Implementation:OpenGVLab InternVL Trainer Save Model

Knowledge Sources	InternVL HuggingFace Trainer
Domains	Model_Deployment, Training
Last Updated	2026-02-07 00:00 GMT

Overview

HuggingFace Trainer's model saving method used to export trained InternVL models after training.

Description

The trainer.save_model() method serializes the trained InternVLChatModel and associated files to disk. This is a Wrapper Doc — the method comes from HuggingFace Transformers Trainer, configured for InternVL's multimodal model.

Usage

Called at the end of training in all InternVL training scripts. No additional configuration needed beyond output_dir in TrainingArguments.

Code Reference

Source Location

• Repository: InternVL
• File: internvl_chat/internvl/train/internvl_chat_finetune.py
• Lines: L1057-1058

Signature

# Called after training completes:
trainer.save_model()
# Saves to training_args.output_dir

Import

from transformers import Trainer
# trainer.save_model() is a built-in method

External Reference

• HuggingFace Trainer Documentation

I/O Contract

Inputs

Name	Type	Required	Description
training_args.output_dir	str	Yes	Directory to save model files

Outputs

Name	Type	Description
Model directory	Directory	Contains config.json, model weights (safetensors), tokenizer files

Usage Examples

Save After Training

trainer = Trainer(model=model, args=training_args, ...)
trainer.train()
trainer.save_model()  # Saves to training_args.output_dir

Related Pages

Implements Principle

• Principle:OpenGVLab_InternVL_Model_Export

Requires Environment

• Environment:OpenGVLab_InternVL_PyTorch_CUDA