---
license: apache-2.0
task_categories:
- visual-question-answering
- image-text-to-text
language:
- en
tags:
- vision
- multimodal
- regenerated
pretty_name: ALLaVA-4V Train Dataset (Regenerated)
size_categories:
- 100K<n<1M
---

# ALLaVA-4V Train Dataset (Regenerated)

This dataset is a regenerated version of the ALLaVA-4V training dataset, processed using **Qwen/Qwen3-VL-8B-Instruct** model.

## Dataset Details

- **Source**: ALLaVA-4V training data
- **Model Used**: [Qwen/Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct)
- **Original Format**: JSONL (252,924 samples)
- **Output Format**: Parquet
- **Temperature**: 0.0 (deterministic generation)
- **Processing Status**: In progress (~21% complete as of upload)

## Generation Details

The dataset was regenerated using the following configuration:
- Model: Qwen/Qwen3-VL-8B-Instruct
- Temperature: 0 (deterministic)
- Server: Auto-launched on port 30000
- Processing script: `regenerate_train_data.py`

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("vincent-4/allava4v-train-regenerated")
```

## Note

This dataset is being actively generated. The current version contains a partial subset of the full dataset. Check back for updates as processing continues.