test_hotwords: ensures hotwords argument is respected and working in both sequential and batched
test_transcribe_signature: ensures WhisperModel.Transcribe and BatchedInferencePipeline.transcribe have the same signature so that they can be used interchangeably with the exclusion of batch_size argument
test_monotonic_timestamps: ensures word and segment timestamps are always increasing and do not overlap
moved tokenizer tests to a separate file and added test_split_on_unicode which is ported from OpenAI Whisper
Added the following tests:
test_hotwords: ensureshotwordsargument is respected and working in both sequential and batchedtest_transcribe_signature: ensuresWhisperModel.TranscribeandBatchedInferencePipeline.transcribehave the same signature so that they can be used interchangeably with the exclusion ofbatch_sizeargumenttest_monotonic_timestamps: ensures word and segment timestamps are always increasing and do not overlaptest_split_on_unicodewhich is ported from OpenAI Whisper