def _decay_weights_op(
self,
var: tf.Variable,
apply_state: Optional[dict[tuple[str, tf.DType], tf.Tensor]] = None,
):
if self._use_weight_decay(var.name):
var_device, var_dtype = var.device, var.dtype.base_dtype
coefficients = (apply_state or {}).get(
(var_device, var_dtype)) or self._fallback_apply_state(
var_device, var_dtype)
m = self.learning_rate_multiplier.get(var.name, 1)
return var.assign_sub(
coefficients["lr_t"] / m * coefficients["wd_t"] * var,
self._use_locking)
return tf.no_op()
# initialize var with shape print(tf.Variable(tf.constant(4.2, shape=(3, 3)))) # Change vars a = Variable(0.) b = a + 1 print(a) print(b) print(type(b)) a.assign_add(1) print(a) a.assign_sub(4) print(a) # Tensors # Constant tensor c = tf.constant([[1, 1, 1], [2, 3, 3], [4, 4, 4]]) print(c.dtype, c.shape) print(c.numpy()) d = tf.constant([[1, 1, 1], [2, 3, 3], [4, 4, 4]], dtype=tf.float32) print(d) # Coefficients - split list into different shaes
