Please direct future correspondence to the LAMMPS mailing list, copied above.
You should use fix deform. That fix allows you to hold the voume fixed under uniaxial stretching. alternative, you could try combining that with a Berendsen or NPH barostat on the lateral dimensions. No matter which method you choose, trying to attach any meaning to the observed temperature rise may prove difficult. If you are interested in macroscopic dissipative processes, these will not be capturedwith a few thousand atoms, and may be not even with a few billion.
Aidan