The model inference seems to be broken

#2
by sovitrath - opened

Getting weird outputs in both grounding and free ocr model for even simple documents which are working fine directly via Hugging Face.
I was going to fine-tune but don't want to fine-tune over a broken pipeline at the moment.
Uploading the image I tested among many along wih the output in the free ocr mode. Anyone else facing similar issue?
image_3

TRADER JOE’S
2001 Greenville Ave
Dallas TX 75206
Store #403 - (469) 334-0614

OPEN 8:00AM TO 9:00PM DAILY

R-CARROTS SHREDDED 10 OZ
R-CUCUMBERS PERSIAN 1 LB
BANANAS 4 1/2 LB
TOMATOES CRUSHED NO SALT
ORGANIC WHOLE NO SALT
MINI-PEARL TOMATOES
1.29
1.99
2.49
0.87
0.99
0.99
0.99
0.99
0.99
0.98
0.99
0.99
0.99
0.99
0.89
0.99
0.99
0.99
0.99
0.68
0.68
0.68

The text output goes on like this.

Sign up or log in to comment