whisper timestamp issue

System Info

i am testing whisper with notimestamps tag but whisper is generating notimestamps tag in the output.Any idea why this is happening? I am running an audio of 42 secs first 30secs i am getting timestamps but for the later part its generating notimestamps tag

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

I chunked the audio for 30 secs and added padding to the end part and created engine by quantizing whisper for int8 weights only and fp16 weights with largev3 model .The behaviour is same in both cases

Expected behavior

It should predict the timestamps in both audio chunks.

actual behavior

Its only predicting timestamps in first audio segment

additional notes

Is it because of large v3 ?

Start by reproducing the behavior with the 42-second audio, 30-second chunks, end padding, and the largev3 model in both int8-only and fp16-weight quantization modes. Compare timestamp output for both chunks and check whether the issue remains; done means identifying the cause or documenting the missing reproduction details needed to resolve it.