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)