ggml-org/LightOnOCR-1B-1025-GGUF

Quantization (fewer layers) vs lightonai/LightOnOCR-1B-1025. ggml-org/LightOnOCR-1B-1025-GGUF matches part of lightonai/LightOnOCR-1B-1025 (311 of its 533 tensors; the rest of lightonai/LightOnOCR-1B-1025 has no counterpart). On the shared part: ggml-org/LightOnOCR-1B-1025-GGUF is a GGUF quantization (Q8_0) of the same weights: 113 tensors match exactly and 198 within quantization error.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/113, MLP 0/84, Attention 0/112

On the Hub · Downloads on Model Pulse