a .Sic§1ã@szdZddlmZddlmZddlmZddlmZddlmZddlm Z ddl m Z e d gd Gd d „d ej ƒƒZ d S)aQOne-line documentation for rmsprop module. rmsprop algorithm [tieleman2012rmsprop] A detailed description of rmsprop. - maintain a moving (discounted) average of the square of gradients - divide gradient by the root of this average mean_square = decay * mean_square{t-1} + (1-decay) * gradient ** 2 mom = momentum * mom{t-1} + learning_rate * g_t / sqrt(mean_square + epsilon) delta = - mom This implementation of RMSProp uses plain momentum, not Nesterov momentum. The centered version additionally maintains a moving (discounted) average of the gradients, and uses that average to estimate the variance: mean_grad = decay * mean_grad{t-1} + (1-decay) * gradient mean_square = decay * mean_square{t-1} + (1-decay) * gradient ** 2 mom = momentum * mom{t-1} + learning_rate * g_t / sqrt(mean_square - mean_grad**2 + epsilon) delta = - mom é)Úops)Ú array_ops)Úinit_ops)Úmath_ops)Ú optimizer)Ú training_ops)Ú tf_exportztrain.RMSPropOptimizer)Úv1csReZdZdZd‡fdd„ Zd d „Zd d „Zd d„Zdd„Zdd„Z dd„Z ‡Z S)ÚRMSPropOptimizera Optimizer that implements the RMSProp algorithm (Tielemans et al. 2012). References: Coursera slide 29: Hinton, 2012 ([pdf](http://www.cs.toronto.edu/~tijmen/csc321/slides/lecture_slides_lec6.pdf)) @compatibility(TF2) tf.compat.v1.train.RMSPropOptimizer is compatible with eager mode and `tf.function`. When eager execution is enabled, `learning_rate`, `decay`, `momentum`, and `epsilon` can each be a callable that takes no arguments and returns the actual value to use. This can be useful for changing these values across different invocations of optimizer functions. To switch to native TF2 style, use [`tf.keras.optimizers.RMSprop`] (https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/RMSprop) instead. Please notice that due to the implementation differences, `tf.keras.optimizers.RMSprop` and `tf.compat.v1.train.RMSPropOptimizer` may have slight differences in floating point numerics even though the formula used for the variable updates still matches. #### Structural mapping to native TF2 Before: ```python optimizer = tf.compat.v1.train.RMSPropOptimizer( learning_rate=learning_rate, decay=decay, momentum=momentum, epsilon=epsilon) ``` After: ```python optimizer = tf.keras.optimizers.RMSprop( learning_rate=learning_rate, rho=decay, momentum=momentum, epsilon=epsilon) ``` #### How to map arguments | TF1 Arg Name | TF2 Arg Name | Note | | ------------------ | ------------- | ------------------------------- | | `learning_rate` | `learning_rate`| Be careful of setting | : : : learning_rate tensor value computed from the global step. : : : : In TF1 this was usually meant to imply a dynamic learning rate and : : : : would recompute in each step. In TF2 (eager + function) it will : : : : treat it as a scalar value that only gets computed once instead of : : : : a symbolic placeholder to be computed each time. : | `decay` | `rho` | - | | `momentum` | `momentum` | - | | `epsilon` | `epsilon` | Default value is 1e-10 in TF1, | : : : but 1e-07 in TF2. : | `use_locking` | - | Not applicable in TF2. | #### Before & after usage example Before: ```python x = tf.Variable([1,2,3], dtype=tf.float32) grad = tf.constant([0.1, 0.2, 0.3]) optimizer = tf.compat.v1.train.RMSPropOptimizer(learning_rate=0.001) optimizer.apply_gradients(zip([grad], [x])) ``` After: ```python x = tf.Variable([1,2,3], dtype=tf.float32) grad = tf.constant([0.1, 0.2, 0.3]) optimizer = tf.keras.optimizers.RMSprop(learning_rate=0.001) optimizer.apply_gradients(zip([grad], [x])) ``` @end_compatibility çÍÌÌÌÌÌì?