[model][WIP]Support InternVL2_5 series

[Kuangdd01]

What does this PR do?

may Fix #6322 #6432 #6236 #3802

• But this PR built upon the PR in Transformers. yonigozlan:add-intern-vl
• The integration of this model into Transformers is still incomplete and may undergo further modifications.
• The current code has been validated on 1B small size model and demo data.

Before submitting

• [x] Did you read the contributor guideline?
• [x] Did you write any new necessary tests?

some demo experiment on InternVL2_5-1B-MPO-hf

1. video lora sft

### model
model_name_or_path: ../InternVL2_5-1B-MPO-hf
trust_remote_code: true

### method
stage: sft
do_train: true
finetuning_type: lora
lora_rank: 8
lora_target: all

### dataset
dataset: mllm_video_demo
# dataset: mllm_demo #, #identity,alpaca_en_demo  # video: mllm_video_demo
template: intern_vl
cutoff_len: 100000
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 4

### output
output_dir: saves/internvl-1b/lora/sft-test-demo-video
logging_steps: 1
save_steps: 500
plot_loss: true
overwrite_output_dir: true

### train
per_device_train_batch_size: 2
gradient_accumulation_steps: 1
learning_rate: 1.0e-4
num_train_epochs: 30.0
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: 500

2. mix data full sft

### model
model_name_or_path: /data_train/code/mllm/kuangdd/workspace/work/models/InternVL2_5-1B-MPO-hf
trust_remote_code: true

### method
stage: sft
do_train: true
finetuning_type: full


### dataset
dataset: mllm_demo, dentity, alpaca_en_demo  # video: mllm_video_demo
template: intern_vl
cutoff_len: 100000
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 4

### output
output_dir: saves/internvl-1b/full/sft-test-demo
logging_steps: 1
save_steps: 500
plot_loss: true
overwrite_output_dir: true

### train
per_device_train_batch_size: 2
gradient_accumulation_steps: 1
learning_rate: 1.0e-4
num_train_epochs: 30.0
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: 500

@hiyouga @BUAADreamer