From various examples of Tensorflow (translation, ptb) it seems like that you need to explicitly change learning rate when using GradientDescentOptimizer. But is it the case while using some more 'sophisticated' techniques like Adagrad, Adadelta etc. Also when we continue training the model from a saved instance, are the past values used by these optimizers saved in the model file ?
It depends on the Optimizer you are using. Vanilla SGD needs (accepts) individual adaption of the learning rate. Some others do. Adadelta for example does not. (https://arxiv.org/abs/1212.5701)
So this depends not so much on Tensorflow but rather on the mathematical background of the optimizer you are using.
Furthermore: Yes, saving and restarting the training does not reset the learning rates, but continuous at the point saved.
Related
This is a more general version of a question I've already asked: Significant difference between outputs of deep tensorflow keras model in Python and tensorflowjs conversion
As far as I can tell, the layers of a tfjs model when run in the browser (so far only tested in Chrome and Firefox) will have small numerical differences in the output values when compared to the same model run in Python or Node. The cumulative effect of these small differences across all the layers of the model can cause fairly significant differences in the output. See here for an example of this.
This means a model trained in Python or Node will not perform as well in terms of accuracy when run in the browser. And the deeper your model, the worse it will get.
Therefore my question is, what is the best way to train a model to use with tfjs in the browser? Is there a way to ensure the output will be identical? Or do you just have to accept that there will be small numerical differences and, if so, are there any methods that can be used to train a model to be more resilient to this?
This answer is based on my personal observations. As such, it is debatable and not backed by much evidence. Some things that I follow to get accuracy of 16-bit models close to 32 bit models are:
Avoid using activations that have small upper and lower bounds, such as sigmoid or tanh, for hidden layers. These activations cause the weights of the next layer to become very sensitive to small values, and hence, small changes. I prefer using ReLU for such models. Since it is now the standard activation for hidden layers in most models, you should be using it in any case.
Avoid weight decay and L1/L2 regularizations on weights while training (the kernel_regularizer parameter in keras), since these increase sensitivity of weights. Use Dropout instead, I didn't observe a major drop in performance on TFLite when using it instead of numerical regularizers.
I followed all the steps mentioned in the article:
https://stackabuse.com/tensorflow-2-0-solving-classification-and-regression-problems/
Then I compared the results with Linear Regression and found that the error is less (68) than the tensorflow model (84).
from sklearn.linear_model import LinearRegression
logreg_clf = LinearRegression()
logreg_clf.fit(X_train, y_train)
pred = logreg_clf.predict(X_test)
print(np.sqrt(mean_squared_error(y_test, pred)))
Does this mean that if I have large dataset, I will get better results than linear regression?
What is the best situation - when I should be using tensorflow?
Answering your first question, Neural Networks are notoriously known for overfitting on smaller datasets, and here you are comparing the performance of a simple linear regression model with a neural network with two hidden layers on the testing data set, so it's not very surprising to see that the MLP model falling behind (assuming that you are working with relatively a smaller dataset) the linear regression model. Larger datasets will definitely help neural networks in learning more accurate parameters and generalize the phenomena well.
Now coming to your second question, Tensorflow is basically a library for building deep learning models, so whenever you are working on a deep learning problem like image recognition, Natural Language Processing, etc. you need massive computational power and will be processing a ton of data to train your models, and this is where TensorFlow becomes handy, it offers you GPU support which will significantly boost your training process which otherwise becomes practically impossible. Moreover, if you are building a product that has to be deployed in a production environment for it to be consumed, you can make use of TensorFlow Serving which helps you to take your models much closer to the customers.
Is there any equivalent/alternate library to fastai in tensorfow for easier training and debugging deep learning models including analysis on results of trained model in Tensorflow.
Fastai is built on top of pytorch looking for similar one in tensorflow.
The obvious choice would be to use tf.keras.
It is bundled with tensorflow and is becoming its official "high-level" API -- to the point where in TF 2 you would probably need to go out of your way not using it at all.
