mradermacher/Dumpling-Qwen2.5-VL-3B-GGUF

Same weights, converted (fewer layers) vs nbeerbower/Dumpling-Qwen2.5-VL-3B. mradermacher/Dumpling-Qwen2.5-VL-3B-GGUF matches part of nbeerbower/Dumpling-Qwen2.5-VL-3B (434 of its 824 tensors; the rest of nbeerbower/Dumpling-Qwen2.5-VL-3B has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16/F32).

Tensors changed: Embeddings 0/1, Norms 0/73, MLP 0/108, Attention 0/252

On the Hub · Downloads on Model Pulse