ç绽×Ùß|Û=FÚRMSPropcsLtt|ƒ ||¡||_||_||_||_||_d|_d|_ d|_ d|_ dS)a:Construct a new RMSProp optimizer. Note that in the dense implementation of this algorithm, variables and their corresponding accumulators (momentum, gradient moving average, square gradient moving average) will be updated even if the gradient is zero (i.e. accumulators will decay, momentum will be applied). The sparse implementation (used when the gradient is an `IndexedSlices` object, typically because of `tf.gather` or an embedding lookup in the forward pass) will not update variable slices or their accumulators unless those slices were used in the forward pass (nor is there an "eventual" correction to account for these omitted updates). This leads to more efficient updates for large embedding lookup tables (where most of the slices are not accessed in a particular graph execution), but differs from the published algorithm. Args: learning_rate: A Tensor or a floating point value. The learning rate. decay: Discounting factor for the history/coming gradient momentum: A scalar tensor. epsilon: Small value to avoid zero denominator. use_locking: If True use locks for update operation. centered: If True, gradients are normalized by the estimated variance of the gradient; if False, by the uncentered second moment. Setting this to True may help with training, but is slightly more expensive in terms of computation and memory. Defaults to False. name: Optional name prefix for the operations created when applying gradients. Defaults to "RMSProp". N) Úsuperr Ú__init__Ú_learning_rateÚ_decayÚ _momentumÚ_epsilonÚ _centeredÚ_learning_rate_tensorÚ _decay_tensorÚ_momentum_tensorÚ_epsilon_tensor)ÚselfÚ learning_rateÚdecayÚmomentumÚepsilonÚ use_lockingÚcenteredÚname©Ú __class__©ú^/var/www/html/django/DPS/env/lib/python3.9/site-packages/tensorflow/python/training/rmsprop.pyrˆs$zRMSPropOptimizer.__init__c Csz|D]p}| ¡ ¡r&tj|jjd}n t |¡}| ||| ¡|jjd|j ¡|j rd|  |d|j ¡|  |d|j ¡qdS)N)ÚdtypeÚrmsÚmgr) Ú get_shapeÚis_fully_definedrÚones_initializerr&Ú base_dtyperÚ ones_likeÚ"_get_or_make_slot_with_initializerÚ_namerÚ _zeros_slot)rÚvar_listÚvZinit_rmsr$r$r%Ú _create_slots¹s  þzRMSPropOptimizer._create_slotscCst| |j¡}| |j¡}| |j¡}| |j¡}tj|dd|_tj|dd|_tj|dd|_ tj|dd|_ dS)Nr)r!rrr) Ú_call_if_callablerrrrrÚconvert_to_tensorrrrr)rÚlrrrrr$r$r%Ú_prepareÆs    zRMSPropOptimizer._preparec Csà| |d¡}| |d¡}|jr„| |d¡}tj||||t |j|jj¡t |j |jj¡t |j |jj¡t |j |jj¡||j d j Stj|||t |j|jj¡t |j |jj¡t |j |jj¡t |j |jj¡||j d j SdS©Nr'rr()r)Úget_slotrrÚapply_centered_rms_proprÚcastrr&r,rrrÚ _use_lockingÚopÚapply_rms_prop©rÚgradÚvarr'Úmomr(r$r$r%Ú _apply_denseÑs6   ö ÷zRMSPropOptimizer._apply_densec Csê| |d¡}| |d¡}|jrŠ| |d¡}tj|j|j|j|jt |j|jj ¡t |j |jj ¡t |j |jj ¡t |j |jj ¡||j d Stj|j|j|jt |j|jj ¡t |j |jj ¡t |j |jj ¡t |j |jj ¡||j d SdSr8)r9rrÚ resource_apply_centered_rms_propÚhandlerr;rr&r,rrrr<Úresource_apply_rms_propr?r$r$r%Ú_resource_apply_denseís6   ö ÷z&RMSPropOptimizer._resource_apply_densec Csè| |d¡}| |d¡}|jrˆ| |d¡}tj||||t |j|jj¡t |j |jj¡t |j |jj¡t |j |jj¡|j |j |jd Stj|||t |j|jj¡t |j |jj¡t |j |jj¡t |j |jj¡|j |j |jd SdSr8)r9rrÚsparse_apply_centered_rms_proprr;rr&r,rrrÚvaluesÚindicesr<Úsparse_apply_rms_propr?r$r$r%Ú _apply_sparse s:   õ özRMSPropOptimizer._apply_sparsec CsÞ| |d¡}| |d¡}|jr„| |d¡}tj|j|j|j|jt |j|j¡t |j |j¡t |j |j¡t |j |j¡|||j d Stj |j|j|jt |j|j¡t |j |j¡t |j |j¡t |j |j¡|||j d SdSr8)r9rrÚ'resource_sparse_apply_centered_rms_proprErr;rr&rrrr<Úresource_sparse_apply_rms_prop)rr@rArJr'rBr(r$r$r%Ú_resource_apply_sparse's:   õ öz'RMSPropOptimizer._resource_apply_sparse)r r r FFr) Ú__name__Ú __module__Ú __qualname__Ú__doc__rr3r7rCrGrLrOÚ __classcell__r$r$r"r%r 1sWù1  r N)rSÚtensorflow.python.frameworkrÚtensorflow.python.opsrrrÚtensorflow.python.trainingrrÚ tensorflow.python.util.tf_exportrÚ Optimizerr r$r$r$r%Ús