Hugging Face's logo

kz-transformers
/
horde-vision

Image-Text-to-Text
Transformers
Safetensors
Kazakh
qwen3_vl
image-to-text
conversational
4-bit precision
bitsandbytes
Model card Files
xet
Community
horde-vision / README.md
kz-transformers's picture
kz-transformers
Update README.md
f446450 verified
preview code
|
raw
history blame contribute delete
4.9 kB
metadata
language:
  - kk
license: apache-2.0
base_model:
  - Qwen/Qwen3-VL-8B-Instruct
pipeline_tag: image-text-to-text
library_name: transformers

HordeVision: Open-Source Kazakh Vision-Language Model

HordeVision is a vision-language model specifically trained for the Kazakh language, designed to handle OCR, image captioning, visual question answering (VQA), reasoning, and instruction-following tasks.

Model Description

HordeVision is built to address the lack of vision-language models for low-resource languages like Kazakh. The model excels at:

  • Image Captioning: Generating detailed, contextual descriptions in Kazakh
  • Visual Question Answering (VQA): Answering diverse questions about image content
  • OCR: Extracting and reading Kazakh text from images
  • Visual Reasoning: Making inferences about context, causality, and temporal states
  • Instruction Following: Executing multi-step visual tasks based on user commands

Key Features

  • First open-source Kazakh vision-language model
  • Trained on ~50k culturally relevant images covering daily life, education, work, culture, and heritage
  • Two-stage training: Supervised Fine-Tuning (SFT) + Reinforcement Learning (GRPO)
  • Ranks #1 across all evaluation tasks compared to comparable multilingual models

Model Performance Summary

Model caption vqa ocr reason instruct_follow Avg Rank
horde-vision 83.5 (โ†‘12.3%) 68.1 (โ†‘5.3%) 64.7 (โ†‘2.6%) 77.4 (โ†‘5.7%) 70.5 (โ†‘5.9%) #1
Qolda 75.2 (โ†‘8.7%) 61.7 (โ†‘3.0%) 60.6 (โ†‘2.0%) 70.3 (โ†‘2.9%) 62.2 (โ†‘2.8%) #2
Qwen3-VL-8B-Instruct 41.3 (โ†‘0.5%) 53.6 (โ†‘1.1%) 59.3 (โ†‘2.1%) 55.5 (โ†‘0.7%) 49.5 (โ†‘0.9%) #3
gemma-3-4b-it 42.0 (โ†‘0.1%) 41.8 (โ†‘0.4%) 50.3 (โ†‘2.3%) 53.0 (โ†‘0.6%) 42.5 (โ†‘0.5%) #4
Qwen2.5-VL-7B-Instruct 35.4 (โ†‘0.0%) 41.6 (โ†‘0.4%) 51.0 (โ†‘0.9%) 44.6 (โ†‘0.3%) 37.7 (โ†‘0.3%) #5
Llama-3.2-11B-Vision 36.2 (โ†‘0.1%) 38.0 (โ†‘0.3%) 15.0 (โ†‘0.1%) 43.4 (โ†‘0.3%) 36.4 (โ†‘0.3%) #6
InternVL3-8B 26.1 (โ†‘0.6%) 29.0 (โ†‘0.0%) 29.1 (โ†‘0.3%) 27.3 (โ†‘0.0%) 25.7 (โ†‘0.0%) #7

Comparison: Outperforms Google Gemma 3-4B-IT, InternVL3-8B, Qwen2.5-VL-7B-Instruct, Qwen3-VL-8B-Instruct, and ISSAI Qolda across all tasks.

Dataset

The training dataset was collected using a syntactic data generation pipeline:

  • Size: 45k training images, 5k validation images
  • Categories: 21 main categories, 104 subcategories, ~2,600 keyword phrases
  • Coverage: Daily contexts, social life, education, work/economy, media/communications, culture and heritage
  • Quality: Filtered with imagededup for deduplication and aesthetic scoring
  • Annotation: Labeled using GPT-4.1 with structured prompts for consistent quality
  • Split Strategy: Entity-level stratification to ensure models are tested on completely unseen entities

Training Details

Supervised Fine-Tuning (SFT)

  • Data: 46k images
  • LoRA Rank: 128
  • Epochs: 1

Reinforcement Learning (GRPO)

  • Data: 5k images
  • LoRA Rank: 64
  • Epochs: 1
  • Judge: GPT-4.1-mini with custom Kazakh evaluation prompts

How to Use 

from transformers import Qwen3VLForConditionalGeneration, AutoProcessor

# default: Load the model on the available device(s)
model = Qwen3VLForConditionalGeneration.from_pretrained(
    "kz-transformers/horde-vision", dtype="auto", device_map="auto"
)

# We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
# model = Qwen3VLForConditionalGeneration.from_pretrained(
#     "kz-transformers/horde-vision",
#     dtype=torch.bfloat16,
#     attn_implementation="flash_attention_2",
#     device_map="auto",
# )

processor = AutoProcessor.from_pretrained("kz-transformers/horde-vision")

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
            },
            {"type": "text", "text": "ะ‘าฑะป ััƒั€ะตั‚ั‚ั– ัะธะฟะฐั‚ั‚ะฐาฃั‹ะท."},
        ],
    }
]

# Preparation for inference
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_dict=True,
    return_tensors="pt"
)
inputs = inputs.to(model.device)

# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
    out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)

Citation