File size: 431 Bytes
6760278 b35fce4 6760278 7c07c27 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 |
---
license: mit
datasets:
- Vikhrmodels/GrandMaster-PRO-MAX
language:
- ru
- en
base_model:
- Qwen/Qwen2-VL-2B-Instruct
pipeline_tag: text2text-generation
tags:
- multimodal
---
# tvl-mini
## Description
This is finetune of Qwen2-VL-2B on russian language.
tvl was trained in bf16
## Data
Train dataset contains:
- GrandMaster-PRO-MAX dataset (60k samples)
- Translated, humanized and merged by image subset of GQA (TODO) |