TinyDoc-VLM-768-checkpoints (ABANDONED)

Do not use. The 768px retrain failed: the vision tower was dead at initialization (constant output features for any input) while the decoder memorized text. Training was stopped; these step checkpoints are kept only as a record of the failed run.

The 256M 384px base model is also retired (measured OCRBench 0.0%, n=1,000) — see eulogik/TinyDoc-VLM-256M.

The working product of this project is the local grounded-extraction SDK: github.com/eulogik/TinyDoc-VLM.

License: Apache 2.0.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support