transcribe same audio but got different result

[genausz]

When i use the pretrained model (small) and infer the same audio file, but got different result. Anyone meet the same problem as me?

[glangford]

See in particular
#308 (comment)

and the following discussions:

• Different transcribe results using same model to transcribe the same audio #81
• Same model + different GPUs = different transcriptions? #274
• Transcriptions differ between Windows CPU and Debian GPU #308