lambda left: np.tanh(left), "tanh2": lambda left: (np.exp2(2. * left) - 1.) / (np.exp2(2. * left) + 1.), "transpose": lambda left: np.transpose(left), "rint": lambda left: np.rint(left), # binary ops "add": lambda left, right: math_cpu.add(left, right), "sub": lambda left, right: math_cpu.sub(left, right), "mul": lambda left, right: math_cpu.mul(left, right), "div": lambda left, right: math_cpu.div(left, right), "eq": lambda left, right: left == right, "ne": lambda left, right: left != right, "lt": lambda left, right: left < right, "le": lambda left, right: left <= right, "gt": lambda left, right: left > right, "ge": lambda left, right: left >= right, "pow": lambda left, right: np.power(left, right), "minimum":
"exp": lambda left: math_cpu.exp(left), "log": lambda left: math_cpu.log(left), "safelog": lambda left: math_cpu.safelog(left), "exp2": lambda left: np.exp2(left), "log2": lambda left: np.log2(left), "sig": lambda left: 1. / (1. + np.exp(-left)), "sig2": lambda left: 1. / (1. + np.exp2(-left)), "tanh": lambda left: np.tanh(left), "tanh2": lambda left: (np.exp2(2. * left) - 1.) / (np.exp2(2. * left) + 1.), "transpose": lambda left: np.transpose(left), "rint": lambda left: np.rint(left), # binary ops "add": lambda left, right: math_cpu.add(left, right), "sub": lambda left, right: math_cpu.sub(left, right), "mul": lambda left, right: math_cpu.mul(left, right), "div": lambda left, right: math_cpu.div(left, right), "eq": lambda left, right: left == right, "ne": lambda left, right: left != right, "lt": lambda left, right: left < right, "le": lambda left, right: left <= right, "gt": lambda left, right: left > right, "ge": lambda left, right: left >= right, "pow": lambda left, right: np.power(left, right), "minimum": lambda left, right: np.minimum(left, right), "maximum": lambda left, right: np.maximum(left, right), "dot": lambda left, right: np.dot(left, right), # reduction ops "sum": lambda op_dict, left: math_cpu.sum(left, axis=op_dict['axis'], keepdims=True), "max": lambda op_dict, left: np.max(left, axis=op_dict['axis'], keepdims=True), "min": lambda op_dict, left: np.min(left, axis=op_dict['axis'], keepdims=True), "argmax": lambda op_dict, left: CustomNumpy.argmax(left, axis=op_dict['axis'], keepdims=True),
