I implement training and evaluating for binary classification with image data through transfer learning from keras API. I'd like to compare performance each models(ResNet, Inception, Xception, VGG, Efficient Net). The datasets are composed by train(approx.2000ea), valid(approx.250ea), test(approx.250ea).
But I faced unfamiliar situation for me so I'm asking couple of questions here.
As shown below, Valid Accuracy or Loss has a very high up and down deviation.
I wonder which one is the problem and what needs to be changed.
epoch_acc_loss
loss_epoch
acc_epoch
If I want to express validation accuracy with number, what should I say in the above case?
Average or maximum or minimum?
It is being performed using Keras (tensorflow), and there are many examples in the API for
train, valid but the code for Test(evaluation?) is hard to find. When figuring performance,
normally implement until valid? or Do I need to show evaluation result?
Now I use Keras API for transfer learning and set this.
include_top=False
conv_base.trainable=False
Summary
I wonder if there is an effect of transfer learning without includint from top, or if it's not,
is there a way to freeze or learn from a specific layer of conv_base.
I'm a beginner and have not many experience so it could be ridiculous questions but please give kind advice.
Thanks a lot in advance.
It's hard to figure out the problem without any given code/model structure. From your loss graph I can see that your model is facing underfitting (or it has a lots of dropout). Common mistakes, that make models underfit are: very high lr and primitive structure (so model can't figure out the dependencies in your data). And you should never forget about the principle "garbage in - garbage out", so double-check tour data for any structure roughness.
Well, validation accuracy in you model training logs is mean accuracy for validation set. Validation technique is based on statistics - you take random N% out of your set for validation, so average is always better if we're talking about multiple experimets (or cross validation).
I'm not sure if I've understood your question correct here, but if you want to evaluate your model with the metric, that you've specified for it after the training process (fit() function call) you should use model.evaluate(val_x, val_y). Or you may use model.predict(val_x) and compare its results to val_y via your metric function.
If you are using default weights for keras pretrained models (imagenet weights) and you want to use your own fully-connected part with it, you may use ONLY pretrained feature extractor (conv blocks). So you specify include_top=False. Of course there will be some positive effect (I'd say it will be significant in comparison with randomly initialized weights) because conv blocks have params that were trained to extract correct features from image. Also would recommend here to use so called "fine-tuning" technique - freeze all layers in pretrained part except a few in its end (may be few layers or even 2-3 conv blocks). Here's the example of fine-tuning of EfficientNetB0:
effnet = EfficientNetB0(weights="imagenet", include_top=False, input_shape=(540, 960, 3))
effnet.trainable = True
for layer in effnet.layers:
if 'block7a' not in layer.name and 'top' not in layer.name:
layer.trainable = False
Here I freeze all pretrained weights except last conv block ones. I've looked into the model with effnet.summary() and selected names of blocks that I want to unfreeze.
i was trying to use average ensembling on a group of models i trained earlier (i'm creating a new model in the ensemble for each pre-trained model i'm using and then loading the trained weights onto it, it's inefficient this way i know but i'm just learning about it so it doesn't really matter). and I mistakenly changed some of the network's parameters when loading the models in the ensemble code like using Relu instead of leakyRelu which i used in training the models and a different value for an l2 regularizer in the dense layer in one of the models. this however gave me a better testing accuracy for the ensemble. can you please explain to me if/how this is incorrect, and if it's normal can i use this method to further enhance the accuracy of the ensemble.
I believe it is NOT correct to chnage model's parameters after training it. parameters here I mean the trainable-parameters like the weights in Dense node but not hyper-parameters like learning rate.
What is training?
Training essentially is a loop that keeps changing, or update, the parameters. It updates the parameter in such a way that it believes it can reduce the loss. It is also like moving your point in a hyper-spaces that the loss function gives a small loss on that point.
Smaller loss means higher accruacy in general.
Changing Weights
So now, changing your parameters values, by mistake or by purpose, is like moving that point to somewhere BUT you have no logical reason behind that such move will give you a smaller loss. You are just randomly wandering around that hyper-space and in your case you are just lucky that you land to some point that so happened give you a smaller loss or a better testing accuracy. It is just purely luck.
Changing activation function
Also, altering the activation function from leakyRelu to relu is similar you randomly alter the shape of your hype-space. Even though you are at the some point the landscape changes, you are still have no logical reason to believe by such change of landscape you can have a smaller loss staying at the same point
When you change the model manually, you need to retrain.
Though you changed the network's parameters when loading the models. It is not incorrect to alter the hyper-parameters of your ensemble's underlying models. In some cases, the models that are used in an ensemble method require unique tunings which can, as you mentioned, give "you a better testing accuracy for the ensemble model."
To answer your second question, you can use this method to further enhance the accuracy of the ensemble, you can also use Bayesian optimization, GridSearch, and RandomSearch if you prefer more automated means of tuning your hyperparameters.
I am currently using the TensorFlow Object Detection API and am attempting to fine tune a pre-trained Faster-RCNN from the model zoo. Currently, if I choose a different number of classes to the number used in the original network, it will simply not initialise the weights and biases from the SecondStageBoxPredictor/ClassPredictor as this now has different dimensions from the original ClassPredictor. However, as all of the classes I would like to train the network on are classes the original network has been trained to identify, I would like to retain the weights and biases associated with the classes I want to use in SecondStageBoxPredictor/ClassPredictor and prune all the others, rather than simply initialising these values from scratch (similar to the behaviour of this function).
Is this possible, and if so, how would I go about modifying the structure of this layer in the Estimator?
n.b. This question asks a similar thing, and their response is to ignore irrelevant classes from the network output - in this situation, however, I am attempting to fine tune the network and I assume the presence of these redundant classes would complicate the training / evaluation process?
If all the classes you would like to train the network on are the ones the network has been trained to identify, you could simply use the network to detect, isn't it?
However, if you have extra classes and you would like to do transfer-learning, you can have as many variables restored from checkpoint as possible by setting:
fine_tune_checkpoint_type: 'detection'
load_all_detection_checkpoint_vars: True
in field train_config from the pipeline config file.
Finally, by looking at the computation graph, it can be seen that the shape of SecondStageBoxPredictor/ClassPredictor/weights is dependent on the number of output classes.
Note that in tensorflow you can only restore in variables level, if two variables have different shapes, one can not use one to initialize the other. So in your case the idea of preserving some values of the weights variable is not feasible.
I'm in the process of adapting my model to TensorFlow's estimator API.
I recently asked a question regarding early stopping based on validation data where in addition to early stopping, the best model at this point should be exported.
It seems that my understanding of what a model export is and what a checkpoint is is not complete.
Checkpoints are made automatically. From my understanding, the checkpoints are sufficient for the estimator to start "warm" - either using so per-trained weights or weights prior to an error (e.g. if you experienced a power outage).
What is nice about checkpoints is that I do not have to write any code besides what is necessary for a custom estimator (namely, input_fn and model_fn).
While, given an initialized estimator, one can just call its train method to train the model, in practice this method is rather lackluster. Often one would like to do several things:
compare the network periodically to a validation dataset to ensure you are not over-fitting
stop the training early if over-fitting occurs
save the best model whenever the network finishes (either by hitting the specified number of training steps or by the early stopping criteria).
To someone new to the "high level" estimator API, a lot of low level expertise seems to be required (e.g. for the input_fn) as how one could get the estimator to do this is not straight forward.
By some light code reworking #1 can be achieved by using tf.estimator.TrainSpec and tf.estimator.EvalSpec with tf.estimator.train_and_evaluate.
In the previous question user #GPhilo clarifies how #2 can be achieved by using a semi-unintuitive function from the tf.contrib:
tf.contrib.estimator.stop_if_no_decrease_hook(my_estimator,'my_metric_to_monitor', 10000)
(unintuitive as "the early stopping is not triggered according to the number of non-improving evaluations, but to the number of non-improving evals in a certain step range").
#GPhilo - noting that it is unrelated to #2 - also answered how to do #3 (as requested in the original post). Yet, I do not understand what an input_serving_fn is, why it is needed, or how to make it.
This is further confusing to me as no such function is needed to make checkpoints, or for the estimator to start "warm" from the checkpoint.
So my questions are:
what is the difference between a checkpoint and an exported best model?
what exactly is a serving input receiver function and how to write one? (I have spent a bit of time reading over the tensorflow docs and do not find it sufficient to understand how I should write one, and why I even have to).
how can I train my estimator, save the best model, and then later load it.
To aid in answering my question I am providing this Colab document.
This self contained notebook produces some dummy data, saves it in TF Records, has a very simple custom estimator via model_fn and trains this model with an input_fn that uses the TF Record files. Thus it should be sufficient for someone to explain to me what placeholders I need to make for the input serving receiver function and and how I can accomplish #3.
Update
#GPhilo foremost I can not understate my appreciation for you thoughtful consideration and care in aiding me (and hopefully others) understand this matter.
My “goal” (motivating me to ask this question) is to try and build a reusable framework for training networks so I can just pass a different build_fn and go (plus have the quality of life features of exported model, early stopping, etc).
An updated (based off your answers) Colab can be found here.
After several readings of your answer, I have found now some more confusion:
1.
the way you provide input to the inference model is different than the one you use for the training
Why? To my understanding the data input pipeline is not:
load raw —> process —> feed to model
But rather:
Load raw —> pre process —> store (perhaps as tf records)
# data processing has nothing to do with feeding data to the model?
Load processed —> feed to model
In other words, it is my understanding (perhaps wrongly) that the point of a tf Example / SequenceExample is to store a complete singular datum entity ready to go - no other processing needed other than reading from the TFRecord file.
Thus there can be a difference between the training / evaluation input_fn and the inference one (e.g. reading from file vs eager / interactive evaluation of in memory), but the data format is the same (except for inference you might want to feed only 1 example rather than a batch…)
I agree that the “input pipeline is not part of the model itself”. However, in my mind, and I am apparently wrong in thinking so, with the estimator I should be able to feed it a batch for training and a single example (or batch) for inference.
An aside: “When evaluating, you don't need the gradients and you need a different input function. “, the only difference (at least in my case) is the files from which you reading?
I am familiar with that TF Guide, but I have not found it useful because it is unclear to me what placeholders I need to add and what additional ops needed to be added to convert the data.
What if I train my model with records and want to inference with just the dense tensors?
Tangentially, I find the example in the linked guide subpar, given the tf record interface requires the user to define multiple times how to write to / extract features from a tf record file in different contexts. Further, given that the TF team has explicitly stated they have little interest in documenting tf records, any documentation built on top of it, to me, is therefore equally unenlightening.
Regarding tf.estimator.export.build_raw_serving_input_receiver_fn.
What is the placeholder called? Input? Could you perhaps show the analog of tf.estimator.export.build_raw_serving_input_receiver_fn by writing the equivalent serving_input_receiver_fn
Regarding your example serving_input_receiver_fn with the input images. How do you know to call features ‘images’ and the receiver tensor ‘input_data’ ? Is that (the latter) standard?
How to name an export with signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY.
What is the difference between a checkpoint and an exported best model?
A checkpoint is, at its minimum, a file containing the values of all the variables of a specific graph taken at a specific time point.
By specific graph I mean that when loading back your checkpoint, what TensorFlow does is loop through all the variables defined in your graph (the one in the session you're running) and search for a variable in the checkpoint file that has the same name as the one in the graph. For resuming training, this is ideal because your graph will always look the same between restarts.
An exported model serves a different purpose. The idea of an exported model is that, once you're done training, you want to get something you can use for inference that doesn't contain all the (heavy) parts that are specific to training (some examples: gradient computation, global step variable, input pipeline, ...).
Moreover, and his is the key point, typically the way you provide input to the inference model is different than the one you use for the training. For training, you have an input pipeline that loads, preprocess and feeds data to your network. This input pipeline is not part of the model itself and may have to be altered for inference. This is a key point when operating with Estimators.
Why do I need a serving input receiver function?
To answer this I'll take first a step back. Why do we need input functions at all ad what are they? TF's Estimators, while perhaps not as intuitive as other ways to model networks, have a great advantage: they clearly separate between model logic and input processing logic by means of input functions and model functions.
A model lives in 3 different phases: Training, Evaluation and Inference. For the most common use-cases (or at least, all I can think of at the moment), the graph running in TF will be different in all these phases. The graph is the combination of input preprocessing, model and all the machinery necessary to run the model in the current phase.
A few examples to hopefully clarify further: When training, you need gradients to update the weights, an optimizer that runs the training step, metrics of all kinds to monitor how things are going, an input pipeline that grabs data from the training set, etc. When evaluating, you don't need the gradients and you need a different input function. When you are inferencing, all you need is the forward part of the model and again the input function will be different (no tf.data.* stuff but typically just a placeholder).
Each of these phases in Estimators has its own input function. You're familiar with the training and evaluation ones, the inference one is simply your serving input receiver function. In TF lingo, "serving" is the process of packing a trained model and using it for inference (there's a whole TensorFlow serving system for large-scale operation but that's beyond this question and you most likely won't need it anyhow).
Time to quote a TF guide on the topic:
During training, an input_fn() ingests data and prepares it for use by
the model. At serving time, similarly, a serving_input_receiver_fn()
accepts inference requests and prepares them for the model. This
function has the following purposes:
To add placeholders to the graph that the serving system will feed
with inference requests.
To add any additional ops needed to convert
data from the input format into the feature Tensors expected by the
model.
Now, the serving input function specification depends on how you plan of sending input to your graph.
If you're going to pack the data in a (serialized) tf.Example (which is similar to one of the records in your TFRecord files), your serving input function will have a string placeholder (that's for the serialized bytes for the example) and will need a specification of how to interpret the example in order to extract its data. If this is the way you want to go I invite you to have a look at the example in the linked guide above, it essentially shows how you setup the specification of how to interpret the example and parse it to obtain the input data.
If, instead, you're planning on directly feeding input to the first layer of your network you still need to define a serving input function, but this time it will only contain a placeholder that will be plugged directly into the network. TF offers a function that does just that: tf.estimator.export.build_raw_serving_input_receiver_fn.
So, do you actually need to write your own input function? IF al you need is a placeholder, no. Just use build_raw_serving_input_receiver_fn with the appropriate parameters. IF you need fancier preprocessing, then yes, you might need to write your own. In that case, it would look something like this:
def serving_input_receiver_fn():
"""For the sake of the example, let's assume your input to the network will be a 28x28 grayscale image that you'll then preprocess as needed"""
input_images = tf.placeholder(dtype=tf.uint8,
shape=[None, 28, 28, 1],
name='input_images')
# here you do all the operations you need on the images before they can be fed to the net (e.g., normalizing, reshaping, etc). Let's assume "images" is the resulting tensor.
features = {'input_data' : images} # this is the dict that is then passed as "features" parameter to your model_fn
receiver_tensors = {'input_data': input_images} # As far as I understand this is needed to map the input to a name you can retrieve later
return tf.estimator.export.ServingInputReceiver(features, receiver_tensors)
How can I train my estimator, save the best model, and then later load it?
Your model_fn takes the mode parameter in order for you to build conditionally the model. In your colab, you always have a optimizer, for example. This is wrong ,as it should only be there for mode == tf.estimator.ModeKeys.TRAIN.
Secondly, your build_fn has an "outputs" parameter that is meaningless. This function should represent your inference graph, take as input only the tensors you'll fed to it in the inference and return the logits/predictions.
I'll thus assume the outputs parameters is not there as the build_fn signature should be def build_fn(inputs, params).
Moreover, you define your model_fn to take features as a tensor. While this can be done, it both limits you to having exactly one input and complicates things for the serving_fn (you can't use the canned build_raw_... but need to write your own and return a TensorServingInputReceiver instead). I'll choose the more generic solution and assume your model_fn is as follows (I omit the variable scope for brevity, add it as necessary):
def model_fn(features, labels, mode, params):
my_input = features["input_data"]
my_input.set_shape(I_SHAPE(params['batch_size']))
# output of the network
onet = build_fn(features, params)
predicted_labels = tf.nn.sigmoid(onet)
predictions = {'labels': predicted_labels, 'logits': onet}
export_outputs = { # see EstimatorSpec's docs to understand what this is and why it's necessary.
'labels': tf.estimator.export.PredictOutput(predicted_labels),
'logits': tf.estimator.export.PredictOutput(onet)
}
# NOTE: export_outputs can also be used to save models as "SavedModel"s during evaluation.
# HERE is where the common part of the graph between training, inference and evaluation stops.
if mode == tf.estimator.ModeKeys.PREDICT:
# return early and avoid adding the rest of the graph that has nothing to do with inference.
return tf.estimator.EstimatorSpec(mode=mode,
predictions=predictions,
export_outputs=export_outputs)
labels.set_shape(O_SHAPE(params['batch_size']))
# calculate loss
loss = loss_fn(onet, labels)
# add optimizer only if we're training
if mode == tf.estimator.ModeKeys.TRAIN:
optimizer = tf.train.AdagradOptimizer(learning_rate=params['learning_rate'])
# some metrics used both in training and eval
mae = tf.metrics.mean_absolute_error(labels=labels, predictions=predicted_labels, name='mea_op')
mse = tf.metrics.mean_squared_error(labels=labels, predictions=predicted_labels, name='mse_op')
metrics = {'mae': mae, 'mse': mse}
tf.summary.scalar('mae', mae[1])
tf.summary.scalar('mse', mse[1])
if mode == tf.estimator.ModeKeys.EVAL:
return tf.estimator.EstimatorSpec(mode, loss=loss, eval_metric_ops=metrics, predictions=predictions, export_outputs=export_outputs)
if mode == tf.estimator.ModeKeys.TRAIN:
train_op = optimizer.minimize(loss, global_step=tf.train.get_global_step())
return tf.estimator.EstimatorSpec(mode, loss=loss, train_op=train_op, eval_metric_ops=metrics, predictions=predictions, export_outputs=export_outputs)
Now, to set up the exporting part, after your call to train_and_evaluate finished:
1) Define your serving input function:
serving_fn = tf.estimator.export.build_raw_serving_input_receiver_fn(
{'input_data':tf.placeholder(tf.float32, [None,#YOUR_INPUT_SHAPE_HERE (without batch size)#])})
2) Export the model to some folder
est.export_savedmodel('my_directory_for_saved_models', serving_fn)
This will save the current state of the estimator to wherever you specified. If you want a specifc checkpoint, load it before calling export_savedmodel.
This will save in "my_directory_for_saved_models" a prediction graph with the trained parameters that the estimator had when you called the export function.
Finally, you might want t freeze the graph (look up freeze_graph.py) and optimize it for inference (look up optimize_for_inference.py and/or transform_graph) obtaining a frozen *.pb file you can then load and use for inference as you wish.
Edit: Adding answers to the new questions in the update
Sidenote:
My “goal” (motivating me to ask this question) is to try and build a
reusable framework for training networks so I can just pass a
different build_fn and go (plus have the quality of life features of
exported model, early stopping, etc).
By all means, if you manage, please post it on GitHub somewhere and link it to me. I've been trying to get just the same thing up and running for a while now and the results are not quite as good as I'd like them to be.
Question 1:
In other words, it is my understanding (perhaps wrongly) that the
point of a tf Example / SequenceExample is to store a complete
singular datum entity ready to go - no other processing needed other
than reading from the TFRecord file.
Actually, this is typically not the case (although, your way is in theory perfectly fine too).
You can see TFRecords as a (awfully documented) way to store a dataset in a compact way. For image datasets for example, a record typically contains the compressed image data (as in, the bytes composing a jpeg/png file), its label and some meta information. Then the input pipeline reads a record, decodes it, preprocesses it as needed and feeds it to the network. Of course, you can move the decoding and preprocessing before the generation of the TFRecord dataset and store in the examples the ready-to-feed data, but the size blowup of your dataset will be huge.
The specific preprocessing pipeline is one example what changes between phases (for example, you might have data augmentation in the training pipeline, but not in the others). Of course, there are cases in which these pipelines are the same, but in general this is not true.
About the aside:
“When evaluating, you don't need the gradients and you need a
different input function. “, the only difference (at least in my case)
is the files from which you reading?
In your case that may be. But again, assume you're using data augmentation: You need to disable it (or, better, don't have it at all) during eval and this alters your pipeline.
Question 2: What if I train my model with records and want to inference with just the dense tensors?
This is precisely why you separate the pipeline from the model.
The model takes as input a tensor and operates on it. Whether that tensor is a placeholder or is the output of a subgraph that converts it from an Example to a tensor, that's a detail that belongs to the framework, not to the model itself.
The splitting point is the model input. The model expects a tensor (or, in the more generic case, a dict of name:tensor items) as input and uses that to build its computation graph. Where that input comes from is decided by the input functions, but as long as the output of all input functions has the same interface, one can swap inputs as needed and the model will simply take whatever it gets and use it.
So, to recap, assuming you train/eval with Examples and predict with dense tensors, your train and eval input functions will set up a pipeline that reads examples from somewhere, decodes them into tensors and returns those to the model to use as inputs. Your predict input function, on the other hand, just sets up one placeholder per input of your model and returns them to the model, because it assumes you'll put in the placeholders the data ready to be fed to the network.
Question 3:
You pass the placeholder as a parameter of build_raw_serving_input_receiver_fn, so you choose its name:
tf.estimator.export.build_raw_serving_input_receiver_fn(
{'images':tf.placeholder(tf.float32, [None,28,28,1], name='input_images')})
Question 4:
There was a mistake in the code (I had mixed up two lines), the dict's key should have been input_data (I amended the code above).
The key in the dict has to be the key you use to retrieve the tensor from features in your model_fn. In model_fn the first line is:
my_input = features["input_data"]
hence the key is 'input_data'.
As per the key in receiver_tensor, I'm still not quite sure what role that one has, so my suggestion is try setting a different name than the key in features and check where the name shows up.
Question 5:
I'm not sure I understand, I'll edit this after some clarification