guanwy1995
guanwy1995
您好，这个处理过程是我们follow的llama模型的处理，您可以参考 https://github.com/ggml-org/llama.cpp/blob/master/convert_hf_to_gguf.py#L1943 ，这个转换并非MiniCPM模型所独有的特性，应该是为提高矩阵运算效率所做的特殊处理。
We are preparing to open-source the training code for BitCPM. In contrast, we do not recommend using unsloth to directly fine-tune BitCPM, as this may damage the model parameters’ fake-quantized...