aynesss

'_UserObject' object has no attribute 'summary'

4

Hello, I'm trying to load a r50_1x_sk0/saved_model of pretrained checkpoint-tf2 of simCLRv2, to just fine tuned but I'm coming across this error: '_UserObject' object has no attribute 'summary' I would...

I use TF2.4.1 I would like to finetune the savedModel of pretrained .I write this : ``` target_model_path = "gs://simclr-checkpoints-tf2/simclrv1/pretrain/1x/saved_model/" def create_model(): baseModel = tf.saved_model.load(target_model_path) baseModel= tf.saved_model.save('./saved_model.h5') headModel = baseModel.output...

TF2.4. Here is what I write to begin finetuning: ``` target_model_path = "gs://simclr-checkpoints-tf2/simclrv2/pretrained/r50_2x_sk1/saved_model" saved_model = tf.keras.models.load_model(target_model_path) saved_model.model.save_weights('./weights.h5') simclr_model.load_weights('./weights.h5', by_name=True, skip_mismatch=True) ``` But I found many lines :+1: ``` WARNING:absl:Importing a...

Error when Finetuning: WARNING:absl:Importing a function (__inference___call___15879) with ops with custom gradients. Will likely fail if a gradient is requested.

2

I work with TF version 2.4.1, Here is how I finetune the saved checkpoint of ResNet : ``` ###Model ############################## def create_model(): baseModel = tf.keras.models.load_model("gs://simclr-checkpoints-tf2/simclrv1/pretrain/1x/saved_model/") headModel = baseModel.output layer1 =...

supervised baseline was intialized with weight = None or "Imagenet"

2

I would like to know, to compare result of SimCLR with result of supervised (baseline), I must initialised weights of ResNet50 with None or Imagenet `base = tf.keras.applications.ResNet50(weights=None, include_top=False, input_tensor=tf.keras.Input(shape=(224,...

Trainable=False or Trainable = "True" in Finetune

1

I do a finetune of pretrained network, I would like to know,please, in finetune level, I must make trainable =True or False ? `for layer in baseModel.layers:` `layer.trainable = True`

Linear evaluation notebook

1

can you give us linear evaluation notebook ?

Finetuning must be with data augmentation like SimCLR code ?????

1

To finetune, we must do data augmentation for our dataset ?

Why the code test on only one client ?

1

Like the code below mention : ``` def keras_evaluate(state, round_num): # Take our global model weights and push them back into a Keras model to # use its standard `.evaluate()`...

After changing only the dataset, Here is what I found as results: ``` Round 1 \Eval metrics: loss=0.823, accuracy=0.673 Round 2 \Eval metrics: loss=0.823, accuracy=0.673 Round 3 \Eval metrics: loss=0.823,...