From 18a805b377683c1d8944e31217cad0797f3d34da Mon Sep 17 00:00:00 2001 From: www-data Date: Mon, 5 Aug 2024 03:15:58 +0000 Subject: [PATCH] add assets identified by bot --- assets/singapore.yaml | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) diff --git a/assets/singapore.yaml b/assets/singapore.yaml index b8ec7b6e..fdcb9044 100644 --- a/assets/singapore.yaml +++ b/assets/singapore.yaml @@ -47,3 +47,24 @@ prohibited_uses: '' monitoring: unknown feedback: https://huggingface.co/OrionZheng/openmoe-base/discussions +- type: model + name: Whisper Medusa + organization: Aiola + description: Whisper Medusa is an advanced encoder-decoder model for speech transcription and translation, which processes audio through encoding and decoding stages. It is built on a version of the Whisper model, with improvements for prediction speed. Optimization strategies such as Faster-Whisper and Speculative Decoding enhance Medusa's performance. + created_date: Unknown + url: https://huggingface.co/aiola/whisper-medusa-v1 + model_card: https://huggingface.co/aiola/whisper-medusa-v1 + modality: Audio; Text + analysis: The model was trained and evaluated on the LibriSpeech dataset, demonstrating improvements in speed. + size: Unknown + dependencies: [LibriSpeech dataset, Whisper Model] + training_emissions: Unknown + training_time: Unknown + training_hardware: Unknown + quality_control: The model was optimized and validated on English, and its performance has been enhanced by adopting different strategies such as Faster-Whisper and Speculative Decoding. + access: Unknown + license: Unknown + intended_uses: Whisper Medusa can be used for speech translation purposes, specifically for translating English audio. + prohibited_uses: It is not recommended to use the model for languages other than English as it has been optimized for English. + monitoring: Unknown + feedback: Any problems with the model should be reported by following the instructions provided in the Whisper Medusa repository README.