Open former7 opened 4 years ago
bert code i can see has 3 inputs. (token,segment,mask) but this code only 2 inputs (token,segment) why there is no mask input? masked_layer = Masked(name='MLM')([mlm_pred_layer, inputs[-1]]) this layer use inputs (segment) other than (mask)
masked_layer = Masked(name='MLM')([mlm_pred_layer, inputs[-1]])
bert code i can see has 3 inputs. (token,segment,mask) but this code only 2 inputs (token,segment) why there is no mask input?
masked_layer = Masked(name='MLM')([mlm_pred_layer, inputs[-1]])this layer use inputs (segment) other than (mask)