taphuynh/whisper-large-en-medical-2607.26-merged

taphuynh/whisper-large-en-medical-2607.26-merged

Fine-tuned from `openai/whisper-large-v3-turbo` with arca-tuner-lite (finetune_medical_en).

• —Base model: openai/whisper-large-v3-turbo
• —Recipe: LoRA adapter
• —Language(s): e, n
• —Run tags: whisper-turbo, medical-en, lora, r32, warmup500
• —Run group: medical-en-warmup

Evaluation

Metrics on the held-out eval split, on the best checkpoint (the one this repo contains — training used early stopping / load_best_model_at_end):

Usage

from peft import PeftModel
from transformers import WhisperForConditionalGeneration, WhisperProcessor

base = WhisperForConditionalGeneration.from_pretrained("openai/whisper-large-v3-turbo")
model = PeftModel.from_pretrained(base, "taphuynh/whisper-large-en-medical-2607.26-merged")
processor = WhisperProcessor.from_pretrained("taphuynh/whisper-large-en-medical-2607.26-merged")

Training data

• —taphuynh/MayoClinic_00001
• —taphuynh/MayoClinic_00002
• —taphuynh/MayoClinic_00003
• —taphuynh/MayoClinic_00004
• —taphuynh/MayoClinic_00005
• —taphuynh/MayoClinic_00006

Training procedure

The exact resolved configuration and environment are in run_card.json in this repo.

Notes & limitations

• —Fine-tuned on domain-specific speech; expect the usual Whisper failure modes (hallucination on silence/noise, degradation far out of domain).
• —This is a PEFT/LoRA adapter — load it on top of the base model above.
• —Not a medical device and not for clinical decision-making.