mirror of
https://github.com/hiyouga/LLaMA-Factory.git
synced 2025-10-16 08:38:09 +08:00
57 lines
2.2 KiB
Python
57 lines
2.2 KiB
Python
import math
|
|
from contextlib import nullcontext
|
|
from typing import TYPE_CHECKING
|
|
|
|
import torch
|
|
from transformers.integrations import is_deepspeed_zero3_enabled
|
|
|
|
from ...extras.logging import get_logger
|
|
|
|
|
|
if TYPE_CHECKING:
|
|
from transformers import PreTrainedModel, PreTrainedTokenizer
|
|
|
|
|
|
logger = get_logger(__name__)
|
|
|
|
|
|
def _noisy_mean_initialization(embed_weight: torch.Tensor, num_new_tokens: int) -> None:
|
|
embedding_dim = embed_weight.size(1)
|
|
avg_weight = embed_weight[:-num_new_tokens].mean(dim=0, keepdim=True)
|
|
noise_weight = torch.empty_like(embed_weight[-num_new_tokens:])
|
|
noise_weight.normal_(mean=0, std=(1.0 / math.sqrt(embedding_dim)))
|
|
embed_weight[-num_new_tokens:] = avg_weight + noise_weight
|
|
|
|
|
|
def resize_embedding_layer(model: "PreTrainedModel", tokenizer: "PreTrainedTokenizer") -> None:
|
|
r"""
|
|
Resize token embeddings.
|
|
"""
|
|
if is_deepspeed_zero3_enabled():
|
|
import deepspeed # type: ignore
|
|
|
|
params = [model.get_input_embeddings().weight]
|
|
if model.get_output_embeddings() is not None and not model.config.tie_word_embeddings:
|
|
params.append(model.get_output_embeddings().weight)
|
|
|
|
context_maybe_zero3 = deepspeed.zero.GatheredParameters(params, modifier_rank=0)
|
|
else:
|
|
context_maybe_zero3 = nullcontext()
|
|
|
|
with context_maybe_zero3:
|
|
current_embedding_size = model.get_input_embeddings().weight.size(0)
|
|
|
|
if len(tokenizer) > current_embedding_size:
|
|
if not isinstance(model.get_output_embeddings(), torch.nn.Linear):
|
|
logger.warning("Current model does not support resizing token embeddings.")
|
|
return
|
|
|
|
model.resize_token_embeddings(len(tokenizer), pad_to_multiple_of=64)
|
|
with context_maybe_zero3:
|
|
new_embedding_size = model.get_input_embeddings().weight.size(0)
|
|
num_new_tokens = new_embedding_size - current_embedding_size
|
|
_noisy_mean_initialization(model.get_input_embeddings().weight.data, num_new_tokens)
|
|
_noisy_mean_initialization(model.get_output_embeddings().weight.data, num_new_tokens)
|
|
|
|
logger.info("Resized token embeddings from {} to {}.".format(current_embedding_size, new_embedding_size))
|