It is clearly the source of inspiration for fastai to easy the use of pytorch as Keras does for tensorflow, as mentionned by the authors time and again:
Unfortunately, Pytorch was a long way from being a good option for part one of the course, which is designed to be accessible to people with no machine learning background. It did not have anything like the clear simple API of Keras for training models. Every project required dozens of lines of code just to implement the basics of training a neural network. Unlike Keras, where the defaults are thoughtfully chosen to be as useful as possible, Pytorch required everything to be specified in detail. However, we also realised that Keras could be even better. We noticed that we kept on making the same mistakes in Keras, such as failing to shuffle our data when we needed to, or vice versa. Also, many recent best practices were not being incorporated into Keras, particularly in the rapidly developing field of natural language processing. We wondered if we could build something that could be even better than Keras for rapidly training world-class deep learning models.
I'm attempting to train a tensorflow model based on the popular slim implementation of mobilenet_v2 and am observing behaviour I cannot explain related (I think) to batch normalization.
Problem Summary
Model performance in inference mode improves initially but starts producing trivial inferences (all near-zeros) after a long period. Good performance continues when run in training mode, even on the evaluation dataset. Evaluation performance is impacted by batch normalization decay/momentum rate... somehow.
More extensive implementation details below, but I'll probably lose most of you with the wall of text, so here are some pictures to get you interested.
The curves below are from a model which I tweaked the bn_decay parameter of while training.
0-370k: bn_decay=0.997 (default)
370k-670k: bn_decay=0.9
670k+: bn_decay=0.5
Loss for (orange) training (in training mode) and (blue) evaluation (in inference mode). Low is good.
Evaluation metric of model on evaluation dataset in inference mode. High is good.
I have attempted to produce a minimal example which demonstrates the issue - classification on MNIST - but have failed (i.e. classification works well and the problem I experience is not exhibited). My apologies for not being able to reduce things further.
Implementation Details
My problem is 2D pose estimation, targeting Gaussians centered at the joint locations. It is essentially the same as semantic segmentation, except rather than using a softmax_cross_entropy_with_logits(labels, logits) I use tf.losses.l2_loss(sigmoid(logits) - gaussian(label_2d_points)) (I use the term "logits" to describe unactivated output of my learned model, though this probably isn't the best term).
Inference Model
After preprocessing my inputs, my logits function is a scoped call to the base mobilenet_v2 followed by a single unactivated convolutional layer to make the number of filters appropriate.
from slim.nets.mobilenet import mobilenet_v2
def get_logtis(image):
with mobilenet_v2.training_scope(
is_training=is_training, bn_decay=bn_decay):
base, _ = mobilenet_v2.mobilenet(image, base_only=True)
logits = tf.layers.conv2d(base, n_joints, 1, 1)
return logits
Training Op
I have experimented with tf.contrib.slim.learning.create_train_op as well as a custom training op:
def get_train_op(optimizer, loss):
global_step = tf.train.get_or_create_global_step()
opt_op = optimizer.minimize(loss, global_step)
update_ops = set(tf.get_collection(tf.GraphKeys.UPDATE_OPS))
update_ops.add(opt_op)
return tf.group(*update_ops)
I'm using tf.train.AdamOptimizer with learning rate=1e-3.
Training Loop
I'm using the tf.estimator.Estimator API for training/evaluation.
Behaviour
Training initially goes well, with an expected sharp increase in performance. This is consistent with my expectations, as the final layer is rapidly trained to interpret the high-level features output by the pretrained base model.
However, after a long period (60k steps with batch_size 8, ~8 hours on a GTX-1070) my model begins to output near-zero values (~1e-11) when run in inference mode, i.e. is_training=False. The exact same model continues to improve when run in *training mode, i.e.is_training=True`, even on the valuation set. I have visually verified this is.
After some experimentation I changed the bn_decay (batch normalization decay/momentum rate) from the default 0.997 to 0.9 at ~370k steps (also tried 0.99, but that didn't make much of a difference) and observed an immdeiate improvement in accuracy. Visual inspection of the inference in inference mode showed clear peaks in the inferred values of order ~1e-1 in the expected places, consistent with the location of peaks from training mode (though values much lower). This is why the accuracy increases significantly, but the loss - while more volative - does not improve much.
These effects dropped off after more training and reverted to all zero inference.
I further dropped the bn_decay to 0.5 at step ~670k. This resulted in improvements to both loss and accuracy. I'll likely have to wait until tomorrow to see the long-term effect.
Loss and an evaluation metric plots given below. Note the evaluation metric is based on the argmax of the logits and high is good. Loss is based on the actual values, and low is good. Orange uses is_training=True on the training set, while blue uses is_training=False on the evaluation set. The loss of around 8 is consistent with all zero outputs.
Other notes
I have also experimented with turning off dropout (i.e. always running the dropout layers with is_training=False), and observed no difference.
I have experimented with all versions of tensorflow from 1.7 to 1.10. No difference.
I have trained models from the pretrained checkpoint using bn_decay=0.99 from the start. Same behaviour as using default bn_decay.
Other experiments with a batch size of 16 result in qualitatively identical behaviour (though I can't evaluate and train simultaneously due to memory constraints, hence quantitatively analysing on batch size of 8).
I have trained different models using the same loss and using tf.layers API and trained from scratch. They have worked fine.
Training from scratch (rather than using pretrained checkpoints) results in similar behaviour, though takes longer.
Summary/my thoughts:
I am confident this is not an overfitting/dataset problem. The model makes sensible inferences on the evaluation set when run with is_training=True, both in terms of location of peaks and magnitude.
I am confident this is not a problem with not running update ops. I haven't used slim before, but apart from the use of arg_scope it doesn't look too much different to the tf.layers API which I've used extensively. I can also inspect the moving average values and observe that they are changing as training progresses.
Chaning bn_decay values significantly effected the results temporarily. I accept that a value of 0.5 is absurdly low, but I'm running out of ideas.
I have tried swapping out slim.layers.conv2d layers for tf.layers.conv2d with momentum=0.997 (i.e. momentum consistent with default decay value) and behaviour was the same.
Minimal example using pretrained weights and Estimator framework worked for classification of MNIST without modification to bn_decay parameter.
I've looked through issues on both the tensorflow and models github repositories but haven't found much apart from this. I'm currently experimenting with a lower learning rate and a simpler optimizer (MomentumOptimizer), but that's more because I'm running out of ideas rather than because I think that's where the problem lies.
Possible Explanations
The best explanation I have is that my model parameters are rapidly cycling in a manner such that the moving statistics are unable to keep up with the batch statistics. I've never heard of such behaviour, and it doesn't explain why the model reverts to poor behaviour after more time, but it's the best explanation I have.
There may be a bug in the moving average code, but it has worked perfectly for me in every other case, including a simple classification task. I don't want to file an issue until I can produce a simpler example.
Anyway, I'm running out of ideas, the debug cycle is long, and I've already spent too much time on this. Happy to provide more details or run experiments on demand. Also happy to post more code, though I'm worried that'll scare more people off.
Thanks in advance.
Both lowering the learning rate to 1e-4 with Adam and using Momentum optimizer (with learning_rate=1e-3 and momentum=0.9) resolved this issue. I also found this post which suggests the problem spans multiple frameworks and is an undocumented pathology of some networks due to the interaction between optimizer and batch-normalization. I do not believe it is a simple case of the optimizer failing to find a suitable minimum due to the learning rate being too high (otherwise performance in training mode would be poor).
I hope that helps others experiencing the same issue, but I'm a long way from satisfied. I'm definitely happy to hear other explanations.
Tried to training LR model on a large scale dataset via tensorflow with FTRL optimizer for a ctr task. tensorflow/sklearn auc and training/evaluation auc are OK. But performance in product is not good. I've tried to lower down the distributed level, but question can't be totally resolved. Any suggestions?
Found at least two reasons:
First is the underlying implementation is not exactly the same as the original paper. I don't know why they do this, explanation needed.
Second, the gradients used in updating weights are batch gradient, which means update the ps weights once per batch(very trivial in a modern distributed system but not suitable for the scenario in original paper), in a summary it does not utilize the training data record-wise. Personally the second is the key point.