deepseek-ai / deepseek-ai/DeepSeek-OCR

Model starts counting off without stopping for some menu images, or repeats same phrases over and over again

Open
#257 3 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
23.9k
Forks
2.2k
PR merge metrics
No merged PRs in 30d

Description

Using the recommended settings on Huggingface for Free OCR, I am running a pizza restaurant menu image.

Gemini OCR gives me something like:

```
CLASSIC PIZZAS
CHEESE
SMALL $14
MEDIUM $16
LARGE $21
X-LARGE $25

PEPPERONI
SMALL $15
MEDIUM $17.5
LARGE $23
X-LARGE $27.5

HAWAIIAN
SMALL $16
MEDIUM $19
LARGE $25
X-LARGE $30

SPECIALTY PIZZAS
SMALL $17
MEDIUM $22
LARGE $28
X-LARGE $34
```

Whereas this model starts counting endlessly:
```
- CHEESE
- SMALL $14
- MEDIUM $16
- PEPPERONI
- SMALL $15
- MEDIUM $15
- HAMANIAN
- SMALL $14
- MEDIUM $16
- HAMANIAN
- SMALL $14
- MEDIUM $17
- HAMANIAN
- SMALL $14
- MEDIUM $18
- HAMANIAN
- SMALL $14
- MEDIUM $19
- HAMANIAN
- SMALL $14
- MEDIUM $20
- HAMANIAN
- SMALL $14
- MEDIUM $21
```
Up until
```
- HAMANIAN
- SMALL $14
- MEDIUM $293
```
It finally terminates once it hits the output token limit.

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.