vidore/colqwen2-v1.0-hf

Same weights, converted vs vidore/colqwen2-base. Same weights stored in another dtype (F32 → BF16), plus 340 new tensors.

Tensors changed: Attention 0/128, MLP 0/132, Norms 0/130, Embeddings 0/1

2 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse