Dataset Viewer
The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): ('text', {}), NamedSplit('validation'): ('text', {}), NamedSplit('test'): ('json', {})}
Error code:   FileFormatMismatchBetweenSplitsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

GROMAIVN

GROMAIVN is a multilingual AI training dataset developed for the GynixA.I project.

The dataset combines conversational text, programming-related content, multilingual text and image classification data. It is designed to support language model training, text generation, conversational AI and visual recognition experiments.

Dataset Information

Property Value
Dataset Name GROMAIVN
Developer Mixtralrde Cineten Ktinde
Project GynixA.I
Repository GRIMIAI-VNAnhdev/GROMAIVN
Dataset Type Multimodal
Main Content Text, conversations, code and images
Primary Language Vietnamese
Additional Languages English, Chinese and other languages
Training Format Text shards
Repository Visibility Private

Dataset Statistics

Based on the current dataset manifest:

Metric Value
Dataset version 9
Total processed data 10,561,588,876 bytes
Total storage Approximately 9.84 GiB
Training data 10,027,982,238 bytes
Validation data 533,606,638 bytes
Training shards 297
Validation shards 17
Source files scanned 679
Tracked code records 214,357
Train / Validation split 95% / 5%

The final number of unique records has not yet been reported by the current manifest. The size category is therefore an approximate metadata classification, not a verified final row count.

Dataset Contents

1. Multilingual Text

Contains text from multiple language collections, including:

  • Vietnamese
  • English
  • Chinese
  • French
  • German
  • Spanish
  • Indonesian
  • Japanese
  • Korean
  • Thai
  • Arabic
  • Other supported languages

2. Conversational Data

Includes conversational and instruction-oriented data designed to support:

  • Context-aware responses
  • Multi-turn conversations
  • Question answering
  • Instruction following
  • Assistant response generation

3. Programming and Code Data

Contains programming-related material used to support:

  • Code understanding
  • Code generation
  • Programming assistance
  • Technical question answering

The current manifest tracks 214,357 code records.

4. Image Data

Includes categorized image collections for visual recognition experiments.

Image categories include animals, humans and vehicles.

Data Processing

The current dataset pipeline includes:

  • Record-level SHA-256 deduplication.
  • Text normalization using Unicode NFKC.
  • Context-aware whitespace normalization.
  • Role-aware conversation segmentation.
  • Assistant-focused training data preparation.
  • Quality filtering.
  • Training and validation separation.
  • Manifest-based dataset verification.

The current manifest identifies the processing pipeline as grom_record_dedup and the quality profile as response_and_code_v4.

Dataset Structure

GROMAIVN/
β”‚
β”œβ”€β”€ data/
β”‚   β”œβ”€β”€ data/
β”‚   β”‚   β”œβ”€β”€ sources/
β”‚   β”‚   β”œβ”€β”€ raw_text/
β”‚   β”‚   └── external_raw/
β”‚   β”‚
β”‚   β”œβ”€β”€ dataset/
β”‚   β”‚   β”œβ”€β”€ train_shards/
β”‚   β”‚   └── val_shards/
β”‚   β”‚
β”‚   β”œβ”€β”€ dataset_extra/
β”‚   └── picture/
β”‚
β”œβ”€β”€ scripts/
β”œβ”€β”€ reports/
β”œβ”€β”€ tokenizer.json
β”œβ”€β”€ config.json
β”œβ”€β”€ train.py
β”œβ”€β”€ model.py
β”‚
└── trainer_source/
    β”œβ”€β”€ train.py
    β”œβ”€β”€ model.py
    β”œβ”€β”€ tokenizer.py
    β”œβ”€β”€ vision.py
    β”œβ”€β”€ longctx.py
    β”œβ”€β”€ grom_runtime.py
    β”œβ”€β”€ curate_append_quality_v2.py
    β”œβ”€β”€ config.json
    └── requirements.txt

Intended Applications

GROMAIVN is intended for research and development involving:

  • Multilingual language models.
  • Conversational AI.
  • Text generation.
  • Programming assistants.
  • Long-context language modeling.
  • Image classification experiments.
  • Multimodal AI research.

License and Data Usage

The repository combines data from multiple sources. Individual sources may have different licenses and usage conditions.

The overall dataset license has not yet been established. The unknown metadata value is used intentionally rather than claiming a license that has not been verified.

Users should review the original source licenses and usage restrictions before redistributing or using individual data collections.

Important Notes

  • This repository is private.
  • The dataset is intended for AI research and model development.
  • Dataset statistics may change as new data is added or processed.
  • The current manifest does not report a final unique-record count.
  • Dataset access requires Hugging Face repository permissions.

Project

GynixA.I β€” GROMAIVN

Developed by Mixtralrde Cineten Ktinde.

Hugging Face: https://huggingface.co/datasets/GRIMIAI-VNAnhdev/GROMAIVN

Downloads last month
12