benhaotang/
Nanonets-OCR-s:F16

5,484 2 months ago

OCR model by Nanonets that excels at turning anything into markdown

vision

2 months ago

f6ff1ea481f6 · 7.5GB

model
qwen25vl
·
3.75B
·
F16
template
{{ if .System }}<|im_start|>system {{ .System }}<|im_end|> {{ end }}{{ range .Messages }}{{ if eq .R
system
Extract the text from the above document as if you were reading it naturally. Return the tables in h
params
{ "num_ctx": 4096, "stop": [ "<|im_end|>" ] }

Readme

Quatized nanonets/Nanonets-OCR-s with ollama 0.8.0, a 6GB card can safely run this model at Q8

I only recommend running the default Q8 version unless RAM constrains, as Q4_K_M shows a notable degradation of performance in code block recognition

Nanonets-OCR-s demonstrates strong OCR ability across different domains, turns anything into markdown with proper math and todo list support, following are two examples generated with ollama 0.8.0 embeded in Msty

image.png image.png