deepseek_vl_v2

Model type

deepseek_vl_v2

Class

DeepSeekOCR2CausalLMModel

Task

vision-language

Source

models/deepseek_ocr2.py

Description

DeepSeek-OCR-2 vision-language model (3-model split).

model_type: deepseek_vl_v2

Components:

  • vision_encoder: SAM ViT-B + Qwen2 encoder + Linear projector

  • embedding: embed_tokens + image feature scatter

  • decoder: DeepSeek-V2 LLM (standard attention + MoE)

Usage

mobius build --model <MODEL_ID> output_dir/

from mobius import build

model = build("<MODEL_ID>")