Option 1: you can always train without padding if you accept to train separate batches.

See this answer to a simple way of separating batches of equal length: Keras misinterprets training data shape

In this case, all you have to do is to perform the "repeat" operation in another manner, since you don't have the exact length at training time.

So, instead of RepeatVector, you can use this:

import keras.backend as K

def repeatFunction(x):

    #x[0] is (batch,latent_dim)
    #x[1] is inputs: (batch,length,features)

    latent = K.expand_dims(x[0],axis=1) #shape(batch,1,latent_dim)
    inpShapeMaker = K.ones_like(x[1][:,:,:1]) #shape (batch,length,1)

    return latent * inpShapeMaker

#instead of RepeatVector:
Lambda(repeatFunction,output_shape=(None,latent_dim))([encoded,inputs])

Option2 (doesn't smell good): use another masking after RepeatVector.

I tried this, and it works, but we don't get 0's at the end, we get the last value repeated until the end. So, you will have to make a weird padding in your target data, repeating the last step until the end.

Example: target [[[1,2],[5,7]]] will have to be [[[1,2],[5,7],[5,7],[5,7]...]]

This may unbalance your data a lot, I think....

def makePadding(x):

    #x[0] is encoded already repeated  
    #x[1] is inputs    

    #padding = 1 for actual data in inputs, 0 for 0
    padding =  K.cast( K.not_equal(x[1][:,:,:1],0), dtype=K.floatx())
        #assuming you don't have 0 for non-padded data

    #padding repeated for latent_dim
    padding = K.repeat_elements(padding,rep=latent_dim,axis=-1)

    return x[0]*padding

inputs = Input(shape=(timesteps, input_dim))
masked_input = Masking(mask_value=0.0)(inputs)
encoded = LSTM(latent_dim)(masked_input)

decoded = RepeatVector(timesteps)(encoded)
decoded = Lambda(makePadding,output_shape=(timesteps,latent_dim))([decoded,inputs])
decoded = Masking(mask_value=0.0)(decoded)

decoded = LSTM(input_dim, return_sequences=True)(decoded)
sequence_autoencoder = Model(inputs, decoded)
encoder = Model(inputs, encoded)

Option 3 (best): crop the outputs directly from the inputs, this also eliminates the gradients

def cropOutputs(x):

    #x[0] is decoded at the end
    #x[1] is inputs
    #both have the same shape

    #padding = 1 for actual data in inputs, 0 for 0
    padding =  K.cast( K.not_equal(x[1],0), dtype=K.floatx())
        #if you have zeros for non-padded data, they will lose their backpropagation

    return x[0]*padding

....
....

decoded = LSTM(input_dim, return_sequences=True)(decoded)
decoded = Lambda(cropOutputs,output_shape=(timesteps,input_dim))([decoded,inputs])