# define number of features

I try to understand my time series data for my ANN for a regression problem. I have temperature values measured with a sampling rate of 10.000 values / 1 second. So I measure in 1 second 10.000 different temperature values. This I repeat with certain time intervals. Do I have than 10.000 different features or only one feature as input dimension for my lstm neural network ? I want to shape my data for a lstm encoder decoder.

How I understand your data is a single feature of temperature. LSTM works with RNN time series like (batch_size, timesteps, feature). In your case it is (none, 10000, 1). Forget about the first one, it's technical.