mradermacher/Bpe-vocab-n-OCR-GGUF
Same weights, converted (fewer layers) vs prithivMLmods/Bpe-vocab-n-OCR. mradermacher/Bpe-vocab-n-OCR-GGUF matches part of prithivMLmods/Bpe-vocab-n-OCR (338 of its 729 tensors; the rest of prithivMLmods/Bpe-vocab-n-OCR has no counterpart). On the shared part: Same weights stored in another dtype (F16/BF16 → F16/F32).
Tensors changed: Embeddings 0/1, Norms 0/57, Attention 0/196, MLP 0/84