nvidia/DeepSeek-V3.2-NVFP4
Different architecture vs deepseek-ai/DeepSeek-V3.2. Different architecture: only 1705 of 92425/180857 tensors match by name and shape.
Different architecture vs deepseek-ai/DeepSeek-V3.2. Different architecture: only 1705 of 92425/180857 tensors match by name and shape.