f""" Generates Whisper models and publishes them to hf.co/{MODEL_REPO_ID}

Download WhisperKit model files from Hugging Face Hub (only the files needed for `self.whisper_version`)

Download whisper.cpp model files from Hugging Face Hub (only the ones needed for `self.whisper_version`)

Temporary function to reformat the original Whisper tokenizer

WhisperTextDecoder emits logits over a token vocabulary. The function used to quantify output change is KL divergence (lower the better)

WhisperEncoder emits embeddings. The function used to quantify output change is negative PSNR (lower the better)

Evaluate a single audio folder with whisper_pipeline

Shapes: Inputs: input_embeds: (batch_size, embed_dim, 1, encoder_seq_len) Outputs:

Shapes: Inputs: melspectrogram_features: (batch_size, num_mel_bins, 1, encoder_seq_len * 2) Outpu

Shapes: Inputs: input_embeds: (batch_size, embed_dim, 1, q_seq_len=1) key_cache:

Shapes: Inputs: input_ids: (batch, ) cache_length: (batch, )

Look-up key-value cache prefill values for a given task spec. Args: task: 0->transcribe, 1->translate language:

Apply STFT on input (audio) data

Compute RMSE for audio waveform

Initialize WhisperTextDecoder and a test mini-batch of random data

Initialize WhisperAudioEncoder and obtain a test mini-batch of random data

Replace any other markers, symbols, punctuations with a space, keeping diacritics

Add metadata to Core ML models for traceability

Summarize the profile job results from AI Hub

Test AI Hub compilation and inference job results against local PyTorch test results

Test forward pass functionality and correctness of PyTorch models

Coverage: - torch2torch parity transformers.models.whisper.modeling_whisper.WhisperDecoder and whisperkit.text_decoder.WhisperTextDeco

Coverage: - torch2torch parity transformers.models.whisper.modeling_whisper.WhisperEncoder and argmax.ane.whisper.WhisperEncoder

Coverage: - Build a WhisperDecoderContextPrefill module - Use a test WhisperDecoder instance to compute all valid context pre

Transcribe an audio file using the WhisperKit CLI

Transcribe an audio file using the whisper.cpp CLI

Transcribe an audio file using MLX

Transcribe an audio file using PyTorch OpenAI Whisper

Transcribe an audio file using the Apple Speech Analyzer CLI