Spaces:
Sleeping
Sleeping
Commit History
-
Switch to transformers5 SIE image; LightOnOCR as default recognition
Filip Makraduli- Base image: latest-cpu-default → latest-cpu-transformers5 (where the LightOnOCR adapter actually lives, fixes the ImportError on Space boot) - Recognition default: Florence-2-base-ft → LightOnOCR-2-1B (Florence-2 isn't loadable on transformers5 due to tokenizer API change; tracked in sie-internal#828) - Drop Florence-2 entries from the dropdown so users don't click into an error - Expand alternates: 3 Donut variants, 4 NER models (GLiNER multi/large/ PII + NuNER Zero) - Swap sample images: replace dense table/handwritten/multi-column with event-poster/slide/letter to better suit the new recognition model - Server: /api/health now reports cuda availability; UI auto-disables gpuRequired models on the CPU image4e0f10e
-
Switch default recognition to LightOnOCR; add Try these moments tour
Mirror of brave-new-demos commits 03b4c32 + 1a1b995. - Default recognition is now lightonai/LightOnOCR-2-1B (was Florence-2-base). Better Markdown on dense layouts; ~4 GB cold download. - New 'Try these moments' UI section above the main panels with four concrete prompts visitors can act on without reading the README. The HF entrypoint still has empty PRELOAD by default, so the Space still boots in ~30 s; first sample click will now trigger a LightOnOCR download on the cache miss.0df0841 -
Initial: document-ocr demo for HF Spaces
Single-container build of the document-ocr demo from superlinked/brave-new-demos. Extends ghcr.io/superlinked/sie-server:latest-cpu-default with Node 22, runs SIE on 127.0.0.1:8080 and the UI on 0.0.0.0:7860. Preloads three small models from the default bundle (Florence-2-base, Donut-CORD-v2, GLiNER-multi). Alternates are dropdown-selectable and lazy-load on first click. HF persistent /data is used as HF_HOME so model weights survive Space restarts.ffe59ba
