[BUG] <title> data fetch error image start token != image end tokens while finetuning the MiniCPM-V-2.6

是否已有关于该错误的issue或讨论？ | Is there an existing issue / discussion for this?

[X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions

该问题是否在FAQ中有解答？ | Is there an existing answer for this in FAQ?

[X] 我已经搜索过FAQ | I have searched FAQ

当前行为 | Current Behavior

When I ran the finetune_lora.sh script, the "data fetch error, image start token != image end tokens" kept popping out in terminal. I've tried the approach in other issue posts as well. It just triggered another error.

The dataset is prepared as the one in instruction. I''ve dive into the file already. No other \n is found in the conversion except the one after token.

Here is the finetune script I use.

#!/bin/bash

GPUS_PER_NODE=1
NNODES=1
NODE_RANK=0
MASTER_ADDR=localhost
MASTER_PORT=6001

MODEL="openbmb/MiniCPM-V-2_6-int4" # or openbmb/MiniCPM-V-2, openbmb/MiniCPM-Llama3-V-2_5
# ATTENTION: specify the path to your training data, which should be a json file consisting of a list of conversations.
# See the section for finetuning in README for more information.
DATA="/home/ubuntu/input_automation/MiniCPM-V/finetune/training_data.json"
EVAL_DATA="/home/ubuntu/input_automation/MiniCPM-V/finetune/training_data.json"
LLM_TYPE="minicpm" 
# if use openbmb/MiniCPM-V-2, please set LLM_TYPE=minicpm
#if use openbmb/MiniCPM-Llama3-V-2_5, please set LLM_TYPE=llama3

MODEL_MAX_Length=2048 # if conduct multi-images sft, please set MODEL_MAX_Length=4096

DISTRIBUTED_ARGS="
    --nproc_per_node $GPUS_PER_NODE \
    --nnodes $NNODES \
    --node_rank $NODE_RANK \
    --master_addr $MASTER_ADDR \
    --master_port $MASTER_PORT
"
torchrun $DISTRIBUTED_ARGS finetune.py  \
    --model_name_or_path $MODEL \
    --llm_type $LLM_TYPE \
    --data_path $DATA \
    --eval_data_path $EVAL_DATA \
    --remove_unused_columns false \
    --label_names "labels" \
    --prediction_loss_only false \
    --bf16 false \
    --bf16_full_eval false \
    --fp16 true \
    --fp16_full_eval true \
    --do_train \
    --do_eval \
    --tune_vision false \
    --tune_llm false \
    --use_lora true \
    --lora_target_modules "llm\..*layers\.\d+\.self_attn\.(q_proj|k_proj|v_proj|o_proj)" \
    --model_max_length $MODEL_MAX_Length \
    --max_slice_nums 9 \
    --max_steps 10000 \
    --eval_steps 1000 \
    --output_dir output/output__lora \
    --logging_dir output/output_lora \
    --logging_strategy "steps" \
    --per_device_train_batch_size 1 \
    --per_device_eval_batch_size 1 \
    --gradient_accumulation_steps 1 \
    --evaluation_strategy "steps" \
    --save_strategy "steps" \
    --save_steps 1000 \
    --save_total_limit 10 \
    --learning_rate 1e-6 \
    --weight_decay 0.1 \
    --adam_beta2 0.95 \
    --warmup_ratio 0.01 \
    --lr_scheduler_type "cosine" \
    --logging_steps 1 \
    --gradient_checkpointing true \
    --deepspeed ds_config_zero2.json \
    --report_to "tensorboard" # wandb

期望行为 | Expected Behavior

The finetuning should work as expected.

复现方法 | Steps To Reproduce

No response

运行环境 | Environment

- OS:
- Python:
- Transformers:
- PyTorch:
- CUDA (`python -c 'import torch; print(torch.version.cuda)'`):

备注 | Anything else?