cstr/got-ocr2-crispembed-GGUF

GOT-OCR2 — CrispEmbed GGUF

GGUF conversion of stepfun-ai/GOT-OCR2_0 for use with CrispEmbed.

Architecture

• Vision: SAM ViT-B (12 layers, 768d, 12 heads, 16×16 patches, 1024×1024 input)
  • Windowed attention (ws=14) with global attention at layers [2, 5, 8, 11]
  • Decomposed relative position encoding
  • Neck: Conv(768→256) → LN2d → Conv(256→256) → LN2d
  • Downsample: Conv(256→512→1024, stride 2) → 256 vision tokens
  • Projector: Linear(1024, 1024)
• LLM: Qwen2-0.5B (24 layers, 1024d, MHA 16/16, SiLU SwiGLU, RoPE θ=1M)
• Tokenizer: tiktoken (151860 vocab)
• Total: ~0.7B parameters

Files

File	Quant	Size	Notes
got-ocr2-f16.gguf	F16	1.34 GB	Full precision
got-ocr2-q8_0.gguf	Q8_0	569 MB	Best quantized quality
got-ocr2-q4_k.gguf	Q4_K	422 MB	Smallest, good quality

Parity

All checkpoints verified at cos ≥ 0.999 against Python reference (F32):

• Vision layers (windowed + global attention)
• Neck, downsample, projector
• LLM decoder layers

Usage

crispembed --ocr got-ocr2 image.png

License

Apache-2.0 (same as upstream model)