Zen
Results
5
issues of
Zen
# 平台(如果交叉编译请再附上交叉编译目标平台): x86_64, MacOS # MNN Version : 2.0.0 使用SIMDOCPruner半结构化剪枝对模型进行稀疏化，转换后的模型相较原来的稠密模型在macos上没有加速，稀疏因子使用了0.6、0.7、0.9，速度都差不多
Could you provide more details about the fine-tune process, such as the data volume, hyper-parameter configuration, training duration, etc.?
When cal the L_guide loss, values of the guided landmarks coordinates are normalized to between 0 and 1, or between -1 and 1?
qa