File size: 2,823 Bytes
5036278 f9c002c 5036278 f9c002c 5036278 f9c002c 2c77d1a f9c002c 5036278 f9c002c 5036278 f9c002c 5036278 f9c002c 5036278 2c77d1a 5036278 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
import gradio as gr
import torch
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig, TextStreamer, AutoModelForSpeechSeq2Seq
# Whisper Model Optimization
WHISPER_MODEL = "openai/whisper-large-v3"
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
whisper_quant_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_use_double_quant=True,
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_quant_type="nf4"
)
whisper_model = AutoModelForSpeechSeq2Seq.from_pretrained(
WHISPER_MODEL,
torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32, # Use fp16 if GPU available
device_map="auto"
)
whisper_tokenizer = AutoTokenizer.from_pretrained(WHISPER_MODEL)
transcriber = pipeline(
task="automatic-speech-recognition",
model=whisper_model,
tokenizer=whisper_tokenizer,
chunk_length_s=30,
device=DEVICE
)
# LLaMA Model Optimization
LLAMA = "meta-llama/Llama-2-7b-chat-hf"
llama_quant_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_use_double_quant=True,
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_quant_type="nf4"
)
tokenizer = AutoTokenizer.from_pretrained(LLAMA)
tokenizer.pad_token = tokenizer.eos_token
model = AutoModelForCausalLM.from_pretrained(
LLAMA,
torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
device_map="auto"
)
# Function to Transcribe & Generate Minutes
def process_audio(audio_file):
if audio_file is None:
return "Error: No audio provided!"
# Transcribe audio
transcript = transcriber(audio_file)["text"]
# Generate meeting minutes
system_message = "You are an assistant that produces minutes of meetings from transcripts, with summary, key discussion points, takeaways and action items with owners, in markdown."
user_prompt = f"Below is an extract transcript of a Denver council meeting. Please write minutes in markdown, including a summary with attendees, location and date; discussion points; takeaways; and action items with owners.\n{transcript}"
messages = [
{"role": "system", "content": system_message},
{"role": "user", "content": user_prompt}
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(DEVICE)
streamer = TextStreamer(tokenizer)
outputs = model.generate(inputs, max_new_tokens=2000, streamer=streamer)
return tokenizer.decode(outputs[0], skip_special_tokens=True)
# Gradio Interface
interface = gr.Interface(
fn=process_audio,
inputs=gr.Audio(sources=["upload", "microphone"], type="filepath"),
outputs="text",
title="Meeting Minutes Generator",
description="Upload or record an audio file to get structured meeting minutes in Markdown.",
)
# Launch App
interface.launch()
|