app.py

import gradio as gr
import torch
from transformers import pipeline, AutoTokenizer
from nemo.collections.asr.models import EncDecMultiTaskModel

# load model
canary_model = EncDecMultiTaskModel.from_pretrained('nvidia/canary-1b')

# update dcode params
decode_cfg = canary_model.cfg.decoding
decode_cfg.beam.beam_size = 1
canary_model.change_decoding_strategy(decode_cfg)

pipe = pipeline(
    "automatic-speech-recognition", 
    model="nvidia/canary-1b"
)

# pipe = pipeline(
#     "text-generation", 
#     model="QuantFactory/Meta-Llama-3-8B-Instruct-GGUF", 
#     model_kwargs={"torch_dtype": torch.bfloat16}, 
#     device_map="auto"
# )

gr.Interface.from_pipeline(pipe,
                           title="ASR",
                           description="Using pipeline with Canary-1B",
                           ).launch(inbrowser=True)