Reels · 1 of 55
1:05
Reel · 1:05
Unlimited-OCR 2 on 16 GB: it reads as well as Datalab's Chandra 2, but it is not unlimited on my card
Narrated: Baidu's Unlimited-OCR 2 against DeepSeek-OCR and Tesseract on a 16 GB GPU. On real pages it tied at the top with Datalab's Chandra 2. The promise of unlimited documents: with transformers memory grows and runs out at 80 pages; with SGLang time per token stays flat, but the window mode hung and a single window cut the document without warning.
- Length
- 1:05
- Published
This reel sums up Unlimited-OCR 2 on 16 GB: it reads as well as Datalab's Chandra 2, but it is not unlimited on my card, where the method, the tables and what did not work are.