Request for pre-TFLite ACFT-KO large-v3-turbo checkpoint for CTranslate2 testing
Hi, thank you for sharing the ACFT-KO models.
I am testing faster-whisper with large-v3-turbo on CPU using short Korean telephone recordings, typically around 3–4 seconds long.
I experimented with reducing the encoder input length instead of padding every utterance to 30 seconds. This significantly reduced processing time, but Korean recognition accuracy also dropped substantially. I would like to test whether ACFT-KO can maintain better accuracy with shorter inputs.
Would you be willing to share the ACFT-KO large-v3-turbo checkpoint before TFLite conversion, preferably in Hugging Face Transformers format, including the weights (such as safetensors), config.json, and any required preprocessing or tokenizer configuration?
My goal is to try converting it to CTranslate2 and compare transcription accuracy and latency on the same recordings using 5-second and 10-second encoder windows.
I am specifically looking for the fine-tuned checkpoint before TFLite conversion, rather than an FP32 TFLite export. Any notes on custom model code or changes needed for short-context inference would also be helpful.
If sharing the checkpoint is not currently possible, are there any plans to release it?
Thank you for your time and work on this project.
I apologize, but I am unable to provide the files.
They were deleted due to storage limitations in the workspace.
Since the training process is not difficult, I would recommend trying it out yourself.