injilashah/whisper-small-finetuned-kashmir-words.2

Direct Use

from transformers import WhisperForConditionalGeneration, WhisperProcessor from transformers import pipeline

model = WhisperForConditionalGeneration.from_pretrained("injilashah/whisper-small-finetuned-kashmir-words.2") processor = WhisperProcessor.from_pretrained("injilashah/whisper-small-finetuned-kashmir-words.2")

import gradio as gr

pipe = pipeline(model="from transformers import pipeline import gradio as gr

pipe = pipeline(model="sanchit-gandhi/whisper-small-hi")

def transcribe(audio): text = pipe(audio)["text"] return text

iface = gr.Interface( fn=transcribe, inputs=gr.Audio(source="microphone"), outputs="text", title="ASR", description="Realtime demo for ENGLISH speech recognition using fine-tuned Whisper small model.", )

iface.launch()

Training Hyperparameters

per_device_train_batch_size=2, # increase
gradient_accumulation_steps=1,  # increased by 2x for every 2x decrease in batch size
learning_rate=1e-5,
#max_steps=50,
#evaluation_strategy="steps",
num_train_epochs=7,

• Training regime: <!--fp16 mixed precision