In the Mozilla forum there is some older discussion that suggests that the DeepSpeech architecture does not support phoneme transcription:
Last year there was another discussion thread where a community member did some experiments with phoneme transcription with - to my understanding - mixed results: