Does a tflite file contain data about the model architecture? A graph that shows what operations there where between the weights and features and biases, what kind of layers (linear or convolutional etc), size of layers, and what activation functions are there in-between the layers?
For example a graph you get with graphviz, that contains all the information, or does a tflite file only contain the final weights of the model after training?
I am working on a project with image style transfer. I wanted to do some research on an existing project, and see what parameters work better. The project I am looking at is here:
https://tfhub.dev/sayakpaul/lite-model/arbitrary-image-stylization-inceptionv3-dynamic-shapes/int8/transfer/1
I can download a tflite file, but I don't know much about these files. If they have the architecture I need, how do I read it?
TFLite flatbuffer files contain the model structure as well. For example, there are a subgraph concept in TFLite, which corresponds to the function concept in the programming language and the operator nodes also represent a graph node, which takes inputs and generates outputs. By using the Netron application, the model architecture can be visualized.
Related
I am using TF-Agents for a custom reinforcement learning problem, where I train a DQN (constructed using DqnAgents from the TF-Agents framework) on some features from my custom environment, and separately use a keras convolutional model to extract these features from images. Now I want to combine these two models into a single model and use transfer learning, where I want to initialize the weights of the first part of the network (images-to-features) as well as the second part which would have been the DQN layers in the previous case.
I am trying to build this combined model using keras.layers and compiling it with the Tf-Agents tf.networks.sequential class to bring it to the necessary form required when passing it to the DqnAgent() class. (Let's call this statement (a)).
I am able to initialize the image feature extractor network's layers with the weights since I saved it as a .h5 file and am able to obtain numpy arrays of the same. So I am able to do the transfer learning for this part.
The problem is with the DQN layers, where I saved the policy from the previous example using the prescribed Tensorflow Saved Model Format (pb) which gives me a folder containing model attributes. However, I am unable to view/extract the weights of my DQN in this way, and the recommended tf.saved_model.load('policy_directory') is not really transparent with respect to what data I can see regarding the policy. If I have to follow the transfer learning as I do in statement (a), I need to extract the weights of my DQN and assign them to the new network. The documentation seems to be quite sparse for this case where transfer learning needs to be applied.
Can anyone help me in this, by explaining how I can extract weights from the Saved Model method (from the pb file)? Or is there a better way to go about this problem?
I have the weights of a custom pre-trained model. I need to extract the representations for different inputs that I pass through the model, across its different layers. What would be the best way of doing this?
I am using TensorFlow 2.1.0 and currently load in the weights of the model using either hub.KerasLayer() or tf.saved_model.load()
Any help would be greatly appreciated! I am very new to TensorFlow and have no choice but to use it since the weights were acquired from another source.
tf.saved_model.load() and its wrapper hub.KerasLayer load both the computation graph and the pre-trained weights. I suppose you're dealing with a TF2-style SavedModel that has its computation packaged in TensorFlow functions. If so, there's no easy way to extract intermediate results from within a function. If possible, you could ask the model creator to provide more outputs, or, if you have the model's Python source, build the model from source and initialize its weights with those from the SavedModel (some plumbing required).
I have 2 Tensorflow Lite models (they are Yolo V2 Tiny models):
Model A) Downloaded from the internet, detects and classifies objects
with 80 classes. The .tflite files weights 44,9mb.
Model B) Trained by myself using Darknet, detects and classifies objects with 52
classes. The .tflite files weights 20,8mb. The model is converted
to TFLite using Darkflow.
However both on a mobile phone and on a computer model B takes 10x more time to predict than model A (even if model B detects within less classes and its file is lighter). Also, models seem to work with input images of size 416x416 and use float numbers.
What could be the reason for model A being faster than model B?
How can I find out why model A is faster?
One of the problems I have is that for model A, since I have not trained it myself, I don't have its .cfg file with the whole setup...
You should try the following two approaches to gain more insight, as the reasons to why a model happens to be slower than expected could be several.
Inspect both networks with a tool like Netron. You can upload your flatbuffer (TF Lite) model file and visualize the network architecture after TF Lite conversion.
There you can see where the difference between the two models lies. If e.g. there happen to be additional Reshape operations or alike in Model B compared to A, that could likely be a reason. To download Netron follow https://github.com/lutzroeder/netron.
Measure the time spent by the model on each of its layers. For this you can use the TF Lite benchmark tool provided directly in the Tensorflow repository.
Check it out here https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/tools/benchmark/README.md.
I want train a custom dataset on FasterRCNN with Mobilenetv1 or v2. I want to use the pre-trained models in tensorflow zoo. But I cant find faster Rcnn model with mobilenet as base extractor. Where can I get it?
I have already tensorflow zoo in github. I have previous used SSD+Mobilenet config for the same. Now I want to compare the results with FasterRCNN and RCNN with Mobilenet.
The official repo has not released Faster RCNN with mobilenet models yet. But if you want you can still use some other models with mobilenet trained on COCO, the process is a bit complicated.
There are two important steps to proceed.
First one is to have corresponding feature extractor class. For Faster RCNN, the models directory already contains faster_rcnn_mobilenet feature extractor implementation so this step is OK. But for R-FCN, you will have to implement the feature extractor class yourself.
Second one is to change tensor names available in the checkpoint. For example, if you use ssd_mobilenet_v1_xxx as checkpoint, then all tensors within mobilenet scope are named as FeatureExtractor/MobilenetV1/XXX while if in the faster_rcnn_mobilenet_v1 model, the tensor names within mobilenet scope are FirstStageFeatureExtractor/MobilenetV1/XXX (and SecondStageFeatureExtractor/MobilenetV1/XXX). So essentially you need to remove FirstStage (as well as SecondStage) in the names of all feature extractor tensors, then these tensors will have exactly the same name as in the checkpoint, and will be correctly restored. If you do this, the function you need to modify is
def restore_map(self,
fine_tune_checkpoint_type='detection',
load_all_detection_checkpoint_vars=False):
in file faster_rcnn_meta_arch.py.
I have used the TensorFlow object detection API to train the SSD Inception model from scratch. The evaluation script shows that the model has learned something and now I want to use the model.
I have looked at the object detection ipynb that can feed single images to a trained model. However, this is for SSD with MobileNet. I have used the following line (after loading the meta graph) to print the tensor names of the TensorFlow model I trained.
print([str(op.name) for op in tf.get_default_graph().get_operations()] )
But it does not contain the same input or output tensor names as in the ipynb. I have also searched through the code, but many functions point toward each other and it is difficult to find what I am looking for.
How can I find the tensor names I need? Or is there another method I do not know about?
To use the graph, you need to freeze/export it, using this provided script. The resulting .pb file will contain the nodes you need. I don't know why it's organized like that, but it is.