Anbeeld/Qwen3-4B-DSpark-GGUF
Same weights, converted vs deepseek-ai/dspark_qwen3_4b_block7. Same weights stored in another dtype (BF16 → BF16/F32), plus 3 new tensors.
Tensors changed: Embeddings 0/1, other 0/3, Norms 0/21, MLP 0/15, Attention 0/20
Same weights, converted vs deepseek-ai/dspark_qwen3_4b_block7. Same weights stored in another dtype (BF16 → BF16/F32), plus 3 new tensors.
Tensors changed: Embeddings 0/1, other 0/3, Norms 0/21, MLP 0/15, Attention 0/20