self_attn(tgt, tgt, tgt, tgt_mask, use_cache, cache)
tgt = residual + self.dropout1(tgt)
if not self.normalize_before:
tgt = self.norm1(tgt)
residual = tgt
if self.normalize_before:
tgt = self.norm2(tgt)
tgt = self.dropout2(self.linear2(self.activation(self.linear1(tgt))))
tgt = residual + tgt
if not self.normalize_before:
tgt = self.norm2(tgt)
if not (output_attentions or use_cache):
return tgt
temp_list = [tgt, attn_weights if output_attentions else None, incremental_cache if use_cache else None]
return tuple(v for v in temp_list if v is not None)
def gen_cache(self, memory):
incremental_cache = self.self_attn.gen_cache(memory, type=self.self_attn.Cache)
return incremental_cache
class TransformerDecoder(Layer):
"""
TransformerDecoder is a stack of N decoder layers.
"""
def __init__(self, config: OPTConfig, decoder_layers: List[Layer]):
super(TransformerDecoder, self).__init__()
if config.word_embed_proj_dim != config.hidden_size:
if config.tensor_parallel_degree > 1:
self.project_out = fleet.meta_parallel.ColumnParallelLinear(
config.hidden_size,
config.word_embed_proj_dim,
gather_output=True,
has_bias=False,
)
else:
self.project_out = nn.Linear(config.hidden_size, config.word_embed_proj_dim, bias_attr=False)
else:
self.project_out = None
self.num_layers = config.num_hidden_layers
self.layers = decoder_layers
if config.normalize_before:
self.final_layer_norm = nn.LayerNorm(config.hidden_size)
else:
self.final_layer_norm = None
self.checkpoints = []
def forward(
self,
tgt,
memory,
tgt_mask=None,
memory_mask=None,
use_cache: bool = False,
cache=None,
output_attentions=False,
output_hidden_states=False,
return_dict=False,
):
r"""
Applies a stack of N Transformer decoder layers on inputs. If `norm` is
provided, also applies layer normalization on the output of last decoder
layer.
"""
output = tgt
new_caches = [] if use_cache else None
self.checkpoints = []
all_self_attentions = () if output_attentions else None
all_hidden_states = () if output_hidden_states else None
for i, mod in enumerate(self.layers):
outputs = mod(
output,
memory,
tgt_mask=tgt_mask,
use_cache=use_cache,
cache=cache[i] if cache is not None else cache,
output_attentions=output_attentions,
)
# outputs = hidden_states if both use_cache and output_attentions are False
# Otherwise, outputs = (hidden_states, attention if output_attentions, cache if use_cache)
output = outputs[0] if (use_cache or output_attentions) else outputs
if output_attentions:
all_self_attentions = all_self_attentions + (outputs[1],)
if use_cache:
new_caches.append(outputs[-1])
if output_hidden_states:
all_hidden_states = all_hidden_states + (output,)
self.checkpoints.append(output.name)
if self.final_layer_norm:
output = self.final_layer_norm(output)
if self.project_out:
output = self.project_out(output)
if not return_dict:
temp_list = [output, new_caches, all_hidden_states, all_self_attentions]
if not (use_cache or output_attentions or output_hidden_states):
return output
return tuple(v for v in temp_list if v is not None)
return BaseModelOutputWithPastAndCrossAttentions(
last_hidden_state=output,
past_key_values=new_caches,
hidden_states=all_hidden_states,
attentions=all_self_attentions,
cross_attentions=None,
)
def gen_cache(self, memory, do_zip=False):
r"""
Generates cache for `forward` usage. The generated cache is a list, and
each element in it is a tuple( :code:`(incremental_cache, static_cache)` )
produced by `TransformerDecoderLayer.gen_cache`. See `TransformerDecoderLayer.gen_cache`
for more details. If `do_zip` is True, apply `zip` on these tuples to get
a list with two elements.
"""
cache = [layer.gen_cache(memory) for layer in self.layers]
if do_zip:
cache = list(zip(*cache))
return cache
class OPTLearnedPositionEmbedding(nn.Embedding):
"""this module learns postional embeddings up to a fixed maximum size"""
def __init__(self, num_embeddings: int, embedding_dim: int, initializer_range: float):
"""OPT is set up so that if padding_idx is specified then offset the embedding ids by 2
and adjust num_embeddings appropriately. Other models don't have this hack.
Args:
num_embeddings (int): the number of embedding size
embedding_dim (int): the dim of embedding
"""
self.offset = 2
super().__init__(num_embeddings + self.offset, embedding_dim)
def forward(self, attention_mask, past_key_values_length: int = 0):
"""get the position embedding with attention mask
Args:
attention_mask: (paddle.Tensor): # create positions depending on attention_mask
past_key_values_length (int, optional): the past key value which will . Defaults to 0.
Returns:
paddle.Tensor: the position embedding
"""
# create positions depending on attention_mask
if attention_mask.dtype not in [paddle.bool, paddle.int64]:
attention_mask = attention_mask == 1.0
position_ids = paddle.cumsum(paddle.cast(attention_mask, "int64"), axis=-1) - 1
# cut positions if `past_key_values_length` is > 0
position_ids = position_ids[:, past_key_values_length:]
return nn.Embedding.forward(self, position_ids + self.offset)
class OPTEmbeddings(Layer):
"""
Include embeddings from word and position embeddings.
"""
def __init__(self, config: OPTConfig):
super(OPTEmbeddings, self).__init__()
if config.tensor_parallel_degree > 1:
self.word_embeddings = fleet.meta_parallel.VocabParallelEmbedding(
config.vocab_size,
config.word_embed_proj_dim,
weight_attr=paddle.ParamAttr(
initializer=nn.initializer.Normal(mean=0.0, std=config.initializer_range)
),
)
else:
self.word_embeddings = nn.Embedding(
config.vocab_size,
config.word_embed_proj_dim,
# padding_idx=config.pad_token_id,
weight_attr=paddle.ParamAttr(
initializer=nn.initializer.Normal(mean=0.0, std=config.initializer_range)
),
)
if config.word_embed_proj_dim != config.hidden_size:
if config.tensor_parallel_degree > 1:
self.project_in = fleet.meta_parallel.ColumnParallelLinear(
config.word_embed_proj_dim,
config.hidden_size,
gather_output=True,
has_bias=False,
)
else:
self.project_in = nn.Linear(config.word_embed_proj_dim, config.hidden_size, bias_attr=False)
else:
self.project_in = None
self.position_embeddings = OPTLearnedPositionEmbedding(
num_embeddings=config.max_position_embeddings,
embedding_dim=config.hidden_size,
initializer_range=config.initializer_range,
)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
def forward(self, input_ids=None, attention_mask=None, input_embeddings=None, past_key_values_length=None):
if input_ids is not None and input_embeddings is None:
input_embeddings = self.word_embeddings(input_ids)
if self.project_in:
input_embeddings = self.project_in(input_embeddings)
position_embeddings = self.position_embeddings(attention_mask, past_key_values_length)
embeddings = input_embeddings + position_embeddings
embeddings = self.dropout(embeddings)
return embeddings
[docs]
class OPTPretrainedModel(PretrainedModel):
"""
An abstract class for pretrained OPT models. It provides OPT related
`model_config_file`, `resource_files_names`, `pretrained_resource_files_map`,
`pretrained_init_configuration`, `base_model_prefix` for downloading and
loading pretrained models.
See :class:`~paddlenlp.transformers.model_utils.PretrainedModel` for more details.
"""
config_class = OPTConfig
base_model_prefix = "opt"
pretrained_init_configuration = OPT_PRETRAINED_INIT_CONFIGURATION
pretrained_resource_files_map = OPT_PRETRAINED_RESOURCE_FILES_MAP
@classmethod
def _get_tensor_parallel_mappings(cls, config: OPTConfig, is_split=True):
from paddlenlp.transformers.conversion_utils import split_or_merge_func
fn = split_or_merge_func(
is_split=is_split,
tensor_parallel_degree=config.tensor_parallel_degree,
tensor_parallel_rank=config.tensor_parallel_rank,
num_attention_heads=config.num_attention_heads,
)
actions = {
"embeddings.word_embeddings.weight": partial(fn, is_column=False),
}
for layer_index in range(config.num_hidden_layers):
actions.update(
{
# Column Linear
f"decoder.layers.{layer_index}.self_attn.q_proj.weight": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.self_attn.q_proj.bias": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.self_attn.k_proj.weight": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.self_attn.k_proj.bias": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.self_attn.v_proj.weight": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.self_attn.v_proj.bias": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.linear1.weight": partial(fn, is_column=True),
f"decoder.layers.{layer_index}.linear1.bias": partial(fn, is_column=True),
# Row Linear
f"decoder.layers.{layer_index}.linear2.weight": partial(fn, is_column=False),
f"decoder.layers.{layer_index}.self_attn.out_proj.weight": partial(fn, is_column=False),
}
)
if config.word_embed_proj_dim != config.hidden_size:
actions.update(
{
"decoder.project_out.weight": partial(fn, is_column=True),
"decoder.project_in.weight": partial(fn, is_column=True),
}
)
if cls.__name__ != "OPTModel":
for key in list(actions.keys()):
actions["opt." + key] = actions.pop(key)
return actions
@classmethod
def _get_fuse_or_split_param_mappings(cls, config: OPTConfig, is_fuse=False):
# return parameter fuse utils
from paddlenlp.transformers.conversion_utils import split_or_fuse_func
fn = split_or_fuse_func(is_fuse=is_fuse)
# last key is fused key, other keys are to be fused.
fuse_qkv_keys = (
"decoder.layers.0.self_attn.q_proj.weight",
"decoder.layers.0.self_attn.k_proj.weight",
"decoder.layers.0.self_attn.v_proj.weight",
"decoder.layers.0.self_attn.qkv_proj.weight",
)
fuse_qkv_bias_keys = (
"decoder.layers.0.self_attn.q_proj.bias",
"decoder.layers.0.self_attn.k_proj.bias",
"decoder.layers.0.self_attn.v_proj.bias",
"decoder.layers.0.self_attn.qkv_proj.bias",
)
num_heads = config.num_attention_heads
num_key_value_heads = getattr(config, "num_key_value_heads", num_heads)
fuse_attention_qkv = getattr(config, "fuse_attention_qkv", False)
final_actions = {}
if is_fuse:
if fuse_attention_qkv:
for i in range(config.num_hidden_layers):
for keys in [fuse_qkv_keys, fuse_qkv_bias_keys]:
new_keys = tuple([key.replace("layers.0.", f"layers.{i}.") for key in keys])
final_actions[new_keys] = partial(
fn, is_qkv=True, num_heads=num_heads, num_key_value_heads=num_key_value_heads
)
else:
if not fuse_attention_qkv:
for i in range(config.num_hidden_layers):
for keys in [fuse_qkv_keys, fuse_qkv_bias_keys]:
new_keys = tuple([key.replace("layers.0.", f"layers.{i}.") for key in keys])
final_actions[new_keys] = partial(
fn, split_nums=3, is_qkv=True, num_heads=num_heads, num_key_value_heads=num_key_value_heads
)
return final_actions
@classmethod
def _get_name_mappings(cls, config: OPTConfig) -> list[StateDictNameMapping]:
mappings: list[StateDictNameMapping] = []
model_mappings = [
["decoder.embed_tokens.weight", "embeddings.word_embeddings.weight"],
["decoder.embed_positions.weight", "embeddings.position_embeddings.weight"],
["decoder.final_layer_norm.weight", "decoder.final_layer_norm.weight"],
["decoder.final_layer_norm.bias", "decoder.final_layer_norm.bias"],
]
for layer_index in range(config.num_hidden_layers):
layer_mappings = [
[
f"decoder.layers.{layer_index}.self_attn.k_proj.weight",
f"decoder.layers.{layer_index}.self_attn.k_proj.weight",
"transpose",
],
[
f"decoder.layers.{layer_index}.self_attn.k_proj.bias",
f"decoder.layers.{layer_index}.self_attn.k_proj.bias",
],
[
f"decoder.layers.{layer_index}.self_attn.v_proj.weight",
f"decoder.layers.{layer_index}.self_attn.v_proj.weight",
"transpose",
],
[
f"decoder.layers.{layer_index}.self_attn.v_proj.bias",
f"decoder.layers.{layer_index}.self_attn.v_proj.bias",
],
[
f"decoder.layers.{layer_index}.self_attn.q_proj.weight",
f"decoder.layers.{layer_index}.self_attn.q_proj.weight",
"transpose",
],
[
f"decoder.layers.{layer_index}.self_attn.q_proj.bias",
f"decoder.layers.{layer_index}.self_attn.q_proj.bias",
],
[
f"decoder.layers.{layer_index}.self_attn.out_proj.weight",
f"decoder.layers.{layer_index}.self_attn.out_proj.weight",
"transpose",
],
[
f"decoder.layers.{layer_index}.self_attn.out_proj.bias",
f"decoder.layers.{layer_index}.self_attn.out_proj.bias",
],
[
f"decoder.layers.{layer_index}.self_attn_layer_norm.weight",
f"decoder.layers.{layer_index}.norm1.weight",
],
[
f"decoder.layers.{layer_index}.self_attn_layer_norm.bias",
f"decoder.layers.{layer_index}.norm1.bias",
],
[
f"decoder.layers.{layer_index}.fc1.weight",
f"decoder.layers.{layer_index}.linear1.weight",
"transpose",
],
[f"decoder.layers.{layer_index}.fc1.bias", f"decoder.layers.{layer_index}.linear1.bias"],
[
f"decoder.layers.{layer_index}.fc2.weight",
f"decoder.layers.{layer_index}.linear2.weight",
"transpose",
],
[f"decoder.layers.{layer_index}.fc2.bias", f"decoder.layers.{layer_index}.linear2.bias"],
[
f"decoder.layers.{layer_index}.final_layer_norm.weight",
f"decoder.layers.{layer_index}.norm2.weight",
],
[f"decoder.layers.{layer_index}.final_layer_norm.bias", f"decoder.layers.{layer_index}.norm2.bias"],
]
model_mappings.extend(layer_mappings)
# base-model prefix "OPTModel"
if cls.__name__ != "OPTModel":
for mapping in model_mappings:
mapping[0] = "model." + mapping[0]
mapping[1] = "opt." + mapping[1]
# downstream mappings
mappings = [StateDictNameMapping(*mapping, index=index) for index, mapping in enumerate(model_mappings)]
return mappings
def _init_weights(self, layer):
"""Initialization hook"""
if isinstance(layer, (nn.Linear, nn.Embedding)):
# In the dygraph mode, use the `set_value` to reset the parameter directly,
# and reset the `state_dict` to update parameter in static mode.
if isinstance(layer.weight, paddle.Tensor):
layer.weight.set_value(
paddle.tensor.normal(
mean=0.0,
std=self.initializer_range
if hasattr(self, "initializer_range")
else self.opt.config["initializer_range"],
shape=layer.weight.shape,
)
)
[docs]
@register_base_model
class OPTModel(OPTPretrainedModel):
r"""
The bare OPT Model transformer outputting raw hidden-states.
This model inherits from :class:`~paddlenlp.transformers.model_utils.PretrainedModel`.
Refer to the superclass documentation for the generic methods.
This model is also a Paddle `paddle.nn.Layer <https://www.paddlepaddle.org.cn/documentation
/docs/zh/api/paddle/nn/Layer_cn.html>`__ subclass. Use it as a regular Paddle Layer
and refer to the Paddle documentation for all matter related to general usage and behavior.
Args:
config (:class:`OPTConfig`):
An instance of OPTConfig used to construct OPTModel.
"""
def __init__(self, config: OPTConfig):
super(OPTModel, self).__init__(config)
self.pad_token_id = config.pad_token_id
self.initializer_range = config.initializer_range
self.hidden_size = config.hidden_size
self.vocab_size = config.vocab_size
self.embeddings = OPTEmbeddings(config)
config.fuse_attention_qkv = False
decoder_layers = nn.LayerList()
for i in range(config.num_hidden_layers):
decoder_layers.append(TransformerDecoderLayer(config))
self.decoder = TransformerDecoder(config, decoder_layers)
self.checkpoints = []
def _prepare_decoder_attention_mask(self, attention_mask, input_shape, past_key_values_length):
# create causal mask
# [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
combined_attention_mask = None
if input_shape[-1] > 1:
combined_attention_mask = _make_causal_mask(
input_shape, past_key_values_length=past_key_values_length, dtype=attention_mask.dtype
)
if attention_mask is not None:
# [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
expanded_attn_mask = _expand_mask(attention_mask, tgt_length=input_shape[-1])
if input_shape[-1] > 1:
combined_attention_mask = combined_attention_mask + expanded_attn_mask
else:
combined_attention_mask = expanded_attn_mask
return combined_attention_mask
[docs]
def forward(
self,
input_ids=None,
position_ids=None,
attention_mask=None,
inputs_embeds=None,
use_cache=False,
cache=None,
output_attentions=None,
output_hidden_states=None,
return_dict=None,
):
r"""
The OPTModel forward method, overrides the `__call__()` special method.
Args:
input_ids (Tensor):
Indices of input sequence tokens in the vocabulary. They are
numerical representations of tokens that build the input sequence.
Its data type should be `int64` and it has a shape of [batch_size, sequence_length].
position_ids(Tensor, optional):
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range ``[0,
max_position_embeddings - 1]``.
Shape as `(batch_size, num_tokens)` and dtype as int64. Defaults to `None`.
attention_mask (Tensor, optional):
Mask used in self attention to avoid performing attention to some unwanted positions,
usually the subsequent positions.
It is a tensor with shape broadcasted to `[batch_size, num_attention_heads, sequence_length, sequence_length]`.
For example, its shape can be [batch_size, sequence_length], [batch_size, sequence_length, sequence_length],
[batch_size, num_attention_heads, sequence_length, sequence_length].
Its data type should be float32.
The `masked` tokens have `-1e9` values, and the `unmasked` tokens have `0` values.
Defaults to `None`, which means nothing needed to be prevented attention to.
inputs_embeds (Tensor, optional):
Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation
of shape `(batch_size, sequence_length, hidden_size)`. 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.
Default to None.
use_cache (bool, optional):
Whether or not to use cache. Defaults to `False`. If set to `True`, key value states will be returned and
can be used to speed up decoding.
cache (list, optional):
It is a list, and each element in the list is a tuple `(incremental_cache, static_cache)`.
See `TransformerDecoder.gen_cache <https://github.com/PaddlePaddle/Paddle/blob/release/2.1/python/paddle/nn/layer/transformer.py#L1060>`__ for more details.
It is only used for inference and should be None for training.
Default to `None`.
output_attentions (bool, optional):
Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
tensors for more detail. Defaults to `None`.
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. Defaults to `None`.
return_dict (bool, optional):
Whether to return a :class:`~paddlenlp.transformers.model_outputs.BaseModelOutputWithPastAndCrossAttentions` object. If `False`, the output
will be a tuple of tensors. Defaults to `None`.
Returns:
Tensor: Returns tensor `encoder_output`, which is the output at the last layer of the model.
Its data type should be float32 and has a shape of [batch_size, sequence_length, hidden_size].
Example:
.. code-block::
import paddle
from paddlenlp.transformers import OPTModel, GPTTokenizer
tokenizer = GPTTokenizer.from_pretrained('facebook/opt-125m')
model = OPTModel.from_pretrained('facebook/opt-125m')
inputs = tokenizer("Welcome to use PaddlePaddle and PaddleNLimage.pngP!", return_token_type_ids=False)
inputs = {k:paddle.to_tensor([v]) for (k, v) in inputs.items()}
output = model(**inputs)
"""
if position_ids is not None:
logger.warning("position_ids has not required for OPTModel.")
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 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 = paddle.shape(input_ids)
input_ids = input_ids.reshape((-1, input_shape[-1]))
elif inputs_embeds is not None:
input_shape = paddle.shape(inputs_embeds)[:-1]
else:
raise ValueError("You have to specify either input_ids or inputs_embeds")
self.checkpoints = []
past_key_values_length = paddle.shape(cache[0].k)[2] if cache is not None else 0
seq_length_with_past = input_shape[-1] + past_key_values_length
if attention_mask is None:
attention_mask = paddle.ones((input_shape[0], seq_length_with_past), dtype=paddle.bool)
embedding_output = self.embeddings(
input_ids=input_ids,
attention_mask=attention_mask,
input_embeddings=inputs_embeds,
past_key_values_length=past_key_values_length,
)
attention_mask = self._prepare_decoder_attention_mask(attention_mask, input_shape, past_key_values_length)
attention_mask.stop_gradient = True
outputs = self.decoder.forward(
embedding_output,
memory=None,
tgt_mask=attention_mask,
use_cache=use_cache,
cache=cache,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
if output_hidden_states:
if return_dict:
outputs.hidden_states = (embedding_output,) + outputs.hidden_states
else:
# [last_hidden_state, caches, all_hidden_states, all_self_attentions]
idx = 2 if use_cache else 1
all_hidden_states = ((embedding_output,) + outputs[idx],)
outputs = outputs[:idx] + all_hidden_states + outputs[idx + 1 :]
self.checkpoints.extend(self.decoder.checkpoints)
return outputs
class OPTLMHead(Layer):
def __init__(self, config: OPTConfig, embedding_weights=None):
super(OPTLMHead, self).__init__()
self.config = config
self.decoder_weight = (
self.create_parameter(
default_initializer=paddle.nn.initializer.Uniform(low=-0.1, high=0.1),
shape=[config.vocab_size, config.hidden_size],
dtype=config.dtype,
is_bias=True,
)
if embedding_weights is None
else embedding_weights
)
def forward(self, hidden_states):
if isinstance(hidden_states, BaseModelOutputWithPastAndCrossAttentions):
hidden_states = hidden_states["last_hidden_state"]
logits = paddle.tensor.matmul(hidden_states, self.decoder_weight.cast(hidden_states.dtype), transpose_y=True)
return logits
[docs]
class OPTForCausalLM(OPTPretrainedModel):
"""
The OPT Model with a `language modeling` head on top.
Args:
config (:class:`OPTConfig`):
An instance of OPTConfig used to construct OPTModel.
"""
def __init__(self, config: OPTConfig):
super(OPTForCausalLM, self).__init__(config)
self.opt = OPTModel(config)
self.lm_head = OPTLMHead(
config,
)
def _get_model_inputs_spec(self, dtype: str):
return {
"input_ids": paddle.static.InputSpec(shape=[None, None], dtype="int64"),
}
[docs]
def forward(
self,
input_ids=None,
attention_mask=None,
inputs_embeds=None,
labels=None,
use_cache=False,
cache=None,
output_attentions=None,
output_hidden_states=None,
return_dict=None,
**kwargs,
):
r"""
Args:
input_ids (Tensor):
See :class:`OPTModel`.
attention_mask (Tensor, optional):
See :class:`OPTModel`.
inputs_embeds (Tensor, optional):
See :class:`GPTModel`.
use_cache (bool, optional):
See :class:`OPTModel`.
cache (Tensor, optional):
See :class:`OPTModel`.
labels (paddle.Tensor, optional):
A Tensor of shape `(batch_size, sequence_length)`.
Labels for language modeling. Note that the labels are shifted inside the model, i.e. you can set
`labels = input_ids` Indices are selected in `[-100, 0, ..., vocab_size]` All labels set to `-100`
are ignored (masked), the loss is only computed for labels in `[0, ..., vocab_size]`
Defaults to None.
output_attentions (bool, optional):
See :class:`GPTModel`.
output_hidden_states (bool, optional):
See :class:`GPTModel`.
return_dict (bool, optional):
See :class:`GPTModel`.
Returns:
Tensor or tuple: Returns tensor `logits` or tuple `(logits, cached_kvs)`. If `use_cache` is True,
tuple (`logits, cached_kvs`) will be returned. Otherwise, tensor `logits` will be returned.
`logits` is the output of the opt model.
`cache_kvs` is the cache output of opt model if `use_cache` is True.
Example:
.. code-block::
import paddle
from paddlenlp.transformers import OPTForCausalLM, GPTTokenizer
tokenizer = GPTTokenizer.from_pretrained('facebook/opt-125m')
model = OPTForCausalLM.from_pretrained('facebook/opt-125m')
inputs = tokenizer("Welcome to use PaddlePaddle and PaddleNLP!")
inputs = {k:paddle.to_tensor([v]) for (k, v) in inputs.items()}
output_ids, score = model.generate(input_ids=inputs['input_ids'])
print(tokenizer.batch_decode(output_ids[0]))
"""
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
outputs = self.opt(
input_ids,
attention_mask=attention_mask,
inputs_embeds=inputs_embeds,
use_cache=use_cache,
cache=cache,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
if use_cache:
encoder_outputs, cached_kvs = outputs[:2]
else:
encoder_outputs = outputs
logits = self.lm_head(encoder_outputs)
loss = None
if labels is not None:
loss = nn.functional.cross_entropy(logits, labels)
if not return_dict:
if not use_cache:
return (loss, logits) if loss is not None else logits
outputs = (logits,) + outputs[1:]
return ((loss,) + outputs) if loss is not None else outputs
return CausalLMOutputWithCrossAttentions(
loss=loss,
logits=logits,
past_key_values=outputs.past_key_values,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
cross_attentions=outputs.cross_attentions,
)
def prepare_fast_entry(self, kwargs: Dict[str, Any]):
# import FasterOPT at here to avoid cycling import
from paddlenlp.ops import FasterOPT
use_fp16_decoding = kwargs.get("use_fp16_decoding", False)
decode_strategy = kwargs.get("decode_strategy")
# decoding_lib can be passed into FasterOPT
decoding_lib = kwargs.get("decoding_lib", None)
if decode_strategy == "beam_search":
raise AttributeError("'beam_search' is not supported yet in the fast version of OPT")
# Currently, FasterTransformer only support restricted size_per_head.
size_per_head = self.opt.config["hidden_size"] // self.opt.config["num_attention_heads"]
if size_per_head not in [32, 64, 80, 96, 128]:
raise AttributeError(
"'size_per_head = %d' is not supported yet in the fast version of OPT" % size_per_head
)
if kwargs["forced_bos_token_id"] is not None:
# not support for forced_bos_token_id yet in the fast version
raise AttributeError("'forced_bos_token_id != None' is not supported yet in the fast version")
if kwargs["min_length"] != 0:
# not support for min_length yet in the fast version
raise AttributeError("'min_length != 0' is not supported yet in the fast version")
self._fast_entry = FasterOPT(self, use_fp16_decoding=use_fp16_decoding, decoding_lib=decoding_lib).forward
return self._fast_entry
def prepare_inputs_for_generation(
self, input_ids, use_cache=False, cache=None, attention_mask=None, inputs_embeds=None, **kwargs
):
if cache is not None:
input_ids = input_ids[:, -1:]
# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
if inputs_embeds is not None and cache is None:
model_inputs = {"inputs_embeds": inputs_embeds}
else:
model_inputs = {"input_ids": input_ids}
model_inputs.update(
{
"cache": cache,
"use_cache": True,
"attention_mask": attention_mask,
}
)
return model_inputs
@staticmethod
def prepare_attention_mask_for_generation(input_ids, pad_token_id, eos_token_id):
is_pad_token_in_inputs_ids = (pad_token_id is not None) and paddle.any(input_ids == pad_token_id).item()
is_pad_token_not_equal_to_eos_token_id = (eos_token_id is None) or (
(eos_token_id is not None) and (pad_token_id != eos_token_id)
)
if is_pad_token_in_inputs_ids and is_pad_token_not_equal_to_eos_token_id:
attention_mask = (input_ids != pad_token_id).astype("int64")
else:
attention_mask = paddle.ones_like(input_ids, dtype="int64")
return attention_mask
def __getattr__(self, name):
try:
return super().__getattr__(name)
except AttributeError as e:
try:
return getattr(getattr(self, self.base_model_prefix), name)
except AttributeError:
try:
return getattr(self, self.base_model_prefix).config[name]
except KeyError:
raise e
OPTForConditionalGeneration = OPTForCausalLM
