微调后词表长度不一致怎么办

[topology1]

Reminder

• I have read the README and searched the existing issues.

System Info

您好，intern2.5-20B微调合并后出现了added_tokens.json，转换gguf模型的时候是成功的，加载的时候显示tensor 'token_embd.weight' has wrong shape; expected 6144, 92550, got 6144, 92544，好像是词表长度不一致，请问怎么办？

Reproduction

model

model_name_or_path: /home/denghui/share/model/org/Shanghai_AI_Laboratory/internlm2_5-20b-chat

method

stage: sft
do_train: true
finetuning_type: lora
lora_target: all

dataset

dataset: identity,alpaca_en_demo
template: intern2
cutoff_len: 1024
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16

output

output_dir: /home/denghui/share/model/wtgguf/internlm2.5
logging_steps: 5
save_steps: 1000
plot_loss: true
overwrite_output_dir: true

train

per_device_train_batch_size: 2
gradient_accumulation_steps: 8
learning_rate: 1.0e-4
num_train_epochs: 2.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

eval

val_size: 0.1
per_device_eval_batch_size: 1
eval_strategy: steps
eval_steps: 5

###add
flash_attn: auto
dataset_dir: data
dataset: our_nlp
max_grad_norm: 1.0
warmup_steps: 0
optim: adamw_torch
packing: False
report_to: none
include_num_input_tokens_seen: True
lora_rank: 8
lora_alpha: 16
lora_dropout: 0
以上是sft的参数文件
llamafactory-cli train train_lora/intern2.5-20_lora_sft.yaml

model

model_name_or_path: /home/denghui/share/model/org/Shanghai_AI_Laboratory/internlm2_5-20b-chat
adapter_name_or_path: /home/denghui/share/model/wtgguf/internlm2.5
template: intern2
finetuning_type: lora

export

export_dir: /home/denghui/share/model/wtgguf/intern
export_size: 2
export_device: cpu
export_legacy_format: false
以上是merge的文件
llamafactory-cli export merge_lora/intern_lora_sft.yaml
合并后出现了added_tokens.json，转用llama.cpp的convert_hf_to_gguf.py 转出来的模型，加载的时候崩溃了，
出现tensor 'token_embd.weight' has wrong shape; expected 6144, 92550, got 6144, 92544，请问应该怎么出来

Expected behavior

有参数能不改变词表长度吗

Others

或者有什么办法重新生成